Lockheed’s costs on Lot 20, the next batch of F-35 jets, outran the government’s authorized funding during the quarter, according to its earnings release.

Lot 20 runs on an early-funding contract that lets work start before the full deal is signed. In plain terms, Lockheed is paying out of pocket for jets Washington hasn’t signed for.

The company warned that its results and cash flow will suffer without more funding. That warning sat below the guidance table in a release packed with good news. I’d put it near the top.

The good news was real. Sales rose 11% to $20.7 billion, according to the release, and Lockheed raised its full-year outlook for the second straight quarter.

Also read: Lockheed bets on ‘revolutionary’ Patriot to stop faster missiles

The F-35 has a 10-year path but no Lot 20 contract

The timing makes the gap stranger. On the earnings call, CEO Jim Taiclet said that a multiyear framework the Pentagon announced that morning sets a path to 850 U.S. jets over 10 years.

Yet the Lot 20 contract remains unsigned. Chief Financial Officer Evan Scott said a deal this year would be upside, but it “is not counted on in our guide.”

Investor relations head Mark Kvasnak said Lockheed pushed its assumption for the next F-35 award into 2027. To me, that is the real tell: Lockheed now plans to carry the gap through year-end.

The strain already shows. Contract assets track work Lockheed has done but cannot yet bill. They rose to $16.9 billion from $13 billion at the end of 2025, according to the release. Not all of it is F-35 work, but it is cash still waiting to be collected.

Lockheed’s missile deals are built to protect its cash

The squeeze looks sharper next to Lockheed’s missile business. Missiles and Fire Control backlog has roughly tripled since the end of 2025 to $142.5 billion, according to the release.

That unit now holds about half of Lockheed’s record $292 billion backlog. Aeronautics, home of the F-35, saw its backlog shrink slightly. Growth is moving from jets to missiles, a shift UBS flagged last month.

Funding is the bigger difference. Scott said the government keeps Lockheed cash neutral while it ramps production under the missile frameworks, which are still being finalized. The F-35 has no such cushion today. Lockheed’s newest growth engine protects its cash, while its best-known program asks it to front money.

Lockheed stock rallied past analysts’ latest targets

Investors cared more about the beat. Lockheed shares rose nearly 7% on the day of the report. They still sit about 15% below the 52-week high of $692. The 52-week low is $437.25, a wide range for a defense blue chip.

The valuation looks fair, not cheap. Lockheed trades at nearly 19 times forward earnings. That sits below its trailing multiple of about 21, so analysts expect profits to grow.

The 23 analysts tracked by StockAnalysis rate the stock a Buy on average, with a mean target of $618.86. That implies only about 5% upside after the rally. The latest calls, all made before the report, leaned cautious:

• Hold from RBC Capital’s Kenneth Herbert, target cut to $550 from $600, Oct. 16
• Hold from Bernstein’s Douglas Harned, target cut to $564 from $629, Oct. 14
• Buy from Vertical Research’s Robert Stallard, an upgrade, Oct. 14
• Sell from Goldman Sachs’ Noah Poponak, target cut to $486 from $503, Oct. 12
• Hold from TD Cowen’s Gautam Khanna, target cut to $570 from $600, Oct. 9

The rally carried the stock past all four of those targets. Analysts must now chase the shares higher or argue they ran too far.

Shareholders are getting less back, too.

Lockheed bought back no stock in the first nine months of 2026, versus $2.25 billion a year earlier, the release indicated. Its risk language names presidential executive orders as a factor in buybacks, after a January order targeted defense contractor payouts.

More Defense stocks:

• UBS sees a structural shift building at Lockheed Martin
• Boeing lands key role in future of U.S. air power
• Two defense stocks just got a multiyear vote of confidence

Defense contractors are carrying more of Washington’s risk

Lockheed isn’t alone. RTX Corp. (RTX) and Northrop Grumman Corp. (NOC) also raised their forecasts the same week, according to Reuters.

The pattern matters more than any single beat. Washington wants faster production, smaller payouts, and contractors willing to spend before deals are final. Lockheed’s disclosure shows what that looks like on a balance sheet.

I’d watch cash more closely than backlog from here. The first number to check is free cash flow, the cash left after capital spending. Lockheed now expects $7.8 billion to $8 billion for 2026, the release confirmed. Falling short would show the F-35 gap starting to bite.

If Lot 20 is still unsigned at the fourth-quarter report, investors will learn how long Lockheed can fund its flagship jet alone. In this defense boom, winning orders may prove easier than getting paid on time.

More Stocks News

• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off
• JPMorgan just more than doubled its SpaceX price target

Retirement savers feel that split most. A portfolio that is meant to last decades has different needs than a trade meant to pay off next quarter, and the debate comes down to Alphabet versus Apple. Both report earnings this month, so the timing is hard to ignore.

How the Alphabet and Apple gap opened up

Alphabet has lost about $860 billion in market value since its peak, close to Walmart’s $937.1 billion, according to 24/7 Wall St., citing an October 17 post by Charlie Bilello of Creative Planning that put Apple at an all-time high and Alphabet 17% below.

Apple is worth $4.95 trillion and Alphabet is worth $4.21 trillion. Apple leads the Magnificent Seven this year at 25.26%, against 6.47% for Alphabet, which has also fallen 4.84% in the past month.

Also read: Apple just made a surprising move with a popular product

Apple’s run is real. The stock touched an all-time high of $345.34 in September, weeks after John Ternus became CEO on September 1, and was up 25% for the year with a market value just under $5 trillion, according to TheStreet. The iPhone 18 lineup went on sale September 19 with prices raised by $100, and the $1,999 foldable iPhone Duo ships October 23.

Alphabet went the other way after July’s earnings. Revenue beat estimates and Google Cloud grew 82% to $24.8 billion, but the company lifted its 2026 capex outlook to $195 billion to $205 billion, up from $180 billion to $190 billion, and the stock sank.

The next test arrives on October 28. Alphabet brought in $119.8 billion in second-quarter sales, up 24%, yet Bank of America expects about $199 billion of capex this year and $301 billion in 2027, plus a $12.4 billion free cash flow loss in the third quarter. The bank also expects Search growth to cool from 17% to 14% while Cloud accelerates to 93%.

What Alphabet’s spending is doing to its cash

Alphabet spent $44.9 billion on capital expenditures in the second quarter, double the year-earlier amount, and posted its first negative free cash flow since going public in 2004, according to Axios, citing FactSet. Net income got a lift from roughly $99 billion in net gains on equity securities.

Alphabet was the first hyperscaler to report this season, Axios noted, so investors read its spending as a signal for the group.

To pay for the buildout, Alphabet announced plans to raise $80 billion in equity in June, including a $10 billion private placement to Berkshire Hathaway at $351.81 per Class A share and $348.20 per Class C share, Bloomberg reported. The company said demand for its AI services exceeds the supply it can deliver.

Plainly, Alphabet is funding growth partly with new shares rather than only cash from operations. For existing holders, that means dilution while the spending keeps climbing. A company known for generating its own cash substantially now turns to the market for funding.

That spending picture does not cancel out what Alphabet is actually building. Revenue climbed 24% last quarter and Cloud jumped 82%, according to Reuters. Investors are not questioning the growth. They are questioning when the bills stop growing faster than the returns do.

How Apple stays out of the spending race

Apple’s June quarter showed what its model delivers. iPhone sales jumped 22% year over year, according to TheStreet. Tim Cook said Apple had underestimated demand, while calling on-device AI a competitive weapon. The redesigned Siri was still in public beta and not yet a proven source of new revenue.

Apple is not skipping AI so much as sourcing it. It overhauled Siri in part by tapping Google’s AI technology behind the scenes. That arrangement helps explain why Apple’s capital spending sits far below Alphabet’s. Analysts say a successful Siri revamp by year-end could put Apple on even footing with Google’s Gemini assistant.

More Apple:

• Apple’s biggest AI advantage may face new privacy challenge
• Apple secret new device is a challenge to Amazon
• Bank of America warns Apple investors because of Meta’s Muse

Wall Street watched the strategy at the June 8 developer conference, where Apple unveiled Siri AI and deeper AI features across its devices. Bank of America had called the event an important marker for Apple’s AI future.

Apple also sells the hardware AI builders need. Meta’s newest model was optimised for Apple’s M5 Ultra chip. Apple shares slipped to $342.00 on October 15. Developers building AI tools need serious local hardware, so every purchase is revenue Apple books today.

What it means for a retirement portfolio

The two stocks have returned almost the same over twelve months, nearly 30% each. The difference is the path. Alphabet runs through larger bills, while Apple runs through iPhone demand and a valuation near 39 times earnings. A premium price leaves much less room for error.

A saver near retirement leans on the dependable payouts. Apple pays a $0.27 quarterly dividend. The dividend was raised 4% in May alongside a $100 billion buyback authorization. It also guided September-quarter revenue growth of 9% to 11%. That kind of cash return supports a steadier profile than a company spending beyond its free cash flow.

For a retirement account, the gap between these two companies is not just about AI. It is about what kind of risk you are paid to take. Both report within a day of each other this month. When the numbers land, the real question is whether the story each stock is telling still holds.

Related: The Fed has a new warning for Wall Street and the stock market

Two-thirds of the $41 billion in capital expenditures in the latest quarter went to shorter-lived assets such as central processing units and graphics processing units. These are the costly devices that lie behind solutions like Copilot and Azure.

Microsoft now wants to retain a larger portion of that expenditure inside the company. With multi-gigawatt deployments starting in 2027, Microsoft will use its Maia chips to run AI inference across its products and services.

There is an obvious potential loser if Microsoft succeeds: Nvidia (NVDA).

One of the large cloud firms purchasing Nvidia chips to develop AI infrastructure is Microsoft. Microsoft may not need to run all of its workloads on someone else’s processors in order to transfer them cheaply onto Maia.

There is just one issue.

Microsoft’s proprietary processors are still “nowhere close” to the more advanced technology of Nvidia or Alphabet (GOOGL), according to Bloomberg Intelligence analyst Mandeep Singh.

This creates a test worth billions of dollars.

Nvidia doesn’t have to be replaced everywhere by Microsoft. Maia only has to become proficient at the AI tasks that Microsoft does millions of times every day.

Microsoft has a powerful reason to build its own chips

Long after a model has been trained, artificial intelligence may be costly.

Computers must provide the answer each time someone asks Copilot a query, requests code from GitHub Copilot, or ultimately dispatches an AI agent to do a job. That process is called inference.

Because Microsoft runs such services on a massive scale, even modest efficiency improvements may have a significant impact when replicated over millions of users.

Microsoft said its current Maia 200 chip is already serving Copilot workloads. Maia 300, expected to enter production deployments in 2027, is designed to become the company’s primary inference system.

Microsoft also said the Maia 300 can produce seven times the output of the Maia 200 at full rack size and 40% more AI per megawatt than top GPUs. In other words, the company believes it can get more AI work from the infrastructure and power it currently pays for.

Because of the massive expenditure, that is important.

Microsoft added nearly $116 billion of property and equipment during fiscal 2026, compared with about $65 billion a year earlier.

In the June quarter alone, capital expenditures reached $41 billion.

It is not necessary for Microsoft to create a semiconductor that is superior than Nvidia for every use if Maia enables the company to offer Copilot and other repetitive AI tasks at a lower cost.

It could simply lessen the quantity of costly external processors Microsoft need for tasks that its own silicon is capable of handling.

Nvidia is not disappearing from Microsoft data centers

The catch is that making an AI chip and replacing Nvidia are two completely unique things.

Singh contended that when it comes to proprietary AI silicon, Microsoft is still far behind both Nvidia and Google.

Google is especially significant, since it demonstrates what Microsoft is attempting to develop.

More Microsoft:

• Microsoft’s $665 target hinges on a new AI advantage
• Meta and Microsoft just sent employees a memo about Anthropic
• Microsoft just sent a strong message to Anthropic

Alongside external chips, Alphabet has spent years creating its Tensor Processing Units. This gives Google an additional way to manage the price and efficiency of the AI operating in its cloud.

With Maia, Microsoft is attempting to establish a comparable edge.

However, Nvidia has spent years creating the software environment that developers use in addition to CPUs. This makes replacing Nvidia an undertaking that involves more than making competitive silicon.

Additionally, Microsoft has no justification for giving up on Nvidia entirely.

Different hardware is needed for different AI tasks. Serving Copilot replies to clients on a regular basis may need different computer power than training massive frontier models.

Therefore, whether Microsoft continues to purchase Nvidia is not the crucial point. It’s whether Microsoft can alter the economics by moving enough high-volume work to Maia.

Microsoft’s AI bill makes Maia worth watching

The most recent results from Microsoft demonstrate why investors should be concerned.

The tech giant’s 2026 annual report showed that Azure revenue surpassed $100 billion for fiscal 2026, rising 41%, while Microsoft Cloud revenue reached more than $214 billion.

AI is contributing to that company’s growth, but Microsoft is investing heavily to provide the processing power that consumers want.

The June quarter’s $41 billion capital-expenditure bill was larger than Microsoft’s $35.8 billion quarterly net income.

That doesn’t mean the spending is ill-advised. Microsoft is investing because demand is strong. But the larger the infrastructure bill becomes, the more valuable efficiency becomes. That is where Maia could matter.

Microsoft may increase the economics of every encounter without waiting for Nvidia to reduce pricing if it can execute more Copilot requests, coding activities, and AI-agent workloads using processors built around its own software.

And for this to matter, Microsoft doesn’t need to defeat Nvidia. Even a small fraction of an AI infrastructure operation could save tens of billions of dollars.

For this reason, what might otherwise seem to be just another piece of arcane semiconductor news should be of interest to a Main Street investor.

Microsoft is more than just a chip manufacturer. It is attempting to manage one of the largest costs associated with its AI business.

Microsoft may retain a larger portion of Copilot and Azure’s earnings if Maia is successful. If it doesn’t, Microsoft may continue spending heavily on its silicon while Nvidia keeps collecting on the AI boom anyway.

Related: Morgan Stanley sees new hurdle for Microsoft stock

When I wrote about SpaceX’s IPO buyers in September, the stock had already split them into very different outcomes depending on when they got in. Every new forecast since has pulled that crowd in a different direction.

The company handed the bulls plenty of fuel this week. SpaceX’s revenue jumped 161% to $14.6 billion in the third quarter, and it swung to a $4.4 billion profit from a $1.1 billion loss, according to Quartz.

Elon Musk said on the Oct. 20 call that he is “fairly confident” SpaceX can add about 10 gigawatts of computing capacity in 2027.

That is the kind of number that makes Wall Street rewrite its spreadsheets overnight.

JPMorgan analyst Doug Anmuth raised his SpaceX price target to $500 from $240 on Oct. 21 and kept an Overweight rating, Investing.com reported. The new target implies 238% upside from the stock’s $147.79 close on Oct. 20.

What JPMorgan’s $500 target is built on

Anmuth more than doubled his 2027 and 2028 estimates. JPMorgan now models 10 gigawatts of compute by the end of 2027 and total 2027 revenue of $254 billion, the report said.

That points to an annual revenue run rate of about $400 billion by the end of 2027. Management’s own range is far higher, at $500 billion to $1 trillion, Investing.com reported.

Related: Vivian Tu’s IPO warning lands as SpaceX buyers split by $1,851

Most of that growth has to come from artificial intelligence. SpaceX’s Super Intelligence segment brought in $7.6 billion in the third quarter, up 830%, while its connectivity business grew 69% to $5.2 billion, Quartz reported.

The target rests on 20 times JPMorgan’s 2028 earnings estimate of $24.41 a share. I ran that math, and it works out to about $488, so the published $500 either rounds up or uses a slightly richer multiple than the summary describes.

Shares had already moved before the note landed. The stock rose 6.6% after hours to $157.59 on Oct. 20, according to Investing.com, as revenue beat the $12.87 billion that analysts expected.

How the rest of Wall Street sized up the quarter

JPMorgan is far out ahead of its peers, and the size of that gap is the real story in this week’s notes.

Here is where other banks landed after the report, according to Investing.com:

• Morgan Stanley: Kept an Overweight rating and a $300 target while raising its AI revenue forecasts, Investing.com reported.
• UBS: Raised its target to $230, citing strength in AI compute leasing, the same report showed.
• RBC Capital: Reiterated Outperform with a $225 target, pointing to a favorable long-term outlook.
• Mizuho: Raised its target to $205 with an Outperform rating after revenue and adjusted EBITDA beat its own forecasts.
• Stifel: Reiterated a Buy rating and a $190 target.

By my math, JPMorgan’s target is 67% above Morgan Stanley’s, the next highest, and more than 2.6 times Stifel’s. A spread that wide tells you analysts agree the quarter was strong but disagree sharply on what the company is worth.

Goldman Sachs reworked its own numbers ahead of the report, as TheStreet covered in Goldman’s pre-earnings revamp of its SpaceX estimates.

Why the bull case needs revenue to nearly double again

Here is the part I think matters most for anyone holding the stock. SpaceX guided to first-quarter 2027 revenue of $33 billion to $37 billion, according to the Motley Fool transcript.

Multiply that range by four, and you get $132 billion to $148 billion for the year. JPMorgan’s $254 billion forecast needs an average of about $63.5 billion a quarter, close to double the midpoint of that first-quarter guide.

More SpaceX:

• Elon Musk wants every American kid to own a SpaceX share
• SpaceX has a plan to make your phone carrier obsolete
• Elon Musk doubles down on the Roadster with SpaceX thrusters

The backlog covers only part of that gap. SpaceX ended the quarter with $63 billion in backlog, about $42 billion of which it expects to turn into revenue within 12 months, according to the transcript. By my math, that $42 billion equals roughly 17% of JPMorgan’s 2027 revenue forecast.

The spending bill is just as steep. SpaceX spent $19.6 billion on capital projects in the third quarter and guided to $35 billion to $40 billion in the fourth.

That means the stock’s next big moves may come from execution updates rather than earnings beats. Watch the compute capacity SpaceX reports each quarter, because JPMorgan’s target lives or dies on that 10-gigawatt figure.

Anmuth acknowledged the risk. Hitting his numbers requires building out the compute, raising substantial capital and a demand environment that stays strong, Investing.com reported.

There is also a corporate wild card. Musk said SpaceX and Tesla (TSLA) have formed special committees to study a possible merger, which would need shareholder approval, and JPMorgan flagged that regulators could weigh in, Investing.com reported.

Much will likely depend, however, on how quickly the new capacity comes online. Musk said the shortage of AI compute “is going to remain for a long time.” In my analysis, that scarcity is what supports the pricing JPMorgan is counting on.

For your portfolio, a 238% upside call is a reason to understand the assumptions, not a reason to chase. If you own SpaceX, or plan to buy shares as a gift after Musk’s pitch to put a SpaceX share in every American kid’s hands, it helps to know which bank’s math you’re really betting on.

Elon Musk has given investors one more number to watch. Starship Flight 15 is expected next month with a catch attempt, and Musk said success “is not guaranteed.”

Related: JPMorgan sees 67% upside in a hidden SpaceX bet

One crucial item that the new stock will deny them is a vote.

Starting in January 2027, Meta will issue publicly traded Class C shares without voting rights. The economic effect of a two-for-one stock split will be one Class C share per Class A and Class B share.

There is more to the move than just making Meta shares seem less expensive.

As the corporation invests enormous sums of money in artificial intelligence, Meta claims that the new shares may be used to raise money, pay staff, and cover acquisitions.

Also Read: Cathie Wood sells $16 million of surging tech stock

However, Meta is able to raise that capital without diminishing Mark Zuckerberg’s power as CEO by issuing non-voting shares.

Because of these factors, the new stock provides an exceptionally clear view into Meta’s goals.

As it purchases processors, constructs data centers, and creates AI products, the corporation anticipates spending between $135 billion and $145 billion on capital expenditures in 2026.

In order to continue spending, Meta needs more financial freedom.

It does not want Zuckerberg to give up power in order to get it.

Meta’s new shares are not free money

The first thing that regular investors need to realize is that Meta is not going to instantly double everyone’s wealth.

The intended distribution would ultimately leave an investor with 10 Class A shares and ten Class C shares if they held ten Meta Class A shares.

However, Meta anticipates that the market value of its current Class A shares will change to reflect a two-for-one stock split.

In other words, investors receive additional shares, but the company’s economic value does not immediately double.

Voting power makes a difference.

Voting rights are attached to Meta’s current publicly listed Class A shares. Despite holding less than a majority of the company’s economic worth, Zuckerberg has been able to keep control of it because of his Class B shares, which have far more voting power.

There are no votes on the new Class C shares.

Meta stressed that no shareholder’s voting power would be altered by the distribution.

That becomes especially important after the initial stock dividend.

Thereafter, Meta will have a currency that it may provide to staff members, possible acquisition targets, or new investors without giving them more power over business choices.

Reuters reported that Meta designed the structure partly to finance growth while preserving Zuckerberg’s control.

That’s the real story for stockholders.

Meta is doing more than just dividing its shares.

To fund the next phase of its AI evolution, it is altering its toolbox.

Meta’s AI bill explains why it wants more flexibility

Meta has a lot of money.

Cash, cash equivalents, and marketable securities totaled $82.45 billion at the end of September.

However, its expenditures are increasing at an incredibly rapid rate.

Purchases of property and equipment increased from $18.83 billion to $34.91 billion in the third quarter alone.

After a year of positive free cash flow of $10.63 billion, free cash flow fell to negative $2.12 billion.

Meta projects $135 billion to $145 billion in capital expenditures for the whole year.

More Tech:

• Meta and Microsoft just sent employees a memo about Anthropic
• Microsoft cutting off popular software with no extension Oct. 13
• Apple’s biggest AI advantage may face new privacy challenge

Artificial intelligence is becoming more and more connected to expenditure.

While attempting to transform its new Muse AI agent into yet another massive consumer platform, Meta is constructing data centers, purchasing cutting-edge CPUs, and creating its own models.

In its first six weeks, over 10 million users in the US and Canada registered for Muse accounts, according to Zuckerberg.

He thinks Muse will someday have billions of users.

However, providing for such a large number of users is costly.

Reuters reported that Zuckerberg acknowledged Muse costs more to operate than many products Meta has launched in the past, prompting the company to think about monetizing it sooner.

This increases the value of having access to more capital.

Zuckerberg keeps control while Meta keeps spending

The Class C shares become very intriguing at this point.

By issuing a lot of voting stock, a traditional corporation may progressively reduce the power of current shareholders.

That may ultimately jeopardize the founder’s authority over a firm they own.

Meta’s suggested structure mostly avoids that issue.

The company’s board said one reason for creating Class C shares was to ensure Zuckerberg continues leading Meta into its next era.

Meta has also entered into what it calls a Founder Alignment Agreement with Zuckerberg covering his Class C shares and potential future stock distributions.

This implies that Meta might use shares with economic value to pay staff or fund acquisitions more often without generating fresh votes that would compete with Zuckerberg’s.

For regular investors, there is a trade-off.

The arrangement provides Meta with more freedom to pursue AI opportunities without necessarily relying solely on money or debt.

However, investors purchasing the new Class C shares will control a portion of Meta without having any influence over the company’s governance.

When one CEO already has significant control over the course of the business, that difference is important.

Meta is preparing for an expensive AI future

A large portion of Meta’s goal can be financed by its current operations.

Third-quarter revenue rose 26% to $64.6 billion, while advertising impressions increased 10% and average advertising prices climbed 14%.

The funding for Meta’s AI revolution is still coming from that advertising machine.

However, it has become more difficult to overlook the price of competing.

During the quarter, Meta invested about $35 billion in capital projects, and it anticipates further aggressive spending in 2027.

It now has an additional means of funding that growth, thanks to the new Class C shares.

This makes it more than just an odd stock split.

Meta is preparing its financing structure for a future where artificial intelligence will require years of massive investment before products like Muse can become significant stand-alone companies.

It has built that framework around a single, crucial need.

Meta can issue more shares.

It can use them to buy companies.

It can use them to attract employees.

It can potentially use them to raise enormous amounts of capital.

But those shares do not have to make Mark Zuckerberg any less powerful.

Related: Meta Platforms, Inc. Q3 2026 Earnings: Recap of $META Earnings Call, Stock Forecast

For years, that was the comfortable assumption behind the software that runs the internet. Bugs existed, but finding the dangerous ones took rare skill and a lot of time.

That buffer has quietly disappeared. Artificial intelligence is now finding old flaws faster than humans can patch them, and attackers are using the same tools to turn fixes into weapons.

If you have a website for your business, your side hustle or your church, this one lands on you too.

The fix is free and already out. The hard part is getting it onto millions of servers before attackers get there first, and that race started earlier than most people realize.

It also explains why Wikipedia, the Port of Rotterdam, and France’s Interior Ministry all pulled sites offline on Oct. 21.

I spent the morning going through the disclosures, and one detail stands out more than any outage list.

Amazon (AMZN) says its security team caught someone trying to exploit the newly discovered flaws in PHP, the programming language behind most of the web, on Oct. 17. That was four days before the developers released a fix, according to a post on Amazon’s security blog.

Read more: IBM, Amazon quantum pricing gets a challenger in Postquant Labs

Amazon’s honeypot saw the attack before the public did

Amazon runs decoy servers, known as honeypots, to watch what attackers do. One of them recorded an exploit attempt from an anonymized internet address while the flaws were still under embargo, wrote CJ Moses, chief information security officer of Amazon Integrated Security.

It is still unclear how the attacker learned about the bugs early. Moses said proof-of-concept exploits were circulating as early as 2 a.m. UTC on Oct. 21, about 12 hours before the official release.

“PHP2Shell is now being broadly exploited in the wild,” Moses wrote.

More Technology:

• Anthropic announces big plan to help workers disrupted by AI
• Chip toolmakers can’t catch a break, even as TSMC spends more
• Apple’s new CEO just teamed up with Tim Cook’s old rival

The flaws, tracked as Common Vulnerabilities and Exposures (CVE) entries CVE-2026-109618 and CVE-2026-109619, let an attacker run code on a vulnerable server without logging in. Cybersecurity firm Wiz, which Alphabet (GOOGL) owns, said it found them with help from its own AI tool plus Google’s Gemini 4 Argon and OpenAI’s GPT-5.6 Sol, according to its research post.

Wiz said the bugs reach back to PHP 5.4 and had been hiding in the code for more than a decade. “Applying the PHP security update remains the only complete remediation,” the firm wrote.

Moses explained why the window keeps shrinking. “AI-assisted analysis makes it cheaper to turn a patch diff or advisory into a working exploit,” he wrote.

Governments pulled sites offline as PHP shipped fixes

The scale is what spooked officials. “All servers using any version of PHP released in the last 20 years will be vulnerable,” Tiffany Farriss, interim chief executive of the Drupal Association, told Reuters.

PHP runs roughly 75% of websites whose programming language can be identified, according to web tracker W3Techs, as cited by Reuters. It sits under WordPress, Drupal, Joomla and Magento.

Here’s how the response played out across the first 48 hours:

• The Netherlands’ National Cyber Security Centre took its own site offline, and the Port of Rotterdam, government portal Overheid.nl and several hospitals followed, according to Reuters.
• European cyber agencies, including the European Union Agency for Cybersecurity (ENISA), told organizations to “consider isolating exposed applications” if they cannot patch right away, according to a joint statement.
• Attackers began altering their payloads to slip past filters, Cloudflare (NET) said, as reported by The Hacker News.
• PHP released fixed versions 8.5.12, 8.4.27, 8.3.36 and 8.2.35 at 14:00 UTC on Oct. 21, according to The Repository, a WordPress industry newsletter.

Costin Raiu, co-founder of security firm TLP:Black, warned that the next phase could be “mass compromise of every vulnerable web server on the internet,” Reuters reported.

Censys count shows 1.5 million servers need upgrades

In my analysis, the outage headlines understate the real exposure. Internet scanning firm Censys counted about 1.5 million of 2.6 million hosts running a PHP version with no fix available, according to Reuters.

That is roughly 58%. Those servers cannot simply be patched; they need a full upgrade to a newer PHP release, which takes longer and can break older websites.

That gap will not close in a week. It means the risk window stays open well after the headlines fade.

Hosting companies are moving at different speeds

Most people with a website never touch PHP directly. Your hosting company does, and the responses so far vary.

Automattic said customers of WordPress.com and its other managed platforms were protected before the release.

“On WordPress.com, security is our job to solve, not yours,” said Ian Stewart, the company’s artistic director and product lead, according to The Repository.

GoDaddy (GDDY) said its Managed WordPress hosting was fully patched, while its standard Web and Windows Hosting still relied on interim protections as of Oct. 22, the newsletter reported. Bluehost warned that customers may see limits on upload file sizes while it finishes patching.

Risk is highest for small sites run by people with limited technical help, often on shared hosting where customers do not manage the servers, Reuters noted.

3 checks for your own website this week

You do not need to be a programmer to protect yourself. Start with three steps:

• Check your PHP version. In WordPress, it appears under Tools, then Site Health, then Info.
• Ask your host directly whether your server runs 8.5.12, 8.4.27, 8.3.36, 8.2.35 or later. Get the answer in writing.
• If you sell anything online, review recent orders and admin accounts for anything you do not recognize, and save your logs from Oct. 17 onward, as Amazon advised.

If you are still on PHP 8.2 or older, plan an upgrade now. PHP 8.2 reaches end of life on Dec. 31, according to The Repository.

Investors get a stress test for Cloudflare and GoDaddy

For investors, the story cuts both ways. Amazon and Alphabet just showed their security teams can spot threats early, while Cloudflare’s filtering business gets a real-world stress test.

GoDaddy carries more operational risk, because its customers skew toward small businesses on shared plans. I will be watching whether any host reports a breach tied to these flaws in the coming weeks.

The bigger shift is the speed. “The scale is huge, and with AI, exploitation may happen instantly,” Lukasz Olejnik, a visiting senior research fellow at King’s College London, told Reuters.

Your website is part of your income now. Treat your host’s patch status the way you treat a bank alert: check it today, not next month.

Related: Food Lion, Stop & Shop online ordering stays down for a 3rd day

These chips have the potential to grow into a massive enterprise.

Evercore ISI predicts $75 billion in 2027 and $113 billion in 2028 sales of Google’s TPUs to outside customers.

As a result, technology that was initially intended to reduce Google’s own computing expenses would become one of Alphabet’s main new sources of income.

Evercore analyst Georgia Anderson estimates TPU sales could represent roughly 35% of Google Cloud revenue in 2027, according to Investor’s Business Daily. She maintained an Outperform rating on Alphabet with a $450 price target.

The significance is clear to investors.

Google doesn’t need to outperform Nvidia in every area of AI computing.

Alphabet can profit from the AI infrastructure boom that has made Nvidia one of the world’s most valuable companies by renting or selling Google’s chips.

Additionally, Google controls the cloud where many of those processors will operate, in contrast to Nvidia.

Google turned an internal cost saver into a product

Google started creating TPUs more than 10 years ago because traditional processors became too costly for some of its biggest machine-learning tasks.

At first, the chips were mostly developed for Google’s internal services.

This tactic is evolving. Earlier in 2026, Alphabet announced that Google Cloud will begin providing TPU hardware to specific clients for use inside their own data centers.

The company expected a small amount of revenue from those agreements in 2026, with the vast majority arriving in 2027.

The opportunity is much increased by such action.

In the past, customers who wanted Google chips had to use Google Cloud to consume them. Now, customers may purchase systems for their own facilities, rent TPUs via Google Cloud, or use them through infrastructure agreements, all of which allow Alphabet to generate revenue.

For instance, Blackstone invested an initial $5 billion in a joint venture with Google that aims to launch 500 megawatts of TPU cloud capacity by 2027. Google will provide the chips, software and services.

As a result, Google begins to resemble a true provider of AI infrastructure rather than just a cloud provider that also designs chips.

Nvidia finally has a different kind of competitor

Google is not attempting to quickly replicate Nvidia’s operations.

With a software environment that developers have spent years understanding, Nvidia continues to be the leading supplier of general-purpose AI accelerators.

Google has a distinct edge.

More Tech:

• Meta and Microsoft just sent employees a memo about Anthropic
• Microsoft cutting off popular software with no extension Oct. 13
• Apple’s biggest AI advantage may face new privacy challenge

The company controls the entire stack around its TPUs. Google creates the processors, runs massive data centers, controls Google Cloud, and creates Gemini models that use the infrastructure.

Customers willing to optimize their software around Google’s framework may find TPUs more appealing.

Evercore’s industry checks suggest TPUs can offer 20% to 40% lower costs for customers with experienced engineering teams, although software limitations and a shortage of specialized talent remain barriers to wider adoption.

One client has already demonstrated the size that Google is pursuing. This year, Anthropic decided to increase the amount of Google TPUs it uses by several gigawatts. The extra capacity is anticipated to be online in 2027.

The Claude developer will use both Google Cloud services and Google-built TPUs provided by Broadcom.

It’s easy to overlook the importance.

Any big AI firm needs a significant amount of processing power. Until recently, much of the investor discourse expected Nvidia to account for the majority of that expenditure.

Google is showing that hyperscalers are becoming increasingly interested in the opportunity.

Wall Street is starting to put a value on Google’s chips

Evercore is not the only company seeing the emergence of a sizable enterprise.

Piper Sandler estimates external TPU hardware sales could climb from roughly $8 billion in 2026 to $51 billion in 2027 and $104 billion in 2028.

While BMO recently lifted its 2027 Google Cloud revenue prediction to $189 billion, citing improved cloud trends and the expanding semiconductor business, Wells Fargo has also increased its Alphabet forecasts due to external TPU sales.

These estimations differ significantly, and investors should remember that none of these figures are certain. Google still has to produce the chips, get sufficient power for data centers, and persuade consumers that implementing its technology is worthwhile.

Additionally, Nvidia has a significant installed edge.

However, TPUs are important, even if Alphabet does not overtake Nvidia.

Even with Evercore’s $75 billion projection, external chip sales would be a significant business inside one of the biggest corporations in the world.

The economics go beyond chip sales, which is more significant. In addition to purchasing storage, networking, databases, cybersecurity, and artificial intelligence services from Google Cloud, a client drawn to Google’s TPUs may also purchase these services.

This implies that a single chip transaction has the potential to increase expenditure inside Alphabet’s ecosystem.

For many years, investors mainly saw Google as an advertising firm with a potential cloud business. But a third option is being made possible by artificial intelligence.

Wall Street is beginning to place a lot of weight on Google, which may have developed one of the only reliable substitutes for the chips driving the AI growth.

Alphabet’s next big venture may already be in its data centers if Evercore is even close.

It was initially designed without external clients in mind.

Related: Bank of America sees a $300 billion test for Google stock

For most of the last century, major industrial employers answered that question the same way. You hired people, trained them, and paid them more as they got better at the job.

That bargain has been fraying for years at Volkswagen, the German group behind VW, Audi, Porsche, Skoda, and Cupra. Its chief executive has spent much of 2026 talking about job cuts, idle factory capacity, and a cost gap with rivals that he puts at roughly 20 percent.

If you’ve ever sat through a budget meeting, you know how this goes. Spending that looks like a cost gets cut. Spending that looks like an investment gets defended, even when money is tight.

So when the company finds money for something new, it’s worth asking where that money is going and who pays for it.

This week, I got an answer.

Volkswagen Group (VLKAF) has equipped more than 330,000 employees with Microsoft (MSFT) Copilot, the artificial intelligence assistant built into Word, Excel, Outlook, and Teams, Microsoft said on Oct. 21.

The timing is hard to miss. Two days earlier, Volkswagen’s own labor representatives said the company wants to hold salaries at 2026 levels for the rest of the decade.

Volkswagen puts Copilot on 330,000 desks

The rollout did not happen overnight. Volkswagen first gave staff broad access to Copilot Chat, the free chat version that comes with eligible Microsoft 365 plans, and saw more than 300,000 monthly active users on it, according to Microsoft’s announcement.

Employees are now using a more advanced version tied to what Microsoft calls Work IQ. The company says it reaches into engineering, production, procurement, and sales.

“The real value of AI comes from making it available at scale in everyday work,” said Sophie Vassarotti, head of group data and AI at Volkswagen Group, in the same release.

Related: Franklin Templeton CEO says AI is hiding a bigger tech story

Microsoft framed the deal as a template for other industrial companies.

“Volkswagen shows how companies can evolve AI from a personal productivity tool into a foundation for future AI innovation,” said Agnes Heftberger, chair of the management board of Microsoft Germany.

Microsoft CEO Satya Nadella cheered the news himself. “Super to see Volkswagen Group putting Copilot to work across the company!” he wrote in an X post on Oct. 22.

Here’s what the deal looks like by the numbers:

• More than 330,000 employees are equipped with Copilot, and more than 26,000 AI agents are active across the group over the past month, according to Microsoft
• About 660,000 people are on Volkswagen’s payroll worldwide, including 275,000 in Germany, according to Daily Sabah, which carried the wire report
• $30 per user per month, paid yearly, for Microsoft 365 Copilot for enterprise, according to Microsoft’s pricing page
• More than 30 million paid Microsoft 365 Copilot seats across all customers as of July, according to Nadella on Microsoft’s fiscal fourth-quarter call, as reported by UC Today

Pay freeze talks hit Europe’s biggest carmaker

Volkswagen’s Global Works Council said the executive board told human resources managers worldwide to suspend pay increases, a directive in effect since early October, according to Daily Sabah.

Germany’s Süddeutsche Zeitung estimated affected staff could lose about 10 percent of their income to inflation over that stretch.

More Jobs:

• Jeff Bezos explains why Amazon is still cutting jobs
• Bill Gates says some jobs should be reserved for humans
• Elon Musk delivers 6-word warning on future of employment

A Volkswagen spokesperson declined to comment on what the company called “speculation” and said labor cost proposals belong in talks with unions, according to the same report.

The union is not backing down. “At Volkswagen, pay development is decided at the negotiating table, not by decree from the executive boardroom,” an IG Metall spokesperson said. The union is seeking a 5 percent raise over 12 months.

The pay fight sits on top of a much bigger restructuring. Volkswagen said in March that it would cut about 50,000 jobs in Germany by 2030 after profit slid to its lowest level since 2016, according to AFP.

In July, CEO Oliver Blume told staff another 50,000 jobs could go, citing that 20 percent cost disadvantage. “We are currently assessing across all brands, companies and regions how many adjustments are actually necessary and feasible,” Blume wrote in an internal memo, according to Just Auto, which cited Reuters.

The company is chasing 6 billion euros in annual net savings by 2030 and had locked in about 1 billion euros so far, according to Quartz, citing Blume’s June 18 shareholder remarks.

Workplace columnist Suzanne Lucas read the pay plan bluntly. In her view, the message to staff is: “We don’t think you’ll find another job, and we don’t care about you when you do stay here,” she wrote in an Oct. 21 opinion piece on Inc.

What a Copilot AI assistant seat costs at list price

Neither company disclosed what Volkswagen pays, or how many of the 330,000 seats are paid for Microsoft 365 Copilot licenses versus the free chat tier. Large customers also negotiate below the list price.

Still, the list price gives you a ceiling. In my analysis, 330,000 seats at $30 a month works out to $118.8 million a year.

That is a small slice of a 6 billion euro savings target. Per worker, it comes to $360 a year.

Now set that against the pay side. If Süddeutsche Zeitung’s estimate holds, a frozen salary could cost those same workers around 10 percent of their purchasing power by 2030.

That contrast is the real story here. A few hundred dollars per head for software clears the budget, and the raise does not.

Whether the tool pays for itself is still an open question. “Most organisations remain uncertain if Copilot’s promised productivity gains justify the added cost,” wrote Daryl Ullman, a partner at Microsoft licensing consultancy SAMexpert, in an Oct. 20, 2025, analysis.

Ullman’s firm sells Microsoft contract negotiation services, so weigh that view with the same skepticism you’d give Microsoft’s own productivity claims. Microsoft has also faced user pushback on AI features in Windows, a reminder that rollout numbers and daily use are different things.

Microsoft earnings Oct. 28 put enterprise demand in the spotlight

You won’t wait long for the next data point. Microsoft reports fiscal first-quarter 2027 results after the market closes on Oct. 28, with a call at 2:30 p.m. PST, the company said in an Oct. 7 release.

Paid Copilot seats climbed from 15 million in January to more than 20 million in April, according to CNBC, before topping 30 million in July.

If every Volkswagen seat turned out to be paid, the deal alone would equal about 1% of that July base. I’ll be listening for whether management singles out industrial customers like Volkswagen when it updates the count.

Volkswagen reports its own third-quarter results on Oct. 29, according to MarketScreener. Investors will want to hear how much of the savings plan is coming from labor and how much from tools like this one.

How this hits your paycheck and your portfolio

If you own an S&P 500 index fund or a target-date fund in your 401(k), you already own a meaningful piece of Microsoft. Deals like this one are how the company is trying to turn its massive AI spending into recurring subscription revenue.

If you work for a large employer, the Volkswagen playbook may look familiar soon. Budgets for software are getting approved while budgets for raises are getting squeezed.

That makes it worth being deliberate. Track what you actually do with the time an AI assistant saves you, and bring that record to your next review.

The companies that win this shift will be the ones that share some of the gains. Your job is to make sure your share shows up in your paycheck, not just in someone else’s savings target.

Related: Volkswagen may cut 100,000 jobs in brutal reset

Monthly installment plans make the shiny-new-object habit easy to feed. Split a four-figure phone into 24 payments and it starts to feel like a utility bill, so most buyers never add up the full tab.

Apple (AAPL) just raised the stakes. Its first foldable phone, the iPhone Duo, starts at $1,999 for 256GB and runs to $3,199 for 2TB, according to Channel Insider. That puts the entry model $700 above the iPhone 18 Pro Max.

Apple’s new CEO pitched it big. The Duo is “the most transformational change to iPhone since the original,” John Ternus said at the Sept. 9 launch event, CFJC Today reported.

Buyers lined up despite the price tag. The Duo sold out within minutes of preorders opening on Oct. 16, and new orders now ship in January, Dexerto reported. Some confirmed preorders resold on eBay for close to $5,000.

Now, one of YouTube’s most-watched tech reviewers is telling most of those shoppers to hold on to their cash. Marques Brownlee posted his full iPhone Duo review on Oct. 21, two days before the phone is due to reach stores on Oct. 23.

Why Marques Brownlee says most upgraders should pass

Brownlee spent about a week and a half with the Duo before publishing. His short version is that Apple’s first foldable is “polished around the edges, but kind of rough in the middle,” according to Dexerto.

For people who already own an iPhone and are wondering whether to trade up, his answer was blunt. For most of them, it’s probably no.

Related: AT&T CEO drops 4-word verdict on new Apple iPhone Duo

He also named the buyers who should walk away. That list includes people who mostly text, scroll social media and browse the web, anyone who picks a phone for its camera, and anyone who tends to break phones, Dexerto reported.

That last group matters more than it sounds, as the repair math below shows.

The trade-offs hiding inside Apple’s first foldable

The problems Brownlee flagged sit in the features you touch most often, from unlocking the phone to opening your favorite apps.

More Apple News: 

• Apple’s $1,999 foldable iPhone creates strange twist for Samsung
• Bank of America resets Apple stock price target after iPhone Duo launch
• Bank of America finds a quicker path to Apple’s iPhone Duo

Here are the four issues behind his verdict:

• Apps that don’t fit yet: Spotify, Google Maps, Discord, X and Google Calendar aren’t properly optimized, showing black bars on the outer screen or shrunken layouts on the inner one. He called app support the Duo’s biggest weakness at launch, Dexerto reported.
• No Face ID: Apple swapped it for a fingerprint reader in the power button, with a built-in delay that makes unlocking feel slow. He expects it to be the feature former iPhone users miss most, according to Dexerto’s account of the review.
• Extra bulk: Samsung’s Galaxy Z Fold 8 is noticeably thinner and more than 20% lighter, he said, per Dexerto.
• Weaker cameras: There’s no dedicated zoom lens and no ProRes or Log video, and low-light shots land closer to the budget iPhone 17e than to the Pro models, MacDailyNews reported.

He expects app support to improve as developers ship updates over the coming months, though some apps may never get a Duo-specific version, Dexerto reported.

What a $1,999 phone really costs over 2 years

I ran the numbers the way most buyers will actually pay: monthly, over two years. Apple lists the Duo at $83.29 a month for 24 months, according to Channel Insider, which works out to the full sticker price with no added interest.

The iPhone 18 Pro Max, at $1,299, comes to about $54.13 a month over the same stretch. That’s a gap of roughly $29 a month, or $700 in total.

Then comes protection. AppleCare+ with Theft and Loss for the Duo costs $19.99 a month, $199.99 a year or $399 upfront, according to 9to5Mac, which cited Apple’s support site.

Stack the monthly plan on top of the phone and you’re paying $103.28 a month. Over two years, that’s $2,478.72 before sales tax.

Skipping coverage is the bigger gamble. Apple estimates $999 to repair the Duo’s inner display without AppleCare+ and $1,499 for other damage, 9to5Mac reported. With coverage, that inner screen fix drops to $99.

In my analysis, that’s where Brownlee’s warning about phone breakers really bites. One cracked inner screen without a plan costs half of what you paid for the phone.

Brownlee framed the decision the same way. He said the Duo is worth jumping into only if neither the $2,000 price nor the bump in your monthly payment means much to you, Dexerto reported.

Where a $700 premium could work harder for you

The $700 gap over a Pro Max is real money, and that cash has other jobs it could be doing.

Park it in an emergency fund and it covers a car repair or surprise medical bill without reaching for a credit card. For many households, that cushion provides more financial security than a bigger screen ever will.

Invest it, and the math compounds. At a hypothetical 7% average annual return, $700 grows to about $1,377 in 10 years. I’m using that rate as an illustration only, since actual returns vary and can turn negative.

There’s also a timing cost. Brownlee said he hopes Apple fixes the Duo’s hardware and software gaps in a second-generation model, Dexerto reported. Buying now means paying full price for the version Apple is still learning from.

Foldable fans who can still justify the iPhone Duo

Brownlee still liked plenty. He praised the 7.6-inch matte display that hides the crease, the battery life, smooth hand-offs between screens and the ability to run two apps side by side, Dexerto reported.

He summed up the pitch as paying $2,000 for an iPad Mini that fits in your pocket, MacDailyNews reported. If you’d otherwise carry a phone and a small tablet, that trade can pencil out.

If you’re open to Android, shop around first. Samsung’s Galaxy Z Fold 8 starts at $1,899.99, according to Tech Advisor.

And if you’re on the fence, waiting costs you nothing right now. New Duo orders already ship in January, and every week you hold off gives app developers more time to catch up.

Apple will sell plenty of Duos. Sellouts and eBay markups make that clear.

Your job is narrower. Price the phone with protection included, and make sure the payment fits a budget that’s still building your savings.

A bigger screen in your pocket is nice. Owning your next upgrade outright, with your emergency fund intact, is nicer.

Related: Jim Cramer’s 4-word Apple rule could save you real money

The facility matures in 364 days, according to an Oct. 22 press release. A one-year term makes it a bridge to longer-term financing, not the financing itself.

The timing says more than the dollar figure. A day earlier, Nebius revenue chief Marc Boroditsky told CNBC that AI computing prices are rising faster than expected. He said a capacity auction on Oct. 20 showed prices still climbing.

He also set Nebius apart from CoreWeave Inc. (CRWV), where the cost of insuring debt against default hit distress levels, CNBC reported. “We’re not carrying the kind of debt that they are,” Boroditsky said.

Together, the moves show a company sure about demand but unwilling to lock in long-term debt at today’s prices.

“Sequencing our financing is how we optimize our long-term cost of capital,” said Chief Operating Officer Ophir Nave.

Also read: Nebius stock breaks below $200 as CoreWeave’s debt scare spreads

A one-year Citi loan buys Nebius time

The loan charges Term SOFR, a benchmark tracking U.S. short-term rates, plus 2.75 percentage points, the release said. That margin can rise to 3.5 points. Because the base rate floats, any Federal Reserve hike would lift Nebius’s interest bill.

Last week, I wrote that cheap money had become the scarcest input for GPU landlords. As I noted then, Nebius’s first secured loan, in July, cost 2.5 points over SOFR. The Citi deal starts a quarter point higher, a modest premium rather than a distress signal.

Alongside customer prepayments and operating cash flow, the money will fund data centers and GPUs, the release says. It stacks on roughly $10 billion of convertible bonds already outstanding, Reuters reported.

A bridge is only as safe as its exit. Nave said Nebius can tap asset-backed financing tied to more than $40 billion of contracts with investment-grade customers. Boroditsky added that Nebius’s 2027 capacity would sell out if brought online, CNBC reported.

Rental rates still determine how much profit that demand throws off. Nebius has just put those rates on public display.

Nebius put its GPU prices where lenders can see them

Nebius began publishing live spot prices for its GPU capacity on Oct. 21, the company said. Spot capacity is computing time that Nebius can interrupt. The prices also appear on outside platforms such as Silicon Data.

Investors are debating how fast Nvidia’s chips lose economic value, CNBC noted. Nebius just handed that depreciation debate a live scoreboard.

Nvidia Vice President Nico Caprez called its hardware “productive, durable, and fungible” in Nebius’ announcement. “Durable” is the word that is disputed.

Crusoe CEO Chase Lochmiller offered a counterpoint. “We’re seeing higher pricing today for A100s than we saw in 2020 or 2021,” he told CNBC about the older Nvidia chips. Strong rents on aging chips are how neoclouds earn back their GPU spending.

A live price feed cuts both ways. If rates slip, Citi and every shareholder will see it live. I’d watch that page more closely than any executive interview.

Nebius is also piloting rebates for customers who release idle GPUs back into its spot pool. That pilot quietly admits that some reserved capacity sits unused, even though Boroditsky told CNBC that demand still outruns supply.

The customers paying those prices are in production

Nebius’ customer list points to everyday workloads rather than hoarding. Netherlands-based Prosus said Nebius’ inference service doubled throughput for its e-commerce businesses.

Neobank bunq’s assistant, Finn, matches leading closed models at 60% of the cost per answer on Nebius, the company said. Cheaper answers tend to invite more questions, so falling unit costs can still grow total compute spending.

Its European capacity now spans six countries, from Finland to Iceland, the company also said. It is adding $10 million in compute grants to draw the region’s builders. Both need more data centers, and the Citi loan funds the next stage of that global build-out.

More Artificial Intelligence:

• AI’s biggest business opportunity may not look like AI
• AI racks are outgrowing old power: 2 chipmakers have a strategy
• BofA splits software into winners and losers as AI agents arrive

Wall Street is betting the bridge holds

The stock has not rewarded either move. Shares rose Oct. 21 after SpaceX (SPCX) reported a ninefold jump in AI revenue, CNBC reported. They closed almost flat, then fell nearly 4% in premarket trading the following day.

Analysts are more patient. The 22 analysts tracked by StockAnalysis rate Nebius a Buy on average, with a price target of $283.60. That target sits about 65% above the latest close, a bet that signed contracts outweigh today’s debt jitters.

Once GPU rents trade in the open, lenders can price the chips behind every neocloud loan. In that market, the cheapest borrower may win the buildout race, not the fastest builder.

The price Nebius pays to refinance this bridge will tell every rival what lenders now believe.

More Stocks News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off

Musk himself has called combining them “a good idea,” and a deal would dwarf anything Wall Street has seen.

The idea itself is not new. Wedbush analyst Dan Ives has put the odds of SpaceX eventually merging with Tesla at 80% to 90% by early 2027, viewing the SpaceX IPO as groundwork for a deal built on shared supply chains and the joint Terafab chip project.

What changed this week is that the two companies have finally started doing the homework.

Musk says SpaceX, Tesla form committee amid potential merger

Musk was careful about his own role, even while saying he thinks a combination would be a good idea, with gains from cooperating on AI software, chip design, robotics, and launching mass to orbit.

He said both companies set up special committees of independent directors at shareholders’ request, and because he is conflicted, he cannot take part in the talks, calling the setup “an odd kabuki dance.”

Also read: Elon Musk takes a direct swipe at Broadcom CEO and AI investors

Musk also tried to cool expectations. The committees have not agreed on terms, he said, and “there’s no guarantee that they will,” adding that treating a merger as a sure thing would be “overstating it.”

The vote is another hurdle. Musk pointed out that Tesla would need approval from 50% of all its shareholders, not just those who cast ballots, and he called that a high bar. Shares that never get voted effectively count as a no.

Investors still liked what they heard. Tesla shares rose a little less than 2% shortly after the call, the Wall Street Journal reported. The bump came just a day before Tesla’s own quarterly results, giving shareholders plenty to ask about on that call.

Gizmodo also noted, citing a Wall Street Journal report, that a merger could lift Tesla’s valuation enough to help Musk qualify for his $1 trillion pay package.

SpaceX brought a blowout quarter to the table

The merger news landed on top of a strong report.

SpaceX posted its first profitable quarter since its June IPO, earning $4.4 billion after a $1.1 billion loss a year earlier, while revenue of about $14.6 billion easily topped the $12.7 billion analysts expected, and it finished the quarter with more than $51 billion in cash and $38.6 billion in marketable securities, according to SpaceX’s SEC filing.

Look under the hood, and AI did most of the heavy lifting. SpaceX’s Super Intelligence unit brought in $7.6 billion, up from $813 million a year earlier. The connectivity side, home to Starlink, also grew, just not nearly as fast, landing at $5.2 billion. No wonder Musk keeps pitching the merger as an AI play.

That growth carries some risk, though. Anthropic and Google are the anchor customers for SpaceX’s computing business, but both contracts include 90-day termination clauses. A merger partner with its own heavy AI needs could make that capacity less dependent on outside buyers.

Tesla heads into its own report with some momentum. The carmaker delivered 486,532 vehicles in the third quarter, well ahead of the roughly 461,100 analysts had forecast, as reported by Quartz.

Its results on Oct. 21 will be the first chance for shareholders to press management on the merger.

The AI logic behind a combination

The two companies are already tangled together. Both are betting heavily on AI; they are partnering on Terafab, Musk’s chip factory project in Texas, and SpaceX has already folded in a potential breakthrough in chip fabrication but does not want to discuss the specifics.

Terafab alone shows the scale of the shared ambition. A May SpaceX filing put the first phase at $55 billion; Intel has signed on as an outside foundry customer.

SpaceX President Gwynne Shotwell said in September that the companies were integrating “faster than I thought” but were “not fully integrated yet,” TheStreet reported.

More Elon Musk:

• SpaceX CEO Elon Musk has 7 words for Delta Airlines after snub
• Elon Musk just handed Nvidia investors another big win
• Billionaire investor makes stunning $15 trillion Musk call

The two companies already do plenty of business with each other. Over 2024 and 2025, SpaceX and its xAI subsidiary bought roughly $697 million in Megapack battery systems from Tesla, plus another $131 million in Cybertrucks last year, according to Reuters.

Tesla, for its part, put $2 billion into xAI in January. Fold everything into one company, and those side deals simply disappear.

Robotics is the other big piece. Tesla raised its 2026 capital spending guidance to $25 billion, with a large share going to turn old Model S and Model X capacity into Optimus robot lines.

Analysts do not expect Optimus to go on sale before late 2027. Musk sees those robots and SpaceX’s computing muscle as natural partners.

Plenty of questions remain before any deal gets done

Not everyone buys the merger story. Investor Ross Gerber has argued that a combination would look more like SpaceX “bailing out Tesla by buying them over, calling it a merger,” according to Benzinga. Analysts have flagged the stock swap terms, which company ends up as the parent, and possible dilution for retail holders.

Then there is the question of who ends up in charge. Jefferies figures Musk could hold on to roughly 55% of the voting power if the deal carried no premium, as reported by TheStreet. JPMorgan described the pairing as strategically coherent. BNP Paribas, which has an underperform rating on Tesla, was less convinced, warning that SpaceX’s cash burn could hold things up.

Some observers think the outcome is all but settled anyway. Gene Munster of Deepwater Asset Management put the chance of a deal within two years at 90% and said: “shareholders would support this,” Benzinga noted.

For now, Musk has offered no timetable, leaving the next move to the committees and, eventually, shareholders.

Related: Elon Musk makes an audacious claim on the growth of SpaceX

Shares of Moderna fell 17.6%, while Merck (MRK) dropped 1.76%.

The sell-off came despite positive trial results for intismeran autogene, Moderna’s personalized mRNA cancer vaccine developed with Merck.

Moderna has been one of the S&P 500’s top-performing stocks this year, surging 587% year-to-date, mostly following its August announcement that the Phase 3 trial had succeeded. Merck shares have gained about 40% over the same period.

Moderna cancer vaccine data leaked ahead of conference

The Phase 3 INTerpath-001 trial evaluated intismeran autogene, a personalized mRNA melanoma vaccine, in combination with Merck’s immunotherapy drug Keytruda in patients with high-risk melanoma.

The study examined whether the combination could reduce cancer recurrence and distant metastasis compared with Keytruda alone.

The detailed results were scheduled for release on Saturday, Oct. 24, at a major medical conference in Madrid.

Also read: Meta stock falls despite Wall Street analysts raising price targets

However, Oncology Central, a cancer news website operated by publishing company Informa, prematurely published an article containing the embargoed data, MarketWatch reported.

Merck and Moderna subsequently released the trial results.

Wall Street analyst said Moderna sell-off was ‘a sell the news event’

Bernstein said the leaked results appeared to show “a strong result” for the combination treatment, The Fly reported.

The firm described the findings as a “strong Phase 3 result and meaningful validation of Moderna’s individualized neoantigen platform.”

Despite the positive data, Moderna shares fell sharply, which Bernstein is “interpreting as a sell the news event.” Bernstein maintained its Market Perform rating and $45 price target on Moderna.

Also read: Intel stock surges on JPMorgan upgrade after 14% weekly plunge

Wells Fargo also said concerns over the trial data “may be overdone.”

The firm noted that recurrence-free survival results exceeded investor expectations, while overall survival data are not yet mature.

Wells Fargo said it would buy Merck shares on weakness ahead of other potential catalysts, including upcoming Sac-TMT lung cancer data at the European Society for Medical Oncology conference.

The firm maintained its Overweight rating and $165 price target on Merck.

Both analysts’ price targets are below Moderna’s closing price of $202.63.

Related: Cathie Wood sells $16 million of surging tech stock

That has fueled worries among some investors, who have been asking whether it’s just panic selling or whether the AI trade is losing ground. 

By Friday, Oct. 16, CRWV closed at $61.39, which is far below its May peak near $138.

Then, this week, a major bank issued a call that could ease some investors’ worries. The bank is positive on the business but much less generous on the price target. 

Goldman Sachs upgrades CoreWeave to Buy with a $106 target

On Oct. 19, 2026, Goldman Sachs analyst Gabriela Borges upgraded CoreWeave from Neutral to Buy, but the bank also lowered its 12-month price target from $139 to $106.

Still, even the reduced target points to roughly a 73% increase from the Oct. 16 close, Benzinga reported.

Also read: Palo Alto CEO wants AI labs out of the ‘fear-mongering business’

CoreWeave rents high-powered Nvidia chips to companies training artificial intelligence models. Major companies such as Meta and OpenAI can pay to use CoreWeave’s data centers instead of spending a lot of money to build their own.

This business made CoreWeave one of 2025’s hottest IPOs, but because the company will need to borrow heavily to build more of these data centers, some investors are worried.

However, Borges and co-analyst Maura Hager argue that investors have overestimated how much CoreWeave’s future borrowing will affect the company. They also believe people are underestimating how profitable each new data center contract can become, according to ZeroHedge.

Why the bond market panicked, and Goldman called it overdone

The sell-off worsened because yields on CoreWeave’s 2031 and 2032 bonds climbed to around 15%. That spike made investors worry about whether CoreWeave could keep funding its planned $35 billion to $39 billion in 2026 capital spending, Barron’s reported.

However, Goldman thinks the 15% yields reflect what traders in the open market are willing to pay for CoreWeave’s existing bonds right now. When CoreWeave takes out new loans, they are usually backed by its Nvidia chips and signed customer contracts, which means lenders charge much lower interest rates.

Nvidia CEO Jensen Huang said something last week that supported that view. “CoreWeave’s strength is improving every day. I’m optimistic. Their cost of capital is actually decreasing,” Bloomberg reported.

What Wall Street is watching beyond Goldman’s note

By the end of the second quarter of 2026, CoreWeave had locked in about $104 billion in future revenue through long-term contracts with major companies like Meta, Reuters reported.

The company also raised its GPU rental prices by 25% in July, then by another 10% in October, according to Seeking Alpha. That shows that the company’s demand is still greater than its supply. 

More AI Coverage:

• OpenAI CEO Sam Altman has a biblical-level warning about AI
• SpaceX turns to debt again to fund a historic order
• Anthropic announces big plan to help workers disrupted by AI

Based on that belief, Cathie Wood’s Ark Invest bought about $25.8 million worth of CRWV shares across two sessions in mid-October, Benzinga confirmed.

JPMorgan, Jefferies, and Deutsche Bank also stayed bullish on the stock even during the sell-off, with Deutsche Bank calling the drop a buying window.

What everyday investors should consider before buying the dip

CoreWeave reported that its second-quarter revenue doubled to $2.575 billion, which is impressive, but the company still posted a $626 million net loss. A major driver of that loss was interest on its debt, which totaled about $640 million that quarter alone.

That is what D.A. Davidson analyst Gil Luria was talking about when he warned that if AI spending slows down and borrowing stays this expensive, the interest costs could eventually swallow the company’s profits, Benzinga noted.

CoreWeave co-founder Brannin McBee pushed back on that fear, pointing out that customers have already signed up for nearly all of the company’s 2027 data center capacity. This means the revenue to cover those interest payments is already locked in.

Goldman’s upgrade reduces the worries of CoreWeave going through a liquidity crisis. However, it does not remove the risk of owning a highly leveraged growth stock in a volatile AI market.

A smart move for anyone considering the stock now is to watch the company’s third-quarter earnings report closely for updates on its debt and demand levels.

More Stocks News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off

The timing is interesting. Nvidia held up better than most chip names during last week’s AI sell-off, helped by second-quarter revenue that more than doubled from a year earlier, mostly on data center sales, as reported by TheStreet.

Now it wants to speed up the slowest part of the whole business: getting the buildings ready.

Nvidia’s plan to build data centers in weeks, not months

The DSX POD is a blueprint rather than a finished product Nvidia sells off the shelf.

It is a reference design for complete Vera Rubin AI factories that partners build inside modular containers, with all the cabling, piping, and testing handled at the factory before anything ships, according to Nvidia.

Also read: Nvidia and Micron are about to lord their successes all over Wall Street

The boxes are packed tight. Each container holds seven Vera Rubin NVL72 racks with 504 GPUs, plus roughly 20 kilometers of copper cabling and 17 kilometers of networking that arrive already installed and tested, The New Stack noted. That’s the kind of fiddly work that usually eats up weeks on a construction site.

The headline promise is speed. Nvidia says a DSX POD can be up and running in as little as 14 days, and Huang used the same stage to tell his audience that every country and company is building their own AI.

Vera Rubin is the engine inside. Back in July, Huang brushed off reports of manufacturing problems and said Rubin hardware was already in production, while Nvidia said Rubin-based products would reach partners in the second half of 2026 and its Vera CPU was in full production.

Why every week of delay costs money

“An AI factory creates value when it starts producing tokens,” Huang said, adding that the value does not come when construction wraps up. Nvidia says cutting deployment at a prepared site from six months to about two weeks gets customers to revenue sooner.

The sums involved are huge. Huang told the Berlin audience that each gigawatt data center costs about $50 billion to build, and he estimated that kind of capacity could bring in about $40 billion a year in revenue, as reported by Investing.com. Every month a site sits idle is money left on the table.

Demand is not the problem. Morgan Stanley analyst Joseph Moore believes Nvidia’s underlying demand is closer to 100% growth than the 70% it has guided for, while CFO Colette Kress warned that supply will “remain a bottleneck” through at least fiscal 2028, TheStreet reported.

There is a catch, though. The two-week timeline only works if the site already has power, cooling, and network connections in place. The container solves the inside of the building, not the grid outside it.

Power is still the bigger bottleneck

The grid is where most AI projects get stuck. Energy firm Currence estimates that 30% to 50% of the large data center capacity expected to come online in 2026 will be delayed, with electricity among the main obstacles, TheStreet reported. A faster container does not change that math.

Microsoft just ran into that wall. The company pulled out of a planned deal for capacity at Nscale’s AI data center in Loughton, England, after grid officials said power for the site may not arrive until the 2030s. Microsoft declined to comment on the withdrawal, and Nscale says it is exploring other power sources, including fuel cells, for the project.

More Nvidia:

• Nvidia stock flashes massive signal as Wall Street leans in
• Nvidia’s AI boom revives a Warren Buffett warning for investors
• Nvidia puts $500 billion on the table, challenges banking as we know it

The site had been planned around roughly 23,000 Nvidia GPUs. Power used to be an assumption in data center planning. Now it is a constraint that shapes the entire project from the start.

Water is the other pressure point, and here the DSX POD helps. Because Vera Rubin can run on coolant as warm as 45 degrees Celsius, dry coolers can replace evaporative cooling towers in most climates.

Nvidia also estimates that this saves about 54 million liters of water a day for a 2-gigawatt deployment, though very hot climates may still need chillers on the worst days.

Partners line up for an early 2027 rollout

Nvidia has a long list of builders ready to go. Seventeen companies, including Anduril, Crusoe, Dell Technologies, HPE, Schneider Electric, Foxconn’s Ingrasys, and Vertiv, are signed on as partners, with worldwide deliveries targeted for early 2027, according to Nvidia.

Europe is clearly part of the pitch. A day before the keynote, the German government agreed to an AI partnership with Nvidia at a Berlin investment summit, with about a dozen German companies, including Siemens and Audi, expected to join.

Huang argued that countries should treat their data as a national resource to build a “national intelligence.”

Sovereign projects are already leaning on the same toolkit. A Japanese consortium called Noetra, backed by SoftBank, Sony, NEC, and Honda, plans an AI factory with about 27,500 Rubin GPUs built on Nvidia’s DSX framework.

If containers get factories like that running faster, Nvidia gets paid faster, too.

Related: Nvidia dodged the AI selloff, and Jensen Huang has a theory why

The company posted adjusted earnings of $4.64 per share, up 161% from $1.78 a year earlier. GAAP earnings totaled $4.63 per share, up 51% year over year. Revenue totaled $4.38 billion, up 55% from the same period last year.

Analysts expected Western Digital to report earnings of $4.08 to $4.11 per share and revenue of $4.14 billion to $4.16 billion.

Adjusted gross margin rose to 58.1% from 43.9% a year earlier.

.

In a research note shared with me on Oct. 20, 2026, BofA Securities analyst Andrew Obin kept his Buy rating on GE Vernova and left his price target at $1,310. That is about 47% above the stock’s recent close of $890.33.

He remains confident in the stock because the company is getting many orders driven by AI data centers’ demand for more energy.

Why Bank of America is sticking with its bullish call on GE Vernova

Obin expects GE Vernova to book orders worth up to $21.0 billion in the third quarter.

That would be 44% higher than the same period last year, with a book-to-bill ratio of 1.70x. The book-to-bill ratio basically means the company is getting new orders 1.7 times faster than it supplies.

Also read: OpenAI CEO Sam Altman has a biblical-level warning about AI

In addition to that, Obin expects revenue of $12.4 billion and adjusted EBITDA of $1.8 billion for the quarter, which are both more than what Wall Street estimated.

“We forecast $21.0bn in 3Q orders ($13.4bn Power, $6.2bn Electrification, $1.5bn Wind). This represents 1.70x book-to-bill and 44% y/y growth,” Obin wrote in the note.

How gas turbines turned GE Vernova into an AI play

GE Vernova builds the gas turbines, grid equipment, and wind machines that keep the lights on across the electric grid. The company was spun off from General Electric in April 2024, GE Vernova reported, and its gas turbine business has quickly become its main growth driver.

The main driver is the AI market. AI data centers consume a lot of energy, which has raised the demand for power. In fact, the U.S. grid has not seen demand like this in two decades, Investing.com reported.

More Energy Stocks

• U.S. loaning $4.2 billion to energy firm with crashing stock
• ‘I think you should own it,’ Cramer says of surging energy stock
• U.S. making bold $1.5 billion investment in defense-critical chipmaker

Due to the high demand, Obin now expects GEV to reach 13 gigawatts of gas power equipment contracts in the third quarter and 14 gigawatts in the fourth. GE Vernova’s management has already said second-half contracts will top 20 gigawatts.

“GE Vernova is uniquely positioned to meet accelerating global electricity demand,” CEO Scott Strazik said earlier this year in a LinkedIn post. “Economic growth, grid modernization, electrification, data center expansion, and the need for more reliable, resilient electricity are creating sustained demand for our platform of solutions.” 

The risks investors should watch out for

GE Vernova shares have fallen by about 9% in the past five trading days, and the stock trades at a price-to-earnings ratio of around 25, which is not cheap for an industrial company.

Obin noted that he’s watching whether data center demand will stay strong and what changes there could be among casting and forging suppliers. He’s also watching whether GE Vernova can keep pushing prices higher.

However, he noted that Q4 orders will likely not top the 24 gigawatts of contractual agreements the company booked in the final quarter of 2025.

Still, Obin pointed out that “a peak in quarterly orders does not mean a peak in backlog,” since contracts are already being signed for 2032 deliveries.

The way I see it, that long production pipeline is what makes the current pullback feel temporary. The orders the company already has give the business years of visibility.

More Stocks News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off

Here’s a closer look at how many Apple shares Ternus owns now, early in his tenure as chief executive.

How much of Apple does John Ternus own? 

John Ternus owned 59,567 shares in Apple, according to an SEC filing on October 5, 2026. His stake represents just a tiny fraction of the company’s 14.59 billion shares outstanding.

Ternus’s shares were valued at over $20.2 million based on a mid-October 2026 share price of $339.50. 

Related: John Ternus’s net worth as Apple’s new CEO

Is John Ternus the largest individual shareholder of Apple?

Compared to other members of Apple’s board, Ternus is a relatively minor insider shareholder, at least for now.

Arthur D. Levinson owned 3.7 million shares, according to an SEC filing in May 2026; he has been accumulating Apple shares since joining the board of directors in 2000.

Tim Cook, whom Ternus replaced as CEO, and who now serves as executive chair of Apple’s board, owned 3.2 million shares, according to an October 2026 SEC filing. His shares were primarily derived from stock awards in the form of restricted stock units (RSUs) for meeting the company’s internal performance-based goals during his tenure as CEO.

Levinson and Cook disclosed their Apple shareholdings as board members under disclosure rules. It is possible (but relatively unlikely) that there are other individuals who own more shares than Cook or Levinson but aren’t required to disclose their ownership because their stake doesn’t represent 5% or more of the company. (A 5% stake in Apple would be equivalent to nearly 730 million shares.)

Institutional investors own the biggest individual stakes in Apple.

Apple’s largest institutional shareholders

Institutional ownership in Apple amounted to 76.6% of total shares, according to Nasdaq. The Vanguard Group was the largest institutional investor, with a 9.8% stake. BlackRock (8%) and Vanguard Capital Management (6.6%) rounded out the top three.

More on Apple:

• Apple’s stock split history: Everything you need to know
• Apple’s dividend explained: Yield, history & more
• Where is Apple’s headquarters? Apple Park explained

How & when did John Ternus accumulate his Apple stock?

Ternus first joined Apple on the company’s product design team in 2001, and he eventually worked his way up to senior vice president overseeing hardware engineering, reporting to his predecessor, Cook, before becoming CEO himself. As he was promoted over two decades, he was awarded shares and likely sold some shares along the way. 

According to his latest SEC filing, he sold over 74,000 shares in early October 2026.

If his predecessor’s stock awards are any guide, Ternus could potentially be awarded a significant number of shares in the coming years, assuming he remains CEO. When Cook became CEO, he only held 13,659 shares in Apple, according to the tech giant’s 2011 proxy statement. After stepping down as CEO, Cook held 3.2 million shares. 

Related: Tim Cook’s net worth: The retired Apple CEO’s wealth in 2026

How much stock does John Ternus get as part of his compensation package? 

Ternus owns restricted stock units (RSUs), awarded to him as part of his performance. He had 99,878 shares that vested on October 1, and over the next two days he sold almost 74,446 shares, fetching nearly $25 million. He has around 250,000 RSUs that will vest in 2027 and into the next few years, based on an SEC filing that disclosed Ternus’s ownership in Apple stock when he became CEO on September 1, 2026. 

As CEO, Ternus’ performance-based stock awards will likely increase in coming years if the company meets or exceeds its internal targets.

Related: Apple’s stock performance under Tim Cook: A retrospective

The decline came as investors grew concerned about the social media giant’s aggressive spending on artificial intelligence.

The selloff extended beyond Meta. Intel (INTC) and Advanced Micro Devices (AMD) dropped about 6% and 7%, respectively, while Nvidia (NVDA) fell more than 4%. All other Magnificent Seven stocks fell.

CoreWeave (CRWV), the AI cloud provider whose debt-heavy balance sheet partly led the tech sell-off last week, also declined nearly 4%.

Tech stocks have become sensitive to headlines, with even strong earnings reports triggering sell-offs as investors grow more concerned about AI risks.

Meta beats Q3 revenue estimates, lifts AI spending 

Meta reported third-quarter revenue of $64.6 billion, beating analysts’ expectations of $63.41 billion.

Earnings came in at $3.83 per share, below the consensus estimate of $5.46, largely due to a $9.89 billion legal charge related to a settlement over teen mental health and social media.

Excluding the charge, Meta said earnings would have been $7.40 per share, while operating income would have reached $22.88 billion instead of the reported $12.99 billion.

Also read: Intel stock surges on JPMorgan upgrade after 14% weekly plunge

The company also issued an upbeat fourth-quarter revenue forecast of $73 billion to $76 billion, compared with Wall Street’s estimate of $73.75 billion.

Perhaps what’s rattling investors is that Meta lifted the lower end of its 2026 capital expenditure forecast to $135 billion to $145 billion from a previous range of $130 billion to $145 billion.

It also raised its full-year expense outlook to $175 billion to $179 billion.

“Our AI investments are already delivering substantial incremental advertising revenue through stronger engagement, better recommendations, and more efficient monetization,” the company said in a statement. 

Meta is pushing deeper into AI agents. In September, it launched Muse, a personal AI agent that can take actions on users’ behalf and help with daily tasks such as sending emails. The app has become one of the most popular personal AI agents since its debut.

Meta expects strong revenue growth to continue through 2027, although it has not provided a full-year revenue forecast. Still, Meta warned that returns from its newer AI initiatives will take longer to materialize than those from its core advertising business.

Related: How much META stock does Facebook founder & CEO Mark Zuckerberg own?

Wall Street analysts reset Meta stock price targets

At least 15 analysts raised their price targets on Meta following its earnings report, according to The Fly.

JPMorgan analyst Doug Anmuth raised his price target to $940 from $920 and maintained an Overweight rating.

The firm pointed to Meta’s strong advertising business, supported by AI-driven improvements and video growth. JPMorgan also sees early momentum in Muse and enterprise products such as Meta Model API and Muse Code.

UBS also raised its price target to $892 from $884 while maintaining a Buy rating.

More Meta:

• Meta Muse gives AMD and Intel investors a reason to cheer
• Meta and Microsoft just sent employees a memo about Anthropic
• Bank of America sends a stark message to Meta stock investors
• History of Meta: Company timeline, facts & milestones

The firm said Meta’s AI investments are beginning to improve advertising performance and product adoption. Faster monetization of Muse could provide additional upside to revenue forecasts for 2027 and 2028.

However, not all analysts became more bullish on Meta stock.

Wells Fargo analyst Ken Gawrelski lowered his price target to $995 from $1,000 while maintaining an Overweight rating.

The firm acknowledged Meta’s strong advertising performance but flagged uncertainty over the timing and size of potential capital raising, along with its 2027 operating expense and capital expenditure plans.

Meta stock traded at $651.47 as of writing, down 1.35% year-to-date. The S&P 500 Index gained 12% over the same period.

Related: Cathie Wood sells $16 million of surging tech stock

The company supplied the electrical and cooling equipment that keeps high-density data centers running. Nvidia’s newest AI system requires large amounts of both, making those systems part of the buildout that has to be completed before expensive GPUs can start processing customer workloads.

Vertiv estimates that bringing the infrastructure online roughly 10 weeks earlier could produce up to 30% more cumulative AI token output during the first 12 months, assuming the same amount of installed computing capacity.

The figure comes from a modeled scenario. Vertiv’s model compares 317 productive days under the new approach with 245 days under a conventional process. Actual results will vary by project.

Vertiv (VRT) shares rose about 4% to around $212.51 in midday trading on Oct. 21 after closing at $203.56 the previous day.

The announcement comes a week before Vertiv reports third-quarter results on Oct. 28.

Vertiv’s model cuts 120 days of work to 48

Vertiv’s estimate breaks the infrastructure process into two stages.

• Design, configuration, and virtual testing: about 30 days under a conventional process, versus three days using pre-engineered designs.
• Power and cooling deployment: about 90 days conventionally, versus 45 days using factory-built infrastructure.

Together, those two stages fall from about 120 days to 48 days in Vertiv’s model. 

Also read: Broadcom fought AT&T over VMware: Now a telecom giant is all in

The first step happens before equipment reaches a construction site.

Instead of designing the power and cooling systems for each AI facility from scratch, engineers can start with reference designs based on Nvidia’s Vera Rubin DSX AI Factory architecture. The proposed setup can then be tested in a digital model before construction begins.

The system also uses technology from the French industrial software company Dassault Systèmes to create a virtual model of the facility, allowing engineers to test how computing, power, and cooling systems work together before construction begins.

After the design is approved, Vertiv will shift more assembly and testing into a factory. Larger sections of the power and cooling infrastructure can arrive at the data center site preassembled rather than being assembled piece by piece during construction.

This is important for AI operators. GPUs generate revenue only after the facility has sufficient power, cooling, and networking to run customer workloads. Shortening the infrastructure schedule gives operators more days to use the computing equipment they have already paid for.

Vertiv CEO Giordano Albertazzi has described the goal as reducing the time from a project’s requirements to productive AI capacity.

The company has been building its relationship with Nvidia around the same constraint. Vertiv has developed power and cooling designs around Nvidia’s newer systems, including infrastructure capable of handling increasingly power-dense server racks.

A 41-megawatt Norway project is applying the approach

Data-center operator Polar Data Centers is using Vertiv’s factory-built deployment approach for the first 41-megawatt phase of an AI data center in Herøya, Norway.

The site is designed to scale to 100 megawatts. Vertiv will provide prefabricated power and thermal modules along with overhead infrastructure for power distribution, liquid cooling, and networking.

The project gives Vertiv a real site where the deployment approach can be used, but neither Vertiv nor Polar has reported that the Norway facility has already achieved the 10-week time saving or 30% token-output increase in Vertiv’s model.

How much time can actually be saved also depends on problems outside the equipment supplier’s control.

A data center can have preassembled cooling and electrical equipment ready to install and still wait for enough electricity to reach the property.

That problem hit Australian data center operator NextDC on Oct. 21.

NextDC shares plunged 41.7% after delays to electricity-network upgrades at its M3 Melbourne facility forced about A$140 million in contracted customer revenue to be pushed into a later period. The company said the next increase in grid supply at the site is now expected in 2028.

Microsoft ran into an even longer delay in Britain.

The company pulled out of an agreement to anchor an Nscale data center in Loughton, England, after the site’s expected power connection was pushed into the 2030s, Reuters reported. The planned facility was supposed to house 23,040 Nvidia GB300 GPUs and provide 50 megawatts of AI computing capacity.

Those cases involve grid connections that Vertiv’s factory-built deployment model cannot remove on its own. They also show why data-center companies are spending more money on ways to secure electricity sooner.

Vertiv is moving closer to the data center’s power source

Vertiv agreed in September to acquire UtilityInnovation Group for about $1.45 billion in upfront cash, adding technology designed to help data centers source and manage power before it reaches server racks.

UtilityInnovation Group builds microgrid controls, coordinates on-site generation, and designs so-called behind-the-meter power systems. Those systems can enable a data center to use local generation or energy storage when the utility grid cannot quickly provide enough electricity.

More Artificial Intelligence:

• AI’s biggest business opportunity may not look like AI
• AI racks are outgrowing old power: 2 chipmakers have a strategy
• BofA splits software into winners and losers as AI agents arrive

The acquisition pushes Vertiv further upstream from the cooling systems and power-distribution equipment it already sells inside data centers.

Its existing equipment helps move electricity from the point where power enters a facility down to individual server racks. UtilityInnovation Group would add tools for deciding where that electricity comes from and how on-site generation can fill gaps when the grid is constrained.

Vertiv has also been expanding production capacity for equipment used in high-density AI facilities and developing systems around newer Nvidia rack designs. The company describes the broader strategy as connecting infrastructure from the power source down to the chip.

Vertiv’s financial growth already reflects the AI infrastructure buildout. Second-quarter revenue rose 24% from a year earlier to $3.27 billion, and management raised its full-year outlook.

The company’s next update comes quickly. Vertiv is scheduled to report third-quarter earnings before the market opens on Oct. 28, followed by an 11 a.m. ET conference call.

Polar’s 41-megawatt Norway project will provide an early look at how Vertiv’s factory-built approach works outside a company model. The remaining question is how much of the proposed 10-week savings survives once construction, permitting, and access to electricity are factored into the schedule.

Related: Goldman Sachs sends a stern warning to stock market investors

That’s what she’s been doing with Palantir Technologies (PLTR), selling shares as the stock rallies ahead of its third-quarter earnings report.

Last year, the flagship Ark Innovation ETF gained 35.49%, far outpacing the S&P 500’s return of 17.88% in the same period. So far this year, Wood’s flagship Ark Innovation ETF (ARKK) is up 11.05% as of Oct. 21, while the S&P 500 surged 13.32%, Yahoo Finance data shows.

Wood gained a reputation after the Ark Innovation ETF delivered a rosy 153% return in 2020. But her style also brings painful losses in bearish markets, as seen in 2022, when the ETF tumbled more than 60%.

Those swings have weighed on Wood’s long-term gains. As of Oct. 21, her Ark Innovation ETF has delivered a five-year annualized return of -6.36%, while the S&P 500 has an annualized return of 11.26% over the same period, according to data from Morningstar.

Cathie Wood remains optimistic about AI and tech revolution

Wood focuses on high-tech companies across artificial intelligence, blockchain, biomedical technology, and robotics. She believes these businesses have strong growth potential, but their volatility often causes fluctuations in the Ark’s funds.

Over the decade ended 2025, the Ark Innovation ETF wiped out nearly $5 billion in investor wealth, according to a report by Morningstar’s analyst Amy Arnott. That made it the fourth-biggest wealth destroyer among mutual funds and ETFs in the ranking. 

Wood defended her investment strategy after an audience member at a recent summit questioned ARK’s performance compared with the Invesco QQQ Trust (QQQ).

Related: Cathie Wood buys $11.8 million of tumbling tech stock

“I welcome the question. Investors deserve to understand both their returns and the decisions behind them,” Wood said in a post on X on Oct. 5.

She argued that the ARK Innovation ETF (ARKK) and QQQ are built differently. QQQ tracks the Nasdaq-100 Index, while ARK actively invests in companies it believes are driving disruptive innovation across industries including healthcare, financial services and transportation.

Wood has long been optimistic about AI, which she sees as a major driver of productivity, economic growth, and corporate profits in the years ahead. She has also pushed back against recent fears that AI could pose an existential threat to humanity.

In September, Wood reposted a post by David Sacks on X, saying he makes a good case that the “AI will kill humanity” headlines were orchestrated.

Meanwhile, she acknowledged that AI, like other technologies, can be used for harmful purposes. She said people such as Elon Musk who highlight AI’s potential risks are “doing us a great service,” adding that “half of the solution—including AI—is understanding the problem.”

In August, Wood said U.S. corporate profits remain unusually strong, with domestic profits before tax at 13.2% of GDP, a level she said is near multi-decade highs. 

Some of that strength came from the massive monetary and fiscal stimulus during the pandemic, but Wood believes another factor is helping sustain margins today: companies are leaning into AI and productivity gains to protect them.

“I think we’re still early in seeing how far that can go,” she said, adding that companies that use AI effectively will “separate themselves from the ones that don’t.”

Not all investors agree with Wood’s optimism. Over the past 12 months through Oct. 20, the Ark Innovation ETF saw roughly $1.64 billion in net outflows, according to data from ETF research firm VettaFi. 

Cathie Wood sells $16 million of Palantir stock

On Oct. 21, 20, 19 and 15, Wood’s Ark funds sold a total of 79,290 shares of Palantir Technologies Inc. (PLTR), according to Ark’s daily trading information sent to TheStreet. These shares were worth about $16 million based on the latest trading price of $201.72.

Palantir shares have rallied this week after falling more than 11% last week amid a broader tech selloff.

On Oct. 19, Palantir announced a partnership with Knox Systems to expand the use of its AI software across U.S. government and defense agencies. The stock also benefited from a rebound in tech shares and optimism ahead of its third-quarter earnings scheduled for Nov. 2.

Related: Cathie Wood buys $17.3 million of surging AI stock

Mizuho analyst Gregg Moskowitz recently raised his price target on Palantir to $245 from $215 while maintaining an Outperform rating, according to The Fly.

The firm updated its software stock targets ahead of third-quarter earnings. Moskowitz said Mizuho’s channel checks were “very healthy overall,” pointing to strong public cloud spending, consumption trends, and continued AI adoption.

On Oct. 8, Goldman Sachs analyst Gabriela Borges upgraded Palantir to Buy from Neutral with a $230 price target. The analyst expects Palantir to outperform through 2027, citing growing demand for sovereign AI, customized applications, and the company’s expansion into specific industries.

Borges said Palantir’s total addressable market “may be setting up for another step function change in depth.”

The firm also noted that Palantir was trading at a discount to companies with similar growth rates after underperforming earlier this year.

In August, Palantir stock saw a single-day rally of 29.45% on August 4, after it posted rosy second-quarter earnings and outlook. 

Also read: CoreWeave stock plunges 12%, but Wall Street analyst says buy the dip

Adjusted earnings came in at 41 cents a share for that quarter, ahead of analysts’ estimates of 35 cents. Revenue jumped 93% year over year to $1.94 billion, beating expectations of $1.80 billion.

“To my knowledge, no businesses at our scale has even grown half this much,” Palantir’s CEO Alex Karp said in a CNBC interview. He also predicted that the strong growth will likely “go on for at least another 18 months.”

Palantir also raised its full-year revenue guidance to between $8.15 billion and $8.16 billion, up from its previous outlook of $7.65 billion to $7.66 billion.

Wood has been active in trading Palantir stock. She greatly increased her position in 2024 and sold many of those shares in 2025 as the stock climbed. In June, she bought back some shares. Now she’s selling again, likely locking in some gains after the recent rally.

Palantir is not among the top 10 holdings in the Ark Innovation ETF. 

Top 10 Holdings in the Ark Innovation ETF by weight as of Oct. 21, 2026:

• Tesla (TSLA) – 10.22%
• SpaceX (SPCX) – 6.47%
• Tempus AI (TEM) – 5.13%
• Circle Internet Group (CRCL) – 5.11%
• OpenAI Group – 4.44%
• Coinbase Global (COIN) – 4.31%
• CRISPR Therapeutics (CRSP) – 4.25%
• Shopify (SHOP) – 3.41%
• Twist Bioscience (TWST) – 3.13%
• Robinhood Markets (HOOD) – 3.07%

Other than buying Palantir shares, Wood’s latest trades on Oct. 21 included selling Amazon (AMZN), Shopify (SHOP), Teradyne (TER), Elbit Systems (ESLT), DraftKings (DKNG), and Personalis (PSNL).

She also bought shares of Pacific Biosciences of California (PACB), Intuitive Surgical (ISRG), X-Energy (XE), Oklo (OKLO), Netflix (NFLX), Scribe Therapeutics (SCTX), Iambic Therapeutics (IAM), and Intuitive Machines (LUNR).

Related: Analysts revamp Super Micro stock price targets after 10% surge on revenue boost

Digital wallets, payment apps and connected banking services have stripped most of the friction from everyday transactions. Moving money is faster, more accessible and more integrated into daily life than it was a decade ago.

That shift has produced a problem the industry has not resolved: how much financial information people give up each time they use those services.

Also read: Kevin Warsh sends a strong warning sign on rates to Wall Street

Every digital transaction can reveal something about the person behind it. Spending patterns can indicate where someone lives, which businesses they frequent and how much they earn.

When banks, payment processors, digital wallets and data brokers each hold portions of that information, individual transactions can combine into a profile that travels further than most customers realize.

The convenience comes with a data trail

The risks compound with routine activity. A single payment reveals limited information. A record of hundreds of transactions, drawn from multiple services, tells a more complete story.

Ermo Eero, managing director of IronWallet, told TheStreet that the exposure builds gradually rather than through any single event. “The biggest risk isn’t one dramatic breach. It’s the quiet build-up of data: every fast, one-tap payment leaves a trail that wallets, payment processors, banks and data brokers can combine into a detailed picture of someone’s life, including where they go, what they buy and what they earn.”

Open banking connections and third-party integrations extend that trail. Each link that makes it easier to move data between platforms also introduces another point where that data can be retained, shared or used in ways customers did not specifically anticipate.

Financial institutions collect transaction data for legitimate reasons: detecting fraud, verifying payments and meeting regulatory requirements. The problem is that the same infrastructure serving those functions can expose information that both individuals and institutions would prefer to keep private.

Consent agreements typically disclose how data is used. Accepting terms does not mean customers understand every downstream use of their information. That gap is likely to widen as financial services become more automated.

When financial records are public by design

Some digital financial networks record transactions on shared ledgers visible to any participant. That transparency supports verification. It also creates confidentiality problems that conventional banking systems were not built to solve.

An account identifier on such a network does not automatically reveal its owner. But once it is linked to a real person or organization, its full transaction history becomes accessible. For businesses, that can expose trading positions, treasury movements and commercial relationships to competitors.

Exposure matters more as financial activity expands across shared digital infrastructure. Whether institutions adopt these systems more widely may depend in part on their ability to protect sensitive information while using them.

Guy Itzhaki, CEO of Fhenix, told TheStreet the issue is about control, not concealment. “Privacy is increasingly becoming less about ‘having something to hide’ and more about having basic control over who can see what.”

Verifying without revealing

Several technologies are being developed to let financial systems run checks and process data without exposing the underlying information. Fully homomorphic encryption, multi-party computation and zero-knowledge proofs each approach the problem differently.

Fully homomorphic encryption allows computation to run directly on encrypted data. Multi-party computation enables computations among multiple participants without requiring each party to have full access to all inputs.

Zero-knowledge proofs allow someone to demonstrate that a condition is met, such as having sufficient funds or having passed an identity check, without disclosing the underlying details.

In financial services, those capabilities could allow systems to verify transactions and assess positions while limiting what each participant in a transaction can see.

More Tech:

• Meta and Microsoft just sent employees a memo about Anthropic
• Microsoft cutting off popular software with no extension Oct. 13
• Apple’s biggest AI advantage may face new privacy challenge

“A financial system should be able to execute rules correctly without requiring every underlying input to be public,” Itzhaki added.

On Oct. 21, Fhenix launched its CoFHE encrypted-computation coprocessor on Ethereum and Arbitrum, allowing smart contracts to process financial data without exposing the underlying values. The release includes a Consensys Diligence security audit and two design partners, Fluton and Alpha Engine, developing confidential financial applications on the network.

The technology still carries trade-offs in cost, performance and implementation complexity. Privacy protections also need to work alongside financial regulations. Anti-money laundering requirements and customer identity checks remain in force regardless of how underlying data is encrypted.

Building privacy into the product, not onto it

Technology does not fix the data exposure problem on its own. Financial services providers also have to decide what information they collect, how long they retain it and who can access it.

Those decisions are most effective when made at the product design stage. Collecting less data by default, shortening retention periods and giving customers usable controls over their information can reduce exposure before any encryption layer is introduced.

Eero said the choices providers make about data will shape how consumers and regulators judge them. “In finance, trust is the real currency, and a service that is convenient but careless with data will eventually lose both its customers and its regulators’ goodwill.”

For technology companies, building privacy into the underlying product rather than treating it as a compliance feature could open access to institutional customers. Banks and asset managers need to demonstrate control over what is visible and to whom before committing sensitive financial activity to shared infrastructure.

Digital payments will continue to get faster and more connected. The industry’s next question is not whether to collect transaction data, but how to limit what gets exposed and to whom.

Related: Michael Burry buys another beaten-down forgotten fintech stock

The feature is part of Apple’s iOS 27 update, which rolled out to iPhones on Sept. 14. It allows customers to use the same phone number across two compatible iPhones running iOS 27, and it requires carrier support.

With iPhone Handoff activated, customers can choose a primary and secondary iPhone. Their phone number automatically switches to whichever phone they unlock.

Calls, voicemail, messages, and “Find My” location also automatically move to whichever iPhone customers are using. 

Verizon faces consumer backlash for $10 iPhone Handoff fee

T-Mobile currently charges its U.S. customers $5 per month for iPhone Handoff; however, Verizon has recently announced it will charge more. 

According to a recent update on its website, Verizon said that iPhone Handoff will be “coming soon” and will cost customers $10 per month.

Some consumers took to social media platform Reddit to express frustration over the $10 fee, with some calling it a “rip-off.”

“What a rip-off. Just transfer the eSIM yourself in settings. It takes 30 seconds,” wrote one consumer.

“That’s basically the same price for an additional line on some plans. WTF are we doing here,” wrote another. 

Related: T-Mobile adds monthly fee to a new iPhone feature for customers

“I didn’t expect this to be something free; but $10 is ridiculous. At most it should be a couple of dollars a month,” wrote another consumer. 

In the U.S., the feature is currently available only to T-Mobile customers. AT&T has not announced plans to offer iPhone Handoff yet. Outside the U.S., Deutsche Telekom customers in Germany have access to the feature, and soon, EE customers in the United Kingdom will as well.

If Verizon customers choose to pay monthly for iPhone Handoff, they should be aware that the feature has a few restrictions, which Apple highlights on its website.

First, the feature does not allow a single phone number to operate on two iPhones at once. Second, while payment cards and car keys stored in Apple Wallet can be used on both iPhones, corporate badges, IDs, and some transit passes may be restricted to one device at a time.

Third, Apple recommends connecting an Apple Watch to the iPhone they use most, as iPhone Handoff does not automatically move the watch connection between devices. 

Lastly, for Dual SIM users, the feature requires that both eSIMs be supported by compatible carriers. When the setup meets those requirements, unlocking either iPhone triggers both eSIMs to switch to that device.

Verizon’s latest fee follows a string of price increases

Verizon’s decision to charge $10 per month for iPhone Handoff comes during a time when it has been asking its customers to pay more in recent months.

For instance, in May, it raised the monthly price of its Unlimited Ultimate wireless plan by $5 and increased the price of its Netflix and HBO Max bundle by $3. In September, Verizon also raised prices for its Disney+, Hulu, and ESPN+ perks.

Amid these changes, a recent survey by Market Force Information found that Verizon customers spend an average of $160 per month on wireless service, the highest among rival carriers. 

More Verizon News:

• Verizon scales back a perk that keeps prices low for customers
• Verizon plans to discontinue a phone service customers rely on
• Verizon hikes prices on multiple discounted offers for customers

Against this backdrop, the survey data also revealed that Verizon’s customer loyalty isn’t as strong as its competitors’. Verizon falls below the 25th percentile with a customer loyalty index (CLI) score of 3.9. Consumer Cellular, Cricket and Mint Mobile claimed the top spots in this category with scores of 4.60, 4.56 and 4.46, respectively. 

Also, 22% of Verizon customers are considering switching carriers, with cost of service being the top reason for exiting. 

In a February Light Reading report, Recon Analytics analyst XJ Wang warned that raising prices has consequences for wireless carriers. 

“Carriers cannot unilaterally raise prices without risking subscriber losses,” wrote Wang. “The market has become commoditized.”

Related: Verizon launches generous customer perks after raising prices

That habit has been worth a fortune to the cable companies that wire American homes. For decades, the choice in most neighborhoods came down to one cable provider, and it priced its plans accordingly.

That grip is slipping. Charter Communications (CHTR), which sells internet under the Spectrum brand, lost 172,000 broadband customers in the second quarter, Light Reading reported July 24.

Comcast (CMCSA) shed 167,000, according to Broadband Breakfast on July 23. Optimum lost another 40,000, Fierce Network reported Aug. 10. The company said it was also exiting some low-density markets in its West footprint, the same report showed.

Charter has mostly pinned its losses on fiber and fixed wireless rivals rather than satellites, Light Reading reported.

Charter’s average residential broadband customer paid $70.16 a month, down 1.5% from a year earlier, Light Reading reported. Now Elon Musk’s SpaceX (SPCX) has put a number right beside it.

On Oct. 20, SpaceX said Starlink’s average consumer paid $69 a month as the service added 1.6 million subscribers in the third quarter. Total revenue jumped 161% to $14.6 billion, according to Quartz.

SpaceX’s $14.6 billion quarter ran on two very different engines

The headline number came mostly from artificial intelligence. SpaceX’s Super Intelligence segment brought in $7.6 billion, up 830%, while its connectivity business grew 69% to $5.2 billion, Quartz reported Oct. 20.

The company also swung to a $4.4 billion profit from a $1.1 billion loss a year earlier. Revenue topped the $12.87 billion analysts expected, and adjusted earnings of 35 cents a share beat the 13-cent forecast, according to Investing.com.

Shares rose 6.6% after hours to $157.59 on Oct. 20, the same report showed. Check the live quote before you act, because that move was still settling on Oct. 21.

In my analysis, the AI number grabs attention, but connectivity is the part that touches your household budget. That is the business now competing for the same monthly payment your cable company counts on.

Why Starlink’s subscriber gain matters for cable companies

The comparison needs one caveat up front.

Starlink’s 1.6 million additions are global consumer subscribers, which Chief Financial Officer Bret Johnsen said came mostly from “more developed markets.” The cable losses are U.S. figures.

More Tech Stocks:

• Meta’s $18 billion teen-safety deal is about to hit earnings
• AI fraud may be exposing a blind spot Wall Street can’t afford
• Intel stock surges on JPMorgan upgrade after 14% weekly plunge

Even with that gap, the trend is hard to ignore. Starlink ended June with 12 million subscribers, as TheStreet reported in Bank of America’s wireless reality check on SpaceX, and reached 13.6 million by the end of September.

Charter had 29.38 million broadband customers at the end of June, Light Reading reported. That means Starlink’s consumer base is already close to half the size of the company behind Spectrum.

The two bills are also moving in opposite directions. Starlink’s average bill rose 5% from the prior quarter, Johnsen said on the call, while Charter’s fell 1.5% from a year earlier. I ran the math, and the gap between them is now $1.16 a month.

Here is what the latest round of filings and calls shows about where this goes next:

• Starlink Mobile: The service passed 22 million monthly users, up 37% from the prior quarter, President Gwynne Shotwell said.
• New spectrum: SpaceX is buying Grain Management’s nationwide 800 MHz spectrum for Starlink Mobile, Telecompetitor reported Oct. 19.
• Carrier ambitions: The spectrum “will pave the way for Starlink Mobile to become a major mobile carrier in the U.S.,” Shotwell said.
• Cable keeps the door open: On satellite wholesale deals, Charter CEO Chris Winfrey said, “If it’s going to be material, it’s certainly something we would think about,” Fierce Network reported.

Spectrum and Comcast bosses say their networks still win

The cable companies are not conceding. Asked whether Starlink would be the main rival in urban and suburban markets, Winfrey said, “Do I think it will be the competitive solution for those? No,” according to Broadband Breakfast.

Comcast was just as blunt. “In reference to fixed wireless and satellite, our network far exceeds what they have,” said Steve Croney, chief executive of Comcast’s connectivity business, the same report showed.

Related: T-Mobile pushes back on SpaceX’s bid to take its customers

Winfrey did concede one thing. Charter’s rural builds were increasingly reaching homes that already had Starlink, which slowed its sign-ups there, Broadband Breakfast reported.

I think the cable executives have a point in dense neighborhoods, where a coaxial or fiber line carries far more traffic than a satellite beam. The risk sits at the edges of their maps, in rural homes where Starlink is now there first.

How the satellite race could lower your internet bill

Do not expect Starlink to start a price war. It raised consumer plan prices by $5 to $10 a month in May, as TheStreet reported in its look at a troubling Starlink pricing trend, and its average bill rose again last quarter.

Your savings come from the pressure Starlink and fiber put on everyone else. Optimum just rolled out a customer appreciation program with giveaways and perks, Telecompetitor reported.

That gives you bargaining power at renewal time. Call your provider, name the alternatives in your ZIP code, and ask for the promotional rate.

Musk is aiming much higher than your monthly plan. He said SpaceX sees “a clear path to several hundred billion dollars a year in connectivity revenue.”

For scale, Charter’s $13.52 billion second quarter works out to about $54 billion a year, based on Light Reading‘s figures. Musk’s target is several times Charter’s entire business, and the phone service tied to that new spectrum isn’t expected until early 2028, Shotwell said on the call.

The next test arrives fast. Comcast reports third-quarter results on Oct. 22, and Verizon follows on Oct. 26.

Watch the broadband line in both reports. If the losses keep coming while Starlink keeps adding, your next renewal call gets a lot more interesting.

Related: SpaceX has a plan to make your phone carrier obsolete

China has just exacerbated the issue.

ByteDance, Alibaba (BABA) and Tencent have placed orders for more than 3 million Nvidia RTX PRO 5500 graphics cards, carrying a combined list-price value above $35 billion, according to Reuters.

However, Nvidia may not be able to do anything like that.

Chinese orders have exceeded 6 million cards, and Nvidia expects to produce only 3 million RTX PRO 5500 units in 2027. Reuters reported that some customers will receive only 5%–10% of their requests.

This does not guarantee Nvidia $35 billion in revenue.

It could disclose something more significant.

China’s largest IT businesses continue to line up for Nvidia gear, even if it isn’t the best, despite years of U.S. regulations intended to restrict China’s access to cutting-edge artificial intelligence processors.

Investors may discover that Nvidia’s China issue is more about determining which customers it can truly serve than about attracting new ones.

China still wants Nvidia chips

Nvidia’s flagship AI processor is not the RTX PRO 5500.

AI inference, or the work a trained AI model does when it reacts to a prompt, creates content, or finishes a job, may be handled by this powerful workstation graphics processing unit.

Because of its capabilities, it is very helpful in China.

U.S. restrictions prevent Nvidia from freely selling some of its most advanced products there. The B300 cannot currently be sold to Chinese customers, while H200 shipments require individual U.S. government approval, Reuters reported.

As a result, Chinese IT firms are vying for access to Nvidia hardware.

They seem to want a lot of it, too.

Together, ByteDance, Alibaba, and Tencent have asked for almost 3 million RTX PRO 5500 cards. China’s overall demand exceeds Nvidia’s projected global output for the next year by more than two times.

Customers should get samples in December, and larger shipments will start in early 2027.

The figures challenge the conventional narrative about China.

Clearly, export limitations have made it more difficult for Nvidia to do business there. However, they haven’t eliminated the need for Nvidia in China.

Rather, it seems that consumers are prepared to go down on Nvidia’s product stack in order to get the remaining processing power.

Also read: How much Nvidia stock does CEO Jensen Huang actually own?

Nvidia can’t make every chip customers want

Strong demand does not automatically mean Nvidia should make millions more RTX PRO 5500 cards.

There is value in manufacturing capacity.

At the same time, Nvidia is expanding its more advanced AI systems for clients like Microsoft (MSFT), Amazon (AMZN), Alphabet (GOOGL), and artificial intelligence firms constructing ever-larger data centers.

As a result, there is a trade-off.

More Nvidia:

• Nvidia stock flashes massive signal as Wall Street leans in
• Nvidia’s AI boom revives a Warren Buffett warning for investors
• Nvidia puts $500 billion on the table, challenges banking as we know it
• History of Nvidia: Company timeline and facts

The scale of Nvidia’s existing business makes the decision easier to understand.

Nvidia generated $89 billion in Data Center revenue in its fiscal second quarter, an increase of 117% from a year earlier. It forecast roughly $108 billion in total revenue for its fiscal third quarter.

One aspect of this approach is particularly intriguing.

According to Nvidia, their prediction does not account for income from data center computing in China.

Because of this, the most recent Chinese demand may be more significant than the $35 billion headline implies.

These Chinese sales are not necessary for Nvidia to meet its current projections.

China may generate more income if it can meet some of the increased demand without compromising the production of more lucrative goods abroad.

However, investors should understand this difference.

A shipment is not the same as an order request. Revenue is not always realized for a shipment. Furthermore, it doesn’t seem feasible that Nvidia will fulfill every single request.

The stronger indication is the demand itself.

Also read: Who owns Nvidia? Top insiders & institutional investors

Nvidia’s China problem comes with a catch

Nvidia cannot expect Chinese businesses to wait indefinitely.

Every chip it is unable to sell offers local rivals another chance.

Huawei is one of the businesses attempting to lessen China’s reliance on American semiconductor technology as a result of U.S. limitations.

The most recent orders are so strange because of this.

Since the RTX PRO 5500 cards are Nvidia’s most cutting-edge technology, ByteDance, Alibaba, and Tencent are not asking for millions of them.

Despite the concessions, they still want them.

This implies that Nvidia still has a client base that firmly favors its hardware and software ecosystem, even in situations where access to its top products is limited, something that rivals have found difficult to match.

Although the $35 billion figure may draw attention, it does not represent a $35 billion sales projection.

Six million could be a more telling figure.

Well over 6 million RTX PRO 5500 cards are apparently wanted by Chinese consumers, although Nvidia anticipates producing around half of that quantity for the global market next year.

That is what most chipmakers would refer to as unusual demand.

It adds another layer of complexity for Nvidia.

Washington continues to demand more chips from Nvidia than it intends to manufacture in one of the places where it has made it most difficult for the company to sell its technology.

China’s desire for Nvidia is now clear.

Related: Nvidia dodged the AI selloff, and Jensen Huang has a theory why

Back in May 2025, Wedbush analyst Dan Ives projected more than $3,000 as a possible price for an iPhone made in the U.S., as reported by Business Insider, compared with $799 for a standard iPhone 16.

Mark Cuban thinks that framing misses the point. In his view, the real contest is machine against machine, and he is not just speaking in theory.

His drug company opened a Dallas plant that uses “advanced robotic and AI computer vision technology,” which gives his argument a little more weight than a typical social media hot take.

Mark Cuban’s drug plant puts his automation argument to the test

Cost Plus Drugs was built to strip out the middlemen. Cuban co-founded the company in 2022, and it sells generic medicines at the manufacturer’s cost plus a flat 15% markup, along with pharmacy and shipping fees, skipping pharmacy benefit managers entirely, as reported by Fierce Healthcare.

The company then pushed into making drugs itself. Its 22,000-square-foot Dallas facility, built for $11 million, started with commercial batches of epinephrine and norepinephrine for intensive care patients, with pediatric chemotherapy drugs to follow. Those are cheap, essential medicines that often run short.

Also read: Mark Cuban has strong words on taxes and wealthy Americans

Automation is the selling point. CEO Alex Oshmyansky said the plant could switch from one drug to another “in principle within four hours,” letting it make whatever product is running short, and he described the generic drug supply chain as extremely brittle, as reported by Fierce Healthcare.

That kind of flexibility is exactly what Cuban means when he says better robots and software could tilt the math back toward American factories.

Skeptics see bigger hurdles than fees. Not everyone thinks robots solve the problem. Rosenblatt analyst Barton Crockett has said U.S. iPhone production is a long way off, naming construction lead times, Apple’s exacting quality bar, and weak American demand for that kind of work.

Cuban’s read is that smarter machines eventually close most of that distance.

China already has a head start on robots

The trouble with Cuban’s plan is that China is running the same race, and faster. Morgan Stanley analysts expect humanoids and robots to become the next key engine of China’s export machine over the coming five to 10 years, comparing the push to the rise of its electric vehicle industry, as reported by Bloomberg.

The early numbers back that up. Morgan Stanley estimated Chinese manufacturers made up roughly nine in 10 of the humanoid robots shipped worldwide last year, while American and Japanese rivals were still mostly building prototypes.

That is the gap Cuban says America has to close.

Tesla is America’s loudest bet on catching up in robotics. The carmaker raised its 2026 capital spending guidance to more than $25 billion, funding a major push into AI automation, with a significant share going to convert Model S and Model X capacity into Optimus robot lines, according to The Motley Fool. Analysts expect Optimus revenue to remain limited until 2029.

China’s robot boom has its own cracks, though. Unitree, the country’s best-known robotics company, has dropped 45% from its all-time high since listing in Shanghai, and regulators have reportedly made it harder for humanoid makers to go public.

At least six more Chinese humanoid makers are still said to be lining up listings. But a head start is not the same as a finished race.

America’s AI chip edge is getting squeezed

Cuban’s other condition, staying ahead on AI, looks shakier inside China. Nvidia’s share of the country’s data center AI chip market slid to 34% in the first half of 2026, while Huawei climbed to 28%, with U.S. export controls limiting Nvidia’s supply, according to Reuters.

That erosion started before this year. In 2025 Nvidia still commanded about 55% of China’s AI accelerator market. Bernstein now puts its share at roughly 8% by year-end, according to TheStreet, a collapse driven in large part by Beijing’s push to replace foreign chips with homegrown ones.

More Artificial Intelligence:

• AI’s biggest business opportunity may not look like AI
• AI racks are outgrowing old power: 2 chipmakers have a strategy
• BofA splits software into winners and losers as AI agents arrive

Memory is the next front. CXMT says it has made key progress on a new memory cell design and plans to launch a server memory product using it by the end of 2026. The layout squeezes each cell into about a third less space than mainstream chips at the same line width.

Wall Street noticed. Micron shares fell about 3% on Oct. 12 and were headed for a weekly drop of more than 10%, partly on worries about competition from CXMT.

Some analysts argued the selloff was overdone, pointing out that CXMT has not yet matched the high-bandwidth memory used in AI servers.

Washington plans to guard its AI lead

The White House sees the race the same way Cuban does. Speaking to reporters on Sept. 26, Trump said he would rather not combine American and Chinese AI efforts “because we’re leading by a lot,” adding that “when you’re leading, you don’t open it up to each other,” according to Al Jazeera.

That leaves the tougher half of Cuban’s formula for companies to solve. Building factories in America still takes patience.

When work began on Cost Plus Drugs’ Deep Ellum facility in 2021, it was scheduled to be finished by late 2022. Its first drugs did not come off the line until March 2024.

Still, his core idea is easy to follow. If American robots, software, and AI can match or beat what China puts on its factory floors, labor costs matter far less. Cuban has even predicted that within five years there would be no more shortages of sterile injectable drugs.

Whether the rest of corporate America follows his lead is the question that will decide the race.

Related: Mark Cuban just sent a strong message to America’s workers

Western Digital makes high-capacity hard drives used by hyperscalers, cloud providers, and other enterprise customers to store large amounts of data.

BofA analyst Wamsi Mohan expects Western Digital’s fiscal first-quarter exabyte shipments to increase 21% from a year earlier, below management’s longer-term target of roughly 25%. Exabyte growth measures the amount of storage capacity the company ships.

At the same time, BofA expects Western Digital’s average price per terabyte to rise 24% from a year earlier. Demand for hard-drive storage continues to exceed available supply, giving manufacturers room to charge more for each unit of storage capacity, according to the bank.

Western Digital shares rose 3.89% to close at $378.51 on Oct. 21.

The company will report fiscal first-quarter results after the market closes Thursday, Oct. 22. BofA expects revenue of $4.22 billion and adjusted earnings of $4.29 per share, slightly above Wall Street estimates of $4.16 billion and $4.11, respectively.

Higher hard-drive prices lift BofA’s earnings outlook

BofA expects tighter HDD supply to keep supporting Western Digital’s prices and profit margins.

The bank forecasts a 56.4% fiscal first-quarter gross margin, slightly above Western Digital’s own guidance range of 55% to 56%. It expects the benefit from higher pricing to continue into the December quarter, when BofA sees gross margin approaching 59%.

Also read: Oracle boss Larry Ellison just suffered a historic wealth collapse

Western Digital’s latest results have already shown how much margins have expanded. Fiscal fourth-quarter revenue rose 44% to $3.75 billion, and adjusted gross margin reached 54.4%, up sharply from a year earlier.

BofA raised its fiscal 2027 earnings estimate to $22.37 per share from $19.17. The increase helped push the bank’s price target to $800.

The valuation itself actually became more conservative. BofA now applies a 16-times earnings multiple, down from 19 times previously. Higher earnings estimates were large enough to push the target higher, even with the lower multiple.

Part of the slower near-term capacity outlook comes from Western Digital’s transition to newer, higher-capacity drives.

The company’s 40-terabyte ePMR hard drive has qualified with two customers and is entering volume production, according to BofA. The bank expects the product transition to help exabyte shipment growth accelerate during the second half of fiscal 2027.

That marks progress from February, when Western Digital said its 40TB UltraSMR ePMR drive was still undergoing qualification with two hyperscale customers, with volume production planned for the second half of 2026.

Toshiba’s expansion could pressure Western Digital’s pricing

Western Digital and rival Seagate Technology (STX) have recently faced a different concern: whether new industry capacity could weaken the pricing power now supporting their earnings.

The stocks sold off earlier this month after reports that Toshiba planned a major expansion in hard-drive production capacity. Investors worried that additional HDD supply could eventually narrow the gap between customer demand and the drives available to ship.

More Tech:

• Meta and Microsoft just sent employees a memo about Anthropic
• Microsoft cutting off popular software with no extension Oct. 13
• Apple’s biggest AI advantage may face new privacy challenge

BofA’s industry checks have yet to show that pressure.

The bank said pricing discussions with hyperscale customers remain constructive, despite Toshiba’s planned expansion. BofA expects strong HDD pricing and margins to continue well into 2028.

Other analysts have also pushed back on the idea that the current HDD cycle is about to end. Morgan Stanley maintained its Overweight rating on Western Digital this week after the recent sell-off, arguing that demand, customer growth, and pricing remain supportive heading into earnings.

BofA still identifies several ways the call could go wrong. A slowdown in cloud or data-center spending, weaker HDD pricing, delays in Western Digital’s higher-capacity products or market-share losses to Seagate could reduce the earnings growth behind its $800 target.

Western Digital’s fiscal first-quarter results will give investors their next read on the two numbers behind BofA’s call: how quickly the company is shipping more storage capacity and how much customers are paying for each terabyte.

Related: SpaceX CEO Elon Musk has 7 words for Delta Airlines after snub

President Donald Trump just asked America’s biggest data centers to learn that lesson at scale. He posted on Truth Social early Wednesday, Oct. 21, 2026. He wrote that new and large data centers are “mandated by me to make their own electricity.”

He also told builders to give or sell any extra power back to the grid and pay host towns a “dividend,” Reuters reported. He pitched it as a way to hold down or even cut power bills for households.

Two companies are already living a version of this plan. Oracle Corp. (ORCL) makes power on site at several of its AI campuses. Bloom Energy Corp. (BE) supplies the fuel cells behind a big part of that effort.

To me, they are the clearest test of whether Trump’s idea works in practice. First, it helps to know why he keeps pushing it.

Also read: Oracle’s ‘force majeure’ warning lands as AI power gets scarce

Why Trump wants data centers to make their own power

The short answer is your power bill. Data centers have turned into a political flashpoint, and the anger now crosses party lines, Reuters noted. The post landed less than two weeks before the midterms, when no candidate wants to defend a higher bill.

Trump has pushed this idea since February. On Wednesday, March 4, 2026, Oracle and six other tech companies joined his Ratepayer Protection Pledge. They promised to build or buy new power and cover grid upgrades.

However, experts told The Associated Press the pledge is likely not enforceable at the federal level. Wednesday’s post dresses that voluntary promise in the language of a mandate.

The bigger question is what happens when a company actually does this. Oracle already has.

Oracle was already making its own power

Oracle’s problem was speed. AI customers wanted capacity now, and grid hookups can take years. So it turned to Bloom, whose fuel cells turn natural gas into electricity without burning it.

Bloom delivered a working system to Oracle in 55 days last year, according to a press release. A master agreement announced Monday, April 13, 2026, lets Oracle buy up to 2.8 gigawatts of Bloom fuel cells, the release said.

More Oracle:

• Oracle layoffs top 2,500 workers as AI spending surges
• Oracle’s ‘force majeure’ warning lands as AI power gets scarce
• Michael Burry just put a date on Big Tech’s AI reckoning, and Oracle’s $664 billion lands in crosshairs

Two weeks later, Oracle said Bloom would fully power Project Jupiter, its giant AI campus in New Mexico. Jupiter reads like an early draft of Trump’s post.

Oracle said it would bear all of Jupiter’s energy costs, and the project committed $360 million to schools, infrastructure and local services. That is the community dividend Trump described, promised months before he asked.

What Oracle could not promise was steady fuel.

Making your own power costs more than it looks

Fuel cells still need fuel. Jupiter’s gas was meant to arrive through a new pipeline called Green Chile, but its start slipped to Monday, February 1, 2027, Bloomberg reported.

So Oracle turned to trucks, hauling compressed gas to Utah and Texas sites while pipelines get built, a separate Bloomberg report found. It is weighing the same fix for Jupiter. East Daley Analytics told Bloomberg that trucked gas costs about four times the price of gas at a major pipeline hub.

Oracle can hardly afford that premium. Its free cash flow was negative $28.7 billion over the past 12 months, according to StockAnalysis. It even invoked force majeure, a contract clause for events beyond its control, to limit its costs if Jupiter runs late.

Here is what I find most telling: Trump’s plan moves the power bill away from households without shrinking it. Shareholders pick it up instead, and the two stocks show who is paying.

Oracle stock carries the bill as Bloom gets paid

Oracle shares have fallen about 53% over the past 52 weeks, according to StockAnalysis. That slide reflects worry about cash, not a shortage of customers.

Even so, the 43 analysts tracked by StockAnalysis rate Oracle a Buy on average, with a mean target near $237. That reads like a bet that the power problems are temporary.

Bloom sits on the other side of the trade. Its shares have more than doubled over the same 52 weeks, according to StockAnalysis, as builders pay up for fast power.

Yet its showcase Oracle project waits on the same delayed pipeline. Bloom reports third quarter results on Thursday, Oct. 22, 2026, and any update on New Mexico could matter more than the headline numbers.

Trump’s plan now runs through a pipeline

Trump’s post pictures data centers with power to spare, selling the surplus back to town. A campus running on trucked gas has no surplus to sell. That gap between the promise and the plumbing is the real story.

The next big decision belongs to the regulators weighing the Green Chile pipeline, not the White House. Their answer will tell every builder what Trump’s model truly costs.

The AI race once turned into a race over who could buy the most chips. Increasingly, it turns on who can get fuel to the front gate.

More Stocks News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off

A combined SpaceX (SPCX) and Tesla (TSLA) would have a market capitalization of at least $3.6 trillion, good enough to make the company the fifth most valuable company in the world after Nvidia (NVDA), Apple (AAPL), Alphabet (GOOGL) and Microsoft (MSFT).

Also read: Elon Musk wants every American kid to own a SpaceX share

That assumes the deal would survive regulatory and political scrutiny. The Trump Administration has close ties to Musk already, but the midterm elections may make the politics of the deal trickier.

SpaceX closed at 7.8% to $159.31. Tesla was up 3.4% to $386.47.

The Wall Street Journal reported in its Oct. 21 editions that Musk ordered SpaceX to slow down a military version of its Starlink mobile telephony networks available to Ukraine. Later in the day, Ukrainian President Volodymyr Zelensky said that U.S. intelligence sharing with Ukraine had been restored, The Washington Post reported.

The move capped the speeds at which Ukrainian hardware could stay connected to the satellite-based network, the report said. That rendered some offensive drones and interceptor drones virtually useless.

A merger of the SpaceX and Tesla has been talked about for some time, but Musk disclosed the two companies’ boards of directors have formed separate committees to study a possible merger. The disclosure came during the question-and-answer section of SpaceX’s third-quarter earnings call on Oct. 21.

Musk, as CEO, can’t be on either board committee for conflict of interest reasons. But that does not preclude either committee talking to him about a deal. But, when questioned about it on the earnings call, he said, “I do think there’s a lot of merit to combining the companies as we are pretty much in the super intelligence singularity.”

SpaceX’s biggest asset is its Starlink business, which is a global satellite system

Related: Meta, Tesla Q3 earnings will be fighting rising AI skepticism

That’s because while I rarely use the streaming service, aside from sometimes watching a stand-up comedy special, it seems like you need to have Netflix just in case they produce something worth watching. In addition, when I travel, my wife has watched a number of series, so it’s probably a good value for us.

Netflix’s co-CEO Greg Peters spoke about price increases during his company’s third-quarter earnings call.

He made it clear that he thinks his company delivers enough to customers that they won’t mind price hikes. Peters noted that members have responded well when the streaming service has raised prices.

“Across all market types where we’ve done price changes, the member response, you mentioned churn engagement usually isn’t really impacted by these things. But that member response is consistent with prior price changes and our expectations,” he said.

His next comments, while they didn’t directly call for a price increase, certainly suggested that the company thinks one would be justified.

Netflix hints at a price increase

Peters made it clear that he thinks Netflix offers strong value to members.

“Again, we think that this is a strong indicator of the value that we’re delivering to members as measured by the quality, the quantity, the variety of viewing that is healthy. Maybe just one also, an external reference point for that value, take the U.S. as an example. Netflix subscribers are paying the least per hour of viewing compared to other SVOD offerings in the country,” he shared.

More Entertainment:

• Bots playing fake AI songs siphoned $8 million from real artists
• Walt Disney’s CEO explains why box office bombs may no longer matter
• Disney+ delivers bad news to millions of subscribers

That comment suggests that Netflix has at least considered raising prices.

Peters went on to provide further justification for why Netflix could be charging more.

“In some cases, they would have to pay more than double per hour for a competitive service. We think that’s a really good marker of how much value we are delivering per price,” he added.

Peters also made it clear that his comments were not just about the U.S.

“Also note that we saw continued healthy retention in every region, churn globally in the U.S. improved year-over-year, so better than forecast. Those things also reinforce the signs of that value and make us increasingly confident in that,” he added.

What does Netflix cost now?

Netflix offers multiple tiers.

• Standard with ads: $8.99 / month
• Standard: $19.99 / month
  • Add 1 extra member for $7.99 / month with ads or $9.99 / month without ads
• Premium: $26.99 / month
• Add up to 2 extra members for $7.99 each / month with ads or $9.99 each / month without ads
  • Source: Netflix

Netflix has also been very active in policing members sharing accounts.

“You cannot share your Netflix account outside of your household, which is a collection of devices connected to the internet at the main residence where you watch Netflix. People who are not in your household can be added as an extra member,” the company shared.

When Netflix does raise prices for existing members, the impact isn’t immediate.

An email is sent one month before the billing date on which a price will increase to notify current members before a rate adjustment takes effect, according to the Netflix Help Center.

The company also notes that plans and pricing may adjust due to “Service improvements, local market changes, inflation, taxes.”

Netflix raises prices regularly

Netflix last raised prices in March. That followed a January 2025 price increase.

RTM Nexus CEO Dominick Miserandino thinks Netflix is getting ready for another U.S. price increase.

“Netflix has figured out that it can raise prices and most people will stick around. That’s a luxury very few companies have. They just raised prices, so I wouldn’t expect another increase tomorrow, but I’d be surprised if they don’t test customers again next year,” he told TheStreet.

The streaming leader has a deep history of raising prices.

“Netflix has been raising prices for 15 years, in markets all over the world, and its annual revenue has grown every single year through all of them,” The Motley Fool’s Daniel Sparks wrote in September after the company raised prices on its lowest-cost tier in the United Kingdom.

Peters reiterated that Netflix has been successful raising prices in the past.

“The ultimate determinant of are we delivering that value is the reaction to price changes. Those recent price changes that we did, including in the United States, have gone well,” he added.

A history of Netflix prices increases

As you can see from the list below, Netflix has no fixed schedule for raising prices. Peters, it should be noted, did not directly say the company was planning a price increase.

His comments instead suggested that customers have historically tolerated increases well and appeared to lay the groundwork to justify an increase.

• 2011: Netflix’s streaming plan cost $7.99 per month.
• May 2014: Netflix raised the U.S. Standard plan from $7.99 to $8.99 per month.
• October 2015: Netflix raised the Standard plan from $8.99 to $9.99 per month.
• October 2017: Netflix raised the Standard plan from $9.99 to $10.99 per month.
• January 2019: Netflix raised the Standard plan from $10.99 to $12.99 per month.
• October 2020: Netflix raised the Standard plan from $12.99 to $13.99 per month.
• January 2022: Netflix raised the Standard plan from $13.99 to $15.49 per month.
• January 2025: Netflix raised the Standard plan from $15.49 to $17.99 per month.
• March 2026: Netflix raised the Standard plan from $17.99 to $19.99 per month.
  • Sources: Netflix’s 2017 SEC filing documents earlier increases, CableTV.com

Related: Walmart launches new membership deal, offers nearly $100 in savings

The two biggest makers of the flash chips in those drives are Korean, according to TrendForce’s sales rankings. Both make much of it in China, at factories caught in a question Washington hasn’t answered.

Samsung Electronics Co. and SK Hynix Inc. (SKHY) are upgrading those plants to make more advanced chips, Reuters reported on Oct. 21. Both confirmed the work, saying it was happening in close consultation with Washington.

That phrase deserves scrutiny. In 2025, the Commerce Department said it intended to license the plants to keep running, but not to expand capacity or upgrade technology. Asked about approvals, it pointed Reuters back to that statement. Reuters could not determine how far U.S. licenses cover the new plans.

Samsung’s Xian plant, which makes about 40% of its NAND flash, is moving from about 200 layers to 286, according to Reuters’ sources.

SK Hynix wants to take its Dalian site from 192 layers to 240 or more, the sources said.

Also read: South Korea’s finance chief bets chip boom outlasts selloff

More layers mean more storage from the same plants

Samsung told Reuters the upgrade complies with U.S. export rules and adds no capacity or new facilities. That covers capacity, not the other half of the 2025 line: technology upgrades, which is exactly what more layers are.

More layers per chip mean more storage density and lower costs, Reuters noted. By my math, Samsung’s move adds about 43% more layers, and SK Hynix’s adds at least 25%.

The same buildings can then turn out more storage. Samsung aims for 5,000 wafers a month of the new chips this December and 40,000 by December 2027, one source told Reuters.

SK Hynix is caught between Beijing and Washington

SK Hynix carries a second complication, and this one comes from Beijing. When China approved its purchase of Intel Corp.’s (INTC) NAND business in 2021, it required SK Hynix to keep raising production for five years, according to the company’s SEC filing.

Put simply, Beijing wants more output, while Washington has said it won’t license upgrades. Both pressures land on Dalian, the only flash plant for the Intel unit now called Solidigm. Beijing’s terms run through this December.

One source told Reuters that SK Hynix is already authorized to make chips with up to 240 layers at Dalian. The timing is unclear. If that permission is recent, Washington has quietly moved past its own 2025 line.

Solidigm has picked Goldman Sachs and Morgan Stanley for a US IPO that could raise about $10 billion, Bloomberg reported. SK Hynix said no specific plans have been confirmed. Buyers will judge that listing partly on how modern Dalian is.

SK Hynix is also discussing a U.S. NAND factory with the Trump administration, according to Reuters. I’d watch whether that factory becomes the price of Washington’s patience in Dalian.

YMTC’s China gains show what the curbs cost

Wuhan’s Yangtze Memory Technologies Co. (YMTC) has gained share from both in China, according to Omdia data cited by Reuters:

• China’s NAND leader in the second quarter was YMTC, with 38.5% of the market, up from 28% for all of 2025.
• Second place belongs to Samsung at 33.5%, down from 36%.
• Third place went to SK Hynix at 17%, down from 20%.

Reuters said the U.S. restrictions contributed to the share loss. That’s an awkward result for controls meant to slow China’s tech progress. Holding back the Korean plants helped hand business to China’s own champion.

More Artificial Intelligence:

• AI’s biggest business opportunity may not look like AI
• AI racks are outgrowing old power: 2 chipmakers have a strategy
• BofA splits software into winners and losers as AI agents arrive

What the upgrades mean for investors and shoppers

Samsung and SK Hynix shares each ended about 6% higher in Seoul after the report, Reuters said. The wider Korean market also rebounded that day, according to TipRanks.

SKHY, the Nasdaq listing of what Bloomberg calls the second-largest memory firm, rose about 3.5% in premarket trading, according to Benzinga.

Meanwhile, Wall Street already backed SK Hynix. Fifteen analysts polled by S&P Global rate SKHY a Strong Buy, with an average price target of $252.21, according to StockAnalysis. That average sits about 80% above Tuesday’s close. SK Hynix reports third-quarter results on Tuesday, Oct. 27, 2026.

Still, shoppers shouldn’t expect quick relief. TrendForce projects NAND contract prices will rise another 15% to 20% this quarter. Samsung’s full Xian target doesn’t land until December 2027.

Export controls work best when the target has nowhere else to turn. Memory never fit that mold, because China has its own supplier and the Korean plants already sit inside its borders.

Samsung even pitched its upgrade as serving customers “including those in the U.S.” Washington’s 2025 policy bet it could freeze these factories in place. The shortage is rewriting that bet one layer at a time.

Related: Chip toolmakers can’t catch a break, even as TSMC spends more

Wall Street expects Meta to report quarterly revenue of roughly $63.3 billion. Analysts are also focused on the impact of a planned roughly $10 billion legal charge tied to the company’s settlement with a bipartisan coalition of state attorneys general over allegations concerning the impact of Facebook and Instagram on children and teens.

In August, Meta agreed to a settlement worth up to $17.1 billion with the states, along with a series of new child-safety measures for its platforms.

The settlement is expected to weigh heavily on reported third-quarter earnings, although it does not change the underlying operating performance of the business.

Beyond the legal charge, investors will be watching Meta’s advertising business, user engagement and the returns it is getting from its rapidly increasing investment in AI infrastructure and products.

Live updates from the company’s earnings report and conference call will be posted on TheStreet.com as they become available. Check back periodically for the latest.

A year later, Oracle shares have fallen about 62% from their September 2025 high. Ellison’s net worth has also dropped by roughly $194 billion, and he now ranks around eighth on the Bloomberg Billionaires Index with a fortune of about $182 billion at the time of writing.

Ellison owns about 40% of Oracle, which is why his wealth moves so much with the stock’s swings. 

Also read: OpenAI CFO offers calming message amid AI slowdown fears

How Oracle’s AI pivot turned into a huge wipeout

Oracle mainly sells cloud services and database software to governments and large companies. However, over the past year, it has focused more on renting out AI computing power to firms such as OpenAI. Because of that, the stock rose. But then it also fell afterward.

In December 2025, Oracle said it expected weaker software sales, and the company also said it would spend far more on data centers than Wall Street expected.

In response, Ellison lost about $25 billion of his wealth in a single day. Then in June 2026, Oracle fell almost 20% in one week, its worst weekly drop since 2001.

Oracle’s heavy debt and a credit downgrade 

Oracle borrowed a lot of money to pay for its AI push.

The company’s total debt is now roughly $169 billion, Yahoo Finance reported. And because of that debt, S&P Global cut Oracle’s credit rating this year.

Michael Burry, the investor who called the housing crash before 2008, has been betting against Oracle. “Oracle didn’t need to do this,” Burry said about the company piling on debt for the large OpenAI deal, Yahoo Finance reported. 

Related: How much Oracle stock is Larry Ellison using as loan collateral?

The OpenAI concentration keeps Wall Street nervous

A big part of Oracle’s reported $664 billion backlog of future contracts is tied to OpenAI. So if OpenAI reduces its spending or runs short on cash, Oracle could be stuck with data centers that are hard to fill with other customers.

Melius Research also warned that Oracle might not be able to keep up with its spending plans if OpenAI or Anthropic ask for more capacity than Oracle can afford to build, Reuters reported. 

More AI Stocks:

• Palo Alto CEO wants AI labs out of the ‘fear-mongering business’
• Jim Cramer sees something Wall Street may be missing in Broadcom
• IBM, Amazon quantum pricing gets a challenger in Postquant Labs

Veteran tech analyst Jack Gold told Techstrong AI that “Oracle is betting big on the need for AI compute and so is investing heavily in ramping its cloud services capability with massive investment in AI systems from NVIDIA and others.

“This has made investors nervous that in a downturn, Oracle’s normally stellar finances and margins could be negatively affected,” Gold added. “As a result, they have been selling off Oracle stock. With Larry Ellison still being a major stockholder, this has negatively affected his net worth.”

What this slide really means for ORCL shareholders now

Oracle’s total revenue increased by about 30% to $19.3 billion in the first quarter of fiscal 2027, and its cloud infrastructure revenue grew 121% year over year, according to Oracle‘s earnings release.

Oracle now trades at about 15.7 times forward earnings, cheaper than the broader tech sector, which is near 21. The stock also has a 12-month average Wall Street price target of about $249. That suggests possible major gains from its Oct. 20 close of $131.14. 

For investors, I would suggest that you keep any position small and watch the next earnings report for any signs that the OpenAI deal is paying off.

More Stocks News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off

But I often wonder how much longer these vehicles can remain affordable for the people who depend on them. And looking at General Motors’ latest numbers, I think they’re expensive now.

GM’s average transaction price climbed above $52,000 in the third quarter of 2026, up from $50,500 a year earlier, CFO Paul Jacobson’s said in an Oct. 20 appearance on CNBC.

I think that’s not a small jump for someone buying a truck to make a living. Yes, the vehicles may be getting better, but they’re also becoming harder to justify financially.

That tension between a company that is genuinely executing well financially and customers feeling the pressure of higher prices ran through GM’s entire Q3 2026 earnings presentation. 

In fact, the company beat on both revenue and earnings. But the stock fell 8% anyway, as of the afternoon of Oct. 20.

That drop has pushed General Motors Company (GM) stock down over 4.5% year to date, according to Yahoo Finance, and a one-year return is fairly strong at 34%.

Also Read: GM wants drivers to be able to text and drive

The $20 billion move and what it says about where GM thinks its stock is valued

The headline from the Oct. 20 earnings was not even the beat but the buyback. GM’s board approved a new $20 billion multi-year share repurchase authorization after completion of an earlier $6 billion program, according to GM’s Q3 earnings announcement. 

The new program has no expiration date. Jacobson had a few words to say about the logic in his CNBC interview. 

We have one of the lowest price-to-earnings multiples on the street and a double-digit free cash flow yield. We know that buying back shares is a really strong investment for our shareholders.

The CFO is simply saying that his GE stock is cheap. GM’s forward P/E ratio on Macrotrends read 6.67 as of Oct. 20, 2026. According to a report by TheStreet’s Tony Owusu, BNP Paribas analysts viewed General Motors’ stock drop after the Q3 earnings as a buying opportunity.

The numbers behind the claim are solid. GM has reduced its fully diluted share count by 37% over the past three years, repurchasing an average of $7.5 billion annually over four years. 

As of Sept. 30, GM had 849 million shares outstanding, down from 995 million at the end of 2024 and 1.2 billion at the end of 2023, according to GM’s earnings release.

The capital allocation framework Jacobson described is to invest in the business first ($10 billion to $12 billion this year), protect the balance sheet second, then return capital. “This is a cyclical industry,” he noted, and the balance sheet protection reflects hard-won lessons from 2008.

The GM earnings that funded the announcement

Q3 2026 revenue was $50.02 billion, up 2.9% year over year and beating consensus expectations of $48.21 billion, according to the GM’s earnings release.

Adjusted EPS was $3.70, up 32.2% year over year. Adjusted EBIT was $4.11 billion with an 8.2% margin. Global wholesale volume reached 1.03 million vehicles.

More Automotive:

• Tesla rival appears to have a demand problem
• Cars are getting older: AutoZone earnings reveal how drivers cope
• Volkswagen’s newest EV runs Chinese chips and starts near $19,000

Full-year adjusted EPS guidance was narrowed to $12.75 to $13.75, up from an earlier range of $12.00 to $14.00 — the second guidance raise of the year, according to Jacobson.

That brings me back to my earlier concern about affordability. Interestingly, Jacobson said Q3 pricing was flat to slightly lower year over year, with a $200 million decline. 

Higher average transaction prices instead reflected a shift toward more expensive trucks and fewer heavily incentivized EVs following the expiration of consumer tax credits.

The hybrid gap GM investors are watching and the Q4 headwinds that spooked them

The 8% stock drop on a beat quarter came from two specific concerns Jacobson addressed.

First, GM plans to cut fourth-quarter pickup truck production by roughly 37,000 wholesale units as plants shut down and reopen to prepare for redesigned 2027 models, he told CNBC. The redesigned full-size trucks represent a major product cycle, but the changeover creates an inevitable sequential earnings step-down.

Second, commodity and logistics inflation has added approximately $400 million in pressure over the last two months of the quarter, with full-year headwinds now projected at $1 billion to $1.5 billion from higher steel, aluminum, and logistics costs.

“The team is digesting it,” Jacobson said, noting that GM held its full-year guidance despite that new pressure.

Another key issue CNBC’s Phil LeBeau pressed Jacobson on was hybrids, which now account for nearly 20% of the U.S. market, while GM has just one offering.

Jacobson acknowledged the success of rival hybrid manufacturers and said hybrids “are coming” to GM, but gave no specific timeline. “We understand how hybrids are growing, and we’re continuing to adapt,” he said.

For buyers who depend on affordable, fuel-efficient trucks to make a living, GM’s next move may matter more than its buyback. For investors, both affordability and shareholder returns deserve attention.

More Stocks News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off

The rally comes after a difficult week for AI stocks. Investors have been concerned about the massive cost of building data centers, particularly as companies such as CoreWeave (CRWV) rely heavily on debt to finance their expansion.

Super Micro works closely with chip makers like Nvidia (NVDA) and builds ready-to-use server hardware that supplies major cloud players like Microsoft (MSFT) and Amazon (AMZN). It also creates liquid and air cooling systems.

The company now expects fiscal first-quarter revenue of $18.5 billion to $19.5 billion, up from its previous forecast of $14.5 billion to $15.5 billion. The midpoint of the new guidance of $19 billion is roughly 27% above Wall Street’s previous consensus estimate of $14.99 billion.

Super Micro also raised its gross margin forecast to 12% to 12.5%, up from its earlier outlook of 10.4% to 10.8%, citing better customer mix and product mix.

The company is scheduled to report its full quarterly results on Nov. 3.

Wall Street reacts to Super Micro’s stronger AI demand

Bank of America raised its price target on Super Micro to $38 from $33 following the announcement, but maintained its Underperform rating, according to The Fly.

The bank remains cautious about the company’s profitability, citing pricing pressure in the AI server market, rising component costs and lower margins typically associated with large deals.

Still, BofA sees the stronger revenue outlook as an encouraging sign for competitors Dell Technologies (DELL) and Hewlett Packard Enterprise (HPE), suggesting that demand for AI infrastructure remains healthy.

Citi maintained its Neutral/High Risk rating and $45 price target for Super Micro stock.

Also read: Intel stock surges on JPMorgan upgrade after 14% weekly plunge

The bank noted that Super Micro’s updated outlook implies approximately 280% year-over-year revenue growth, compared with roughly 195% under its previous guidance.

Citi also estimated that the stronger revenue and margin forecasts could put quarterly earnings per share roughly 60% to 70% above previous expectations.

“Overall, we view the announcement as constructive for the broader server OEM ecosystem, including Dell and HPE,” Citi analysts wrote in a research note sent to TheStreet.

The firm said AI infrastructure is increasingly shifting from individual servers toward integrated rack-level systems, creating opportunities for companies that provide liquid cooling, networking and other data-center equipment.

Super Micro still faces questions over profitability and cash flow

Despite the stronger outlook, Wall Street remains focused on whether Super Micro can turn its rapid revenue growth into sustained positive cash flow.

Citi said investors can focus on the durability of AI demand, particularly among neocloud customers, as well as financing conditions and deployment timelines.

Also read: CoreWeave stock plunges 12%, but Wall Street analyst says buy the dip

“We believe a key investor debate remains the timing and magnitude of free-cash-flow inflection,” Citi analysts wrote.

The bank also pointed to working capital and cash conversion as key issues heading into Super Micro’s Nov. 3 earnings.

Super Micro stock trades at around $35 and is up 22% year-to-date as of writing. 

Related: Cathie Wood buys $11.8 million of tumbling tech stock

That’s roughly the amount of money the bank expects Alphabet (GOOGL) to spend on capital expenditures in 2027, building data centers and AI chips.

Take something as familiar as searching for a restaurant. Over the past several years, Google has steered you toward reviews, menus, and reservation pages. Now, its AI tools can pull together answers and recommendations before you ever click through.

Though that feels like a small convenience. Behind the scenes, though, providing those answers entails costly chips and data centers. And as people’s search habits change, Google needs to find ways to keep its lucrative advertising business growing.

That’s quite a balancing act, which involves paying for the future without weakening the business financing it.

That said, in a note shared with me, Bank of America remains bullish, maintaining a Buy rating and $430 price target. However, the company’s CapEx forecast raises a big question for shareholders: whether Google’s growing AI and cloud businesses can generate enough cash to justify the bill.

Bank of America sees Google’s cloud growth masking a costly problem

On the surface, it’s easy to see why Google will continue on an upward trajectory.

More Google:

• Google is making a change in November that might catch Gemini users off guard
• Google set to challenge Meta Muse in the Agentic AI race
• Google just disclosed something troubling about its AI

Alphabet brought in $119.8 billion in Q2 sales, up 24% from a year earlier. Google Cloud, once a smaller piece of the business, grew 82% to $24.8 billion, an extraordinary pace for a company of Google’s size.

Post expects the momentum to continue when Alphabet reports on October 28. His Q3 forecast calls for $111.4 billion in sales after traffic-acquisition costs and $3.07 in GAAP earnings per share, ahead of Wall Street’s comparable estimates of $111 billion and $2.98.

That said, the growth isn’t evenly spread across Google.

Bank of America expects Search revenue growth to cool from 17% in the previous quarter to 14%, while Cloud accelerates from 82% to 93%. Google’s traditional cash-generating business is growing much more slowly, even as its AI infrastructure operations expand rapidly.

The more surprising finding is BofA’s cash-flow forecast.

The bank expects negative free cash flow of $12.4 billion in Q3, compared with about $3.9 billion expected by other analysts. And it forecasts around $300 billion in capital spending in 2027 alone.

That’s a huge financial balancing act. Google can continue to deliver strong sales, even blow past earnings estimates, while pouring more cash into the data centers powering its AI ambitions.

So why is a business this profitable consuming so much cash?

Google’s AI ambitions could consume $500 billion in just two years

BofA expects Google’s AI ambitions to come with a massive bill.

It estimates Alphabet will spend about $199 billion in CapEx in 2026 and another $301 billion in 2027. That’s roughly half a trillion dollars in just two years, much of it tied to data centers, computing chips, and the infrastructure needed to handle rising AI demand.

But Google isn’t simply buying more servers and hoping customers show up.

It’s building its own Tensor Processing Units (TPUs), expanding cloud capacity, and selling computing power and AI services to other businesses.

Google Cloud CEO Thomas Kurian offered a sense of that demand at a September technology conference. 

He said over 300 customers had each committed over $100 million, while customers were spending, on average, more than 50% above their contractual commitments.

Kurian also said Google’s AI servers pay for themselves in under two years, with its own TPUs expected to pay back faster than conventional graphics processors.

Nevertheless, there are a few caveats to consider. 

Bank of America notes that an index tracking AI-model usage costs fell 33% in Q3. Cheaper AI might attract more customers, but Google needs enough additional usage to offset the lower per-unit earnings.

And its original moneymaker faces a different challenge.

As Google adds AI answers and rivals develop agents that complete tasks directly, users might have fewer reasons to visit websites. Pew Research found traditional search-result clicks occurred in 8% of visits with AI summaries, versus 15% without them.

So even though Google has the customers and technology to make this work. Whether they can generate enough cash to justify the bill is another matter.

Buffett’s big Google bet faces an October 28 reality check

Interestingly, that question also matters to Warren Buffett’s Berkshire Hathaway (BRK.A), too.

Years ago, Buffett admitted he’d missed an opportunity in Google’s advertising business, despite the potential it saw through GEICO’s search-ad spending as reported on CNBC. He eventually initiated Berkshire Hathaway’s Alphabet investment, just as Google was becoming a much more expensive business to build, which I covered a few months ago.

Now, investors have an important date circled which is when Alphabet reports earnings on October 28 for Q3.

Bank of America expects Google Cloud revenue to grow 93%. But another impressive growth number won’t settle the debate.

Investors will want to know whether Search advertising is holding up as AI changes how people find information, and whether Google can turn soaring Cloud demand into stronger margins.

Then there’s the spending. Any increase in Alphabet’s CapEx outlook, or a deeper free cash flow shortfall, would raise fresh questions about how quickly its AI investments can pay off.

Even headline earnings may need a closer look.

BofA estimates Alphabet’s SpaceX stake could produce a significant mark-to-market headwind, complicating comparisons with the unusually large investment gains recorded earlier this year.

The big new challenge is whether those businesses can generate enough cash over the coming years to support the extraordinary infrastructure being built around them.

Related: Morgan Stanley sees new hurdle for Microsoft stock

Then, on October 18, 2026, Anthropic announced the creation of the Anthropic Public Benefit Fund. Subject to stockholder approval, the new entity is set to receive 5% of the company’s shares, and the money will go toward helping workers whose jobs are disrupted by artificial intelligence, CNBC reported.

This is one of the first serious attempts by an AI company to provide support to affected individuals.

Inside Anthropic’s pledge to displaced workers

Anthropic is the company behind Claude AI, and it makes money by selling access to that AI to businesses and everyday users.

Also read: OpenAI CFO offers calming message amid AI slowdown fears

All seven of Anthropic’s co-founders have earlier pledged to give away 80% of their personal shares, Fortune reported. When you add that to the 5% of Anthropic’s equity that the fund is starting with, and a company program that matches employee donations, that means roughly 20% of Anthropic’s total shares are now set aside for public benefit.

If the company hits a $2 trillion valuation for its IPO, that initial 5% stake alone could be worth up to $100 billion when trading begins, Forbes reported. That would be one of the biggest corporate commitments to a social cause ever announced. Investors have recently scaled back their valuation estimates to roughly $1.5 trillion. Still, that would be worth about $75 billion.

Why Anthropic chose an LLC instead of a charity

Anthropic set up the fund as a wholly owned Limited Liability Company, not a traditional nonprofit. That structure gives the company legal room to try things most charities simply can’t do.

For example, the fund can make direct payments to specific populations hardest hit by AI’s labor-market impact. It can also fund retraining programs and groups that help workers transition to new careers.

More AI Coverage:

• Anthropic’s $2 trillion IPO dream hits a new reality
• Bill Gates makes bold predictions on the future of humanity and AI
• OpenAI CEO Sam Altman has a biblical-level warning about AI

“We should do everything we can to prevent it,” Amodei said in June about large-scale AI job losses, calling permanent worker displacement “undesirable and dangerous,” according to Business Today.

Anthropic has also blocked the fund from spending any money on lobbying or political donations. And the fund is set to grow with the company. At the board’s discretion, every time Anthropic’s valuation rises by 50%, the fund gets another 2% of equity.

What this means if your job could be lost to AI

Anthropic has not yet said who qualifies for help or when any direct payments might reach anyone. The company says more details will come in the next few months.

No one knows exactly when people will see real benefits from this fund, so if you work in customer service, basic coding, entry-level finance, or routine administrative work, improving your own skills is the smart move for now. 

The catch behind the big payout promise

Equity doesn’t pay bills until it is sold. If the AI trade or the broader stock market slows after the IPO, it could easily cut into the cash the fund can eventually hand out.

I also think the fact that Anthropic has not presented any guidelines on qualifications and a timeline for adoption means the whole fund could be a thing of the future, rather than a near-term plan.

More Stocks News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off

Then came a twist few people saw coming. Just days after that admission, Zuckerberg publicly thanked Apple, the company whose rules cost Meta so much, for helping one of its new AI models run on Apple hardware. The two moves tell the same story about where Meta wants to stand in the AI era.

Muse Wonder is the first real test of the spending

The payoff Zuckerberg was waiting for arrived on October 12. Meta’s Superintelligence Lab released Muse Wonder and a stronger version called Muse Wonder Full, its first serious shot at a frontier model since Alexander Wang reorganized the company’s AI work in August 2025.

Also read: Mark Zuckerberg says Meta found a way to make more money

Meta is pitching memory as the standout feature.

The company says its own testing puts the models alongside leading rivals and highlights what it calls “active memory,” the ability to hold on to and combine information users share over time, as reported by TechCrunch. Both models will power Meta’s Muse agent, and they were trained at its Prometheus data center in central Ohio.

That training did not come cheap. Wang said building a frontier model now costs “tens of billions of dollars,” and Meta’s performance claims still need independent results to back them up, according to TheStreet. For now, investors are mostly taking the company’s word for it.

Early demand looks encouraging, though. Meta’s Muse personal assistant app logged 2.8 million downloads in its first 12 days, according to TechCrunch. Citi thinks the app could eventually add at least $27 billion in annual revenue by fiscal 2030.

The bill keeps getting bigger

The cost of owning everything is showing up in Meta’s books. When it reported second-quarter results on July 29, Meta raised the low end of its 2026 capital spending forecast by $5 billion, while free cash flow sank 91% from a year earlier to just $784 million. Some investors are already comparing the splurge to the costly metaverse push.

Zuckerberg used that same call to sketch a way out. He said businesses have offered to run their own AI projects on Meta’s capacity at a meaningful premium to what it cost to build, but added that “there will continue to be a significantly higher margin on selling intelligence rather than selling compute directly,” as reported by Reuters. In other words, renting out servers is a side gig, not the plan.

Hyperion shows how quickly the price tag can swell. In July, Meta said its Richland Parish campus in Louisiana would grow to 5 gigawatts of computing capacity from an original design of 2 gigawatts, pushing the cost above $50 billion from a prior estimate of $27 billion, Quartz reported. It will be Meta’s largest data center.

Investors have flinched more than once. On September 25, a Goldman Sachs note argued that the largest U.S. AI spenders need roughly $300 billion in annual AI revenue just to break even, and Meta shares slid about 4% that day. Zuckerberg’s own fortune dropped by billions in that single session.

Old Apple wounds shaped the go-it-alone strategy

The scars from Apple’s privacy change run deep. In February 2022, then-CFO Dave Wehner called the hit from App Tracking Transparency “a pretty significant headwind for our business,” admitted the $10 billion figure was only an estimate and watched the stock sink 23% in extended trading.

Wehner also took a pointed swipe at Apple’s search deal with Google, which he suggested was treated differently under the same privacy rules. “The incentive clearly is for this policy discrepancy to continue,” he said, as reported by CNBC. It was a reminder of how little say Meta had over the platform carrying its apps.

More Mark Zuckerberg:

• How much META stock does Facebook founder & CEO Mark Zuckerberg own?
• Mark Zuckerberg sends bold message to Meta investors after Muse
• Mark Zuckerberg says Meta found a way to make more money

The mood has softened lately. On October 15, Zuckerberg thanked Apple’s chief executive, John Ternus, for helping Meta’s Muse Spark Open model run on Mac Studio computers, in what appeared to be the first public partnership between the two companies in years. Neither side disclosed financial terms.

The power balance may be shifting, too. Bank of America analyst Wamsi Mohan warned on September 29 that AI agents could capture shopping, referrals and payments that now flow through Apple’s ecosystem. If agents become the new front door, owning the full stack could leave Meta holding a much stronger hand.

Wall Street wants answers on October 21

Analysts are mostly upbeat heading into Meta’s October 21 earnings report. Bank of America’s Justin Post wrote that Meta is “back on offense with AI” and said recent launches have strengthened confidence in Meta’s AI execution, though he flagged the risk that heavier AI investment could pressure 2027 earnings, according to Yahoo Finance.

JPMorgan has also come around. Analyst Doug Anmuth upgraded the stock to Overweight in September, arguing Meta is still early in rolling out frontier models and AI products beyond advertising, while projecting capital spending well above Wall Street’s consensus for the next two years. He also noted that the Muse app hit No. 3 in the U.S. App Store on its second day, TheStreet reported.

The harder question is what happens to the ad machine. Bank of America has listed competition from AI-native platforms for users’ attention as one of Meta’s key risks, noting that advertising still accounts for most of its revenue. Zuckerberg is betting that owning every layer gives Meta the best shot at controlling where those eyeballs go next.

Related: Mark Zuckerberg sends bold message to Meta investors after Muse

Released on Sept. 8, downloads for the AI assistant surpassed 5 million in just 22 days. By Oct. 20, it had more than doubled that amount to over 10.9 million downloads, according data from market intelligence firm Sensor Tower viewed by Tech Crunch.

It reached those milestones much faster than more AI-centric rivals like OpenAI’s ChatGPT, X’s Grok and Anthropic’s Claude. ChatGPT hit the 5 million mark after 56 days, Grok hit it after 103 days, and Claude reached that many downloads after 492 days. Muse is only one of 23 apps to have ever hit 5 million downloads within 22 days of launch, according to Forbes.

Muse helps users do everything from book appointments, to fill out forms, and handle customer services. In essence, it “proactively takes tasks off your plate.”

With Meta already mining so much data from its users, one of Muse’s selling points was robust privacy protections. However, a new study from Claroc shows a big hole in Muse’s data security.

Muse privacy protections as a selling point

Meta Platforms undoubtedly knows its reputation among the general public. The terms Facebook and privacy are basically antithetical at this point. So it made sense for CEO Mark Zuckerberg to make sure potential users knew that they could trust Muse with their most personal queries.

“Muse is built from the ground up for privacy and security,” Zuckerberg announced on Sept. 8. “Your data and credentials live on the Muse Secure VM — an isolated linux computer with a browser, CPU, memory, and storage.”

“A Sentinel agent separate from your Muse runs on your VM,” he continued. “Every action or piece of data that goes out to the network has to be approved by the Sentinel. The kernel enforces that and it knows when it needs to get your permission to proceed.”

Read more: Mark Zuckerberg sends bold message to Meta investors after Muse

However, despite all of the clear thought and work that went into making Muse secure, it didn’t take long for people in the know to locate the holes in Meta’s security fence.

Two weeks after it’s release, Dan Goodin at Ars Technica outlined a “zero-day” vulnerability that showed that Muse’s security wasn’t as tight as Zuckerberg suggested. But now, researchers at Claroc also just exposed a vulnerability built into Muse’s design.

Sharing Muse conversations exposes you

The process that allows you to share the conversations you have with Muse lead to a vulnerability that the tech’s designers unwittingly embedded into the product.

When a user asks Muse AI to share a conversation, the AI agent publishes it as a web page that anyone with the link can open.

“That is by design: a public link is the only way the teammate it is sent to can open it,” Claroc said in its report.

But its research team found that those links are also easy to find by anyone with the time to look for them because “passive-DNS and subdomain-discovery tools list the hostnames directly.”

All it took for researchers to find the cache of data was opening a sample from a web browser in incognito mode and, even with no login, they were able to reveal users’ personal files and companies’ internal documents.

That data included a home-sale budget, a to-do note flagging a bank-fraud alert with a partial account number, a small-business loan application, a CEO-only revenue review, and a candidate’s confidential prep for an executive-search call with a frontier AI lab.

More Meta:

• Meta Muse gives AMD and Intel investors a reason to cheer
• Meta and Microsoft just sent employees a memo about Anthropic
• Bank of America sends a stark message to Meta stock investors

Every page was public “…because someone asked Muse to share it, and Muse did what it was asked,” Claroc said in its report. “What makes Muse’s shares easy to find is the title sitting inside the web address, where public DNS records it for anyone to search.”

Researchers had access to a cache of nearly 7,000 hostnames, but they only opened a few hundred of them and left the rest because lucky enough for them, and anyone else with more nefarious plans, the page titles had descriptive titles of the information contained inside. Making the process even less secure, they were able to open them without an account or a sign-in screen.

“Plenty of AI assistants share by public link. What makes Muse’s shares easy to find is the title sitting inside the web address, where public DNS records it for anyone to search,” Claroc said in its report. “Muse builds each share’s web address from the share’s title, a short phrase describing the conversation, then a random ID. The hostname usually gave away what a page held before we opened it.”

“If you had asked us beforehand, we would have guessed that gathering these would take real effort,” Claroc said. “It took one query.”

How did Meta respond?

A Meta spokesperson told TheStreet that the company has paused the conversation sharing link functionality while it reviews Cleric’s report and that it will “consider changes to how these links can be discovered.”

The company said that when users share artifacts, which is the term Meta uses for the “interactive outputs” to your queries that the AI responds with, Muse does automatically create a page that can be accessed, like the Claroc report says. However, Meta also said that the page “can be accessed by those with the link.”

Claroc discovered 6,800 links without being sent them. They checked hundreds in incognito mode without a login, according to their report.

But the company is responding to the report.

“However, we’ve paused this functionality while we review this report and consider changes to how these links can be discovered,” said the spokesperson.

“We are glad to see it closed this quickly,” Claroc said in an update on Tuesday, Oct. 20.

So Claroc went back and rechecked 40 shares that it had previously opened without an account. None opened for them and the hostnames that serve their content returned an error message.

However, the security loophole may not be fully closed as the hostnames themselves, which include those pesky titles that make sorting the mundane data from the valuable so easy, were still in passive-DNS history, according to Claroc.

Related: Meta Muse gives AMD and Intel investors a reason to cheer

The carrier said its contract gave it the option to renew support for two more years. Broadcom would only continue supporting AT&T through new subscription packages requiring three-year contracts, AT&T alleged.

At a hearing on Oct. 23, 2024, AT&T’s lawyer said the carrier would seek to leave VMware over licensing and cost changes. “Without the threat, there’s really no way Broadcom will become more reasonable,” he told the court.

The two sides reported a settlement in principle that December, without disclosing terms.

Now another telecom giant is heading the other way. NTT DOCOMO, the mobile arm of Japan’s NTT, has signed a 10-year deal to keep its private cloud on VMware Cloud Foundation, according to a Broadcom press release on Oct. 19, 2026.

The length is the surprise. Forrester analyst Naveen Chhabra told TechTarget that most customers would sooner reduce their VMware dependence over time than fight. DOCOMO just committed for a decade.

Also read: Elon Musk takes a direct swipe at Broadcom CEO and AI investors

DOCOMO bought the whole bundle, not a sliver

Financial terms were not disclosed. A footnote in the release spells out the scope, and it caught my eye. DOCOMO is taking Cloud Foundation plus VMware’s load balancing, vDefend security, and its Tanzu app platform.

It also gets dedicated account managers and VCF Forte Support, which carries Broadcom’s strictest service targets. To me, that reads like the showcase Broadcom has wanted since it folded VMware’s catalog into two main bundles, as TechTarget confirmed.

The container piece matters more than it looks. DOCOMO will run virtual machines and containers on one platform using VMware’s own Kubernetes service, according to the release.

A carrier modernizing its software could have used that shift to walk away, and DOCOMO chose to stay.

The same outage fear pushed AT&T and DOCOMO apart

AT&T’s lawsuit leaned on one warning. Its lawyers argued that losing VMware support could knock out vital systems for public safety agencies, according to TechTarget. VMware, in other words, was too embedded to unplug quickly.

DOCOMO’s priorities read as that same worry turned into a purchase. The carrier wants to upgrade systems without interruptions and hold quality steady through traffic spikes. “VMware products have long been used across a range of our systems,” said Hiroshi Ishimaru, a DOCOMO general manager, in the release.

That is the uncomfortable math for VMware’s critics. The deeper the software sits inside a network, the riskier a migration looks. DOCOMO has built its private cloud with VMware since 2019, the release said, so this extends a seven-year bet.

The deal gives Broadcom stock a steadier software anchor

Broadcom stock mostly trades on AI chips. AI semiconductor revenue hit $16.7 billion in the fiscal third quarter, more than triple the year-earlier figure. That growth is why investors scrutinize every big cloud order.

Software is the quieter half. Infrastructure software revenue rose 29% to $8.75 billion in the same quarter, the release showed. Its share of total sales fell to 30% only because chip sales grew far faster.

More Broadcom:

• Jim Cramer sees something Wall Street may be missing in Broadcom
• Broadcom CEO echoes Warren Buffett with warning for investors
• A $10,000 Broadcom investment in 2016 now pays $1,991 a year in dividends

This deal helps that half in two ways. A decade of committed licensing and premium support should mean predictable revenue to cushion a lumpy chip cycle. It also hands Broadcom’s sales team a marquee carrier to cite when other operators renew.

AI links the two halves. CEO Hock Tan told analysts in June that processors sold alongside AI chips are “driving some accelerated growth of our VMware business.” DOCOMO’s plan to run future AI workloads on the same platform fits that thesis and the shift toward private clouds for AI.

Investors have not rewarded the story yet. Broadcom shares were little changed early on Tuesday, Oct. 20, according to StockAnalysis.

The stock sits roughly a third below its 52-week high. Even so, the site shows a Strong Buy consensus from 50 analysts.

One contract will not move a company that reported $29.6 billion in sales last quarter. I would treat it as a sales tool rather than a revenue event, and that may be worth more.

Long commitments could become VMware’s real moat

The VMware debate has mostly focused on who leaves. For investors, the sharper test is how long the customers who stay will commit.

DOCOMO’s answer stretches past its next network upgrade. The release says the platform is meant to evolve with AI-driven threats, 6G and IOWN, NTT’s next-generation optical network project. That is a carrier planning around technology still years from wide use, with VMware at the center.

If other carriers follow, the AT&T fight will look less like an exodus and more like a test of leverage. The next big telecom renewal will show which reading is right.

More Stocks News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off

I’ve followed Bank of America’s weekly iPhone lead-time tracker since preorders for the iPhone 18 Pro opened in September. This week’s update is the first one with real data on Apple’s foldable, and it holds a useful surprise for buyers.

Demand for new iPhones is not the question. Apple (AAPL) reported $109.4 billion in revenue for its June quarter, up 16 percent, with iPhone sales up 22 percent to $54.3 billion, according to MacRumors.

“This is the most powerful and most popular iPhone lineup we’ve ever had,” then-CEO Tim Cook said on the company’s July 30 earnings call.

However, the supply side of that story looks very different, depending on where you order.

According to a Bank of America Global Research note dated Oct. 19 and reviewed by TheStreet, the iPhone Duo’s wait on U.S. carrier websites averages about 50 days, roughly half the 102 days Apple’s own store is quoting across all storage sizes and colors.

Duo shoppers face two very different waits

Bank of America tracks ship dates on Apple’s website and on major carrier sites every week. As of Oct. 19, the global weighted average wait for the iPhone Duo stood at 82 days, the note said.

Outside the U.S., the picture is uniform and slow. China, Japan, the U.K., Germany, France and Australia all showed a 95-day wait for every storage size, which pushes delivery to about Jan. 22.

U.S. carrier sites were noticeably faster across the board:

• The 256GB Duo showed an average wait of 53 days, according to the Bank of America tracker.
• The 512GB model was at 48 days and the 1TB version at 50 days, the note showed.
• The top 2TB model, which costs $3,199, showed the shortest wait at 47 days, based on pricing reported by TheStreet.

Analyst Wamsi Mohan, who wrote the note titled “iPhone Duo lead times extend to next year; Pro and Pro Max broadly stabilize,” offered a simple explanation for the gap.

Availability on U.S. carrier sites “appears to be much better than Apple’s own website,” which “could be due to varying allocations of inventory,” wrote Mohan.

Related: Bank of America flags surprising iPhone 18 pre-order trend

The in-store picture is tighter still. The Duo shows no in-store availability for its Oct. 23 launch day, and Apple’s site is telling shoppers to check again on Oct. 24 for pickup, the note said.

There is a catch for your wallet. Buying through a carrier usually ties the phone to that carrier’s installment plan, while Apple’s own iPhone Upgrade Program starts at $57.99 a month on a 24-month lease, TheStreet reported.

A shorter wait is only worth it if the carrier terms fit your budget.

China lead times flash a warning on iPhone 18 Pro demand

The Duo grabs the headlines, but the iPhone 18 Pro, one of just three new iPhones Apple launched this fall, is where wait times keep shrinking.

The global average fell to eight days from nine a week earlier and from 14 days in the first week of preorders, according to Bank of America.

More Apple:

• Apple cuts iPhone 18 Pro orders as AI raises the bill
• Apple CEO rejects biggest fear over iPhone Duo launch
• Apple stock is giving new CEO John Ternus plenty to smile about

China moved the most. Pro lead times there dropped to four days from 11 the week before, “indicating a meaningful improvement in availability,” wrote Mohan. That is now the shortest wait among international regions, where the Pro averages about 11 days.

The Pro Max is a different story. Its global wait held at 15 days, and the note said availability “remains comparatively tighter” for the bigger model.

By my analysis, a sudden drop in China wait times can mean two things: Apple caught up on supply, or buyers there are cooling off. The backdrop makes the second reading hard to dismiss.

China’s smartphone shipments fell 15.4 percent from a year earlier to 57.9 million units in the third quarter, and Apple’s shipments slipped 4.3 percent, according to preliminary IDC data reported by TechNode. Huawei led the market and was the only major vendor with double-digit growth.

IDC blamed higher memory and component costs, which pushed retail prices up and stretched out how long people keep their phones, TechNode reported. Apple raised starting prices for the iPhone 18 Pro and Pro Max by $100, to $1,199 and $1,299.

Given that backdrop, it’s easy to see why the Pro data deserves a closer look than the Duo sellout. Apple told suppliers to cut October component orders for the Pro models by at least 15 percent, Nikkei Asia reported, as TheStreet noted on Oct. 12.

Fewer orders and shorter waits can point to the same thing.

What Bank of America expects before the Nov. 2 report

For now, Mohan reads the week as normalization, not trouble. The Pro data points to “continued normalization,” he wrote, while Pro Max supply stays tighter and initial Duo store inventory “appears to be sold out.”

The next real test is Nov. 2, when Apple reports results for its September quarter, according to MacRumors. Apple guided for revenue growth of nine percent to 11 percent and mid-teens iPhone growth for that period.

That guidance also came with a warning. “For September, we expect to pay even higher memory costs,” Cook said on the July call.

Much will likely depend, however, on whether the Duo’s long wait reflects demand or a thin first production run. Apple hasn’t said how many units it had for preorder, and a 95-day wait can come from either.

Bank of America’s price objective of $370 implies about 7.7 percent upside from Apple’s $343.46 price in the note. It rests on 37 times the bank’s calendar 2027 earnings estimate of $9.98 a share, above Apple’s five-year range of 19 to 35 times, the note said.

That is a rich multiple to pay for a company whose flagship Pro phone is seeing shorter waits. I’d want to see the Nov. 2 iPhone number before treating the Duo sellout as a growth story.

If you are trying to buy a Duo, the practical move is simple: Check your carrier’s site before you settle for Apple’s January date.

Mohan, for his part, is sticking with the stock. He reiterated a Buy rating “on strong capital returns, eventual winner on AI at the edge and optionality from new products/markets.”

Related: Apple extends iPhone Duo delivery, disappointing eager customers

AI stocks have since rebounded, but the sharp swings show how sensitive investors remain to news about AI spending and financing.

For the past few years, the AI boom has been a race to build powerful language models. Companies have spent billions of dollars on data centers, advanced chips, and computing infrastructure, betting that demand for AI will continue to grow.

But the next phase of AI may be less about building bigger models and more about how people interact with the technology in the physical world.

That’s where Ramin Hasani, co-founder and CEO of Liquid AI, sees potential.

In an exclusive interview with TheStreet at Qualcomm’s Snapdragon Summit in September, he discussed where he sees the next opportunities in AI; how AI could move beyond chatbots and data centers into cars, smartphones, and other everyday devices; and whether soaring AI valuations have gone too far.

AI isn’t just about chatbot models

Liquid AI was founded by a team from the Massachusetts Institute of Technology, and it develops AI models that run across various hardware types.

Unlike companies building consumer-facing chatbots, Liquid AI focuses on supplying technology to businesses that want to integrate AI into their own products.

Hasani said there is an important distinction between the consumer AI market, where companies compete to build apps and agents, and the enterprise market, where manufacturers need the underlying technology to make those products work.

Also read: CoreWeave stock plunges 12%, but Wall Street analyst says buy the dip

“If you just look at the model layer, you will see no differentiation,” Hasani said, referring to the growing number of AI models available from companies in the U.S. and China.

He argued that the real differentiation comes from combining models with the software systems needed to enable more sophisticated AI behavior.

For example, a carmaker that develops an intelligent assistant needs more than a language model. The system must understand users’ voice commands, process information from cameras, and access vehicle functions while responding quickly.

Liquid AI provides models and supporting software, developing a platform that enables businesses to customize and improve their AI systems. The idea originated from Hasani’s study of how neurons exchange information in small animal brains.

“Liquid means flexibility,” Hasani said, explaining that the goal is to build AI models that are adaptable after training.

Why the next AI battleground could be your car

Liquid AI is working with Mercedes-Benz on an AI-powered in-car assistant that Hasani expects to roll out in North America by the end of 2026.

Hasani said his company provides an AI system that runs inside a Mercedes vehicle on Qualcomm (QCOM) chips, allowing drivers to communicate with their cars more easily. Its software can access vehicle functions that traditional smartphone assistants cannot.

Hasani used Apple’s (AAPL) Siri as an example.

“What if you ask Siri to open my window?” he asked.

Related: Qualcomm CFO talks AI, data centers, and next big bet on robots

Siri generally cannot control a vehicle’s windows without the necessary integration yet.

“When we work directly with OEMs (Original Equipment Manufacturer), we also get all the access inside the car,” Hasani said. That access could enable thousands of vehicle functions through human language commands.

The auto industry is moving toward vehicles that can gain new features through software updates. That shift could create a growing market for chips and AI software designed to run inside cars.

Liquid AI is also partnering with Qualcomm (QCOM) to bring its AI models to Snapdragon-powered devices, allowing more AI tasks to run directly on the hardware rather than relying on cloud-based data centers.

Businesses might cut costs, not AI usage

Consumers may tolerate waiting several seconds for a chatbot to answer a complicated question. Businesses processing millions of requests often cannot.

Hasani pointed to Liquid AI’s work with Shopify (SHOP) as an example. Requests made through Shopify’s Shop app can be processed using a Liquid foundation model. The company needs to handle large volumes of requests while maintaining extremely low latency.

“You want to conclude a job that is coming in 20 milliseconds with an AI model…. There aren’t that many options in the world that can do that,” Hasani said.

Liquid AI’s approach involves optimizing models for the hardware on which they run.

For many businesses, a smaller model that responds quickly and costs less to operate may be more attractive than a larger model with stronger benchmark results. As AI becomes more deployed, efficiency could become as important as raw computing power.

Related: Micron stock tumbles despite analyst calls semiconductor selloff overblown

Hasani expects businesses to become more selective about AI spending as they look for better returns from the technology.

“There’s a gap between the capabilities of a model and the value that they’re creating,” he said.

He believes companies will find ways to reduce AI costs without necessarily using the technology less.

“Enterprises are getting smarter about the expenditure on AI, so they’re definitely gonna cut that cost, but that doesn’t mean that they’re gonna cut usage,” Hasani said.

How Liquid AI makes money

Liquid AI makes money primarily by licensing its models and software to manufacturers, generally charging an annual per-device fee in addition to development costs. It also charges enterprise customers based on token usage when customizing models.

Hasani has an ambitious target: “We want to be on 50% of all the devices on the planet.”

He said potential annual licensing revenue could range from less than $1 to hundreds of dollars per device, depending on the application, though the company has not disclosed revenue forecasts.

Liquid AI remains privately held, and Hasani said a potential IPO is on his mind, although the company is still too young to pursue one immediately. The company faces challenges including competition, hardware limitations, and the difficulty of securing large-scale commercial deployments.

There are senseless valuations in AI

As investors debate whether the AI boom resembles the dot-com bubble, Hasani sees a distinction between the technology’s potential and the prices investors pay for it.

“There are senseless kinds of valuations that are coming on,” he said. “I can totally see from a traditional economic sense, we would be in a bubble, like that’s how you would define a bubble. It exactly looks like that.”

But from a technological perspective, Hasani believes AI’s biggest breakthroughs are still ahead.

“The promise of this technology is much, much bigger than what it was before,” Hasani said. 

One area he is particularly excited about is healthcare. Hasani pointed to recursive self-improvement, a process in which AI systems help develop more advanced versions of themselves.

He believes that could lead to breakthroughs in cancer research, longevity, and other areas of medicine.

“If you say that slow down the pace of innovation across the finding out how to cure cancer, I’m not gonna slow down because people are losing time,” Hasani said.

Beyond healthcare, Hasani sees opportunities for AI to help scientists answer fundamental questions about physics and the universe.

Still, Hasani pointed to some risks of developing “increasingly powerful” AI technology.

He said he has discussed AI safety with other foundation-model executives and found that researchers across the U.S. and China recognize the need for safeguards, even as competition between companies and countries intensifies.

He warned that some people could misuse AI and stressed the importance of keeping AI under human control. One priority, he said, is making AI outputs verifiable so that humans can understand and check the results.

“AI has to stay under control of humans… as long as we can make sure that’s the case, we should be fine,” he said.

Despite those concerns, Hasani remains optimistic about AI’s long-term potential.

“I am definitely worried,” Hasani said. “But you have to be cautiously optimistic.”

Related: Cathie Wood buys $11.8 million of tumbling tech stock

The company wants its Copilot AI assistant to handle more of the busywork in Word, Excel, Outlook, and Teams. For employers, that sounds like a lot of time saved. For Microsoft, it opens up a lucrative new business built on top of software millions already use.

Investors have bought into that promise, too, though it hasn’t been without hiccups. Microsoft shares have surged 10% in 2026 through Oct. 19, trailing the S&P 500’s near-12% rise, despite Wall Street’s enthusiasm for its AI ambitions, as reported by Seeking Alpha.

However, we’re now seeing that expectations are becoming harder to satisfy.

In an Oct. 19 research note shared with me, Morgan Stanley maintained its lofty $600 Microsoft stock price target, but analyst Adam Wood (4.40 star-rated by Tipranks) noted that improving sentiment and a higher valuation have raised the bar.

With Microsoft planning more than $50 billion in Q1 2027 CapEx, investors aren’t just looking for another impressive cloud-growth figure. They’re looking for proof that the spending is paying off.

Morgan Stanley sees Microsoft’s AI spending raising the stakes 

Morgan Stanley’s Adam Wood believes that Microsoft’s recent success has created a problem.

According to him, its investors are getting used to being pleasantly surprised.

In the June quarter, Microsoft generated $90 billion in revenue, up 18% year over year. Azure and other cloud services grew 43%, while adjusted earnings reached $4.74 a share, beating Wall Street’s consensus by 50 cents.

More Tech:

• Meta and Microsoft just sent employees a memo about Anthropic
• Microsoft cutting off popular software with no extension Oct. 13
• Apple’s biggest AI advantage may face new privacy challenge

That followed earnings beats of 21 cents in the March quarter and 28 cents in December. Microsoft also topped estimates in the preceding September quarter, reporting adjusted earnings of $4.13 against expectations of $3.65.

That’s a lofty record to improve upon.

Morgan Stanley expects Azure growth of roughly 47% in constant currency in Microsoft’s fiscal Q1 2027, ahead of management’s guidance for about 45%. Wood also wants the following quarter’s outlook to show that growth can hold steady, rather than fade as more data-center capacity comes online.

And building that capacity isn’t cheap.

Microsoft dropped a whopping $41 billion on CapEx in the June quarter, as reported by Reuters. Its guidance for the September quarter calls for more than $50 billion, though changes in lease accounting complicate a direct comparison.

Morgan Stanley estimates Microsoft brought about 3.25 gigawatts of additional data-center capacity online in fiscal 2026 and could add another 4.25 gigawatts in fiscal 2027. Its model has Azure AI revenue rising from around $25.7 billion to $52.2 billion over those years.

Those are remarkable numbers, but they’ve also created some enormous expectations.

Microsoft’s Copilot customers face a different AI spending problem

Open Outlook, Teams, or Excel at work, and Microsoft is probably already collecting a subscription fee. 

Now the goal is for employers to spend more on AI assistants that draft emails, summarize meetings, and eventually handle entire tasks.

For perspective, during Microsoft’s July 29 earnings call, CEO Satya Nadella told Morgan Stanley analyst Adam Wood that Copilot could potentially serve entire businesses through a combination of subscriptions and usage-based pricing.

“The TAM is much more expansive. We’re going to be very, very focused on driving customer value and then expanding with it,” Nadella said.

Microsoft already has more than 30 million paid Copilot seats, but Nadella didn’t provide a specific dollar estimate for the broader market opportunity.

Morgan Stanley sees plenty of appetite.

In its latest survey, 64% of 100 U.S. and European technology chiefs said they planned primarily to buy AI capabilities from vendors, compared with 17% planning to build them. A separate survey found 30% expected Microsoft to gain the biggest share of additional generative-AI spending, up from 27%.

But buying AI and making money from it aren’t the same thing.

Management says conversations per user nearly doubled year over year. Yet Morgan Stanley’s partner checks suggest more advanced, usage-based Copilot activity remains concentrated in selected workflows.

Premium E5 and E7 subscriptions could boost sales today, while AI tools and usage-based fees could bring in more down the road. However, customers need clear pricing and proof that these tools are worth the money.

That’s important for Microsoft’s bottom line.

However, third-party research raises questions about that payoff. Gartner reported on Oct. 19 that only 13% of surveyed CIOs had seen significant or transformational value from AI. 

That said, Microsoft is betting the benefits will spread. Investors need to see whether stronger customer usage translates into sufficient revenue and profit to justify the expansion of the AI infrastructure bill.

Microsoft’s next earnings report could determine whether AI costs pay off

Microsoft’s Oct. 28 earnings report will once again reignite the debate over whether its massive AI capital expenditures can drive growth without hurting profitability.

That means listening closely to what management says about the capacity coming online. Is Azure finding paying customers as quickly as Microsoft can build? Are businesses using Copilot beyond basic subscriptions? And can profits keep growing while expensive data centers weigh on margins?

Morgan Stanley believes Microsoft can remain free cash flow positive in fiscal 2027. Still, its longer-term forecasts allow for more borrowing and fewer share buybacks as expansion continues. 

None of this means the AI strategy is failing, though. It means investors may become less forgiving of growth that requires ever-higher spending.

Microsoft wants Copilot to become as routine at work as Outlook or Excel. The payoff depends on whether customers find it useful enough to keep paying.

Related: Elon Musk’s SpaceX could change how Americans buy iPhones

On the anniversary of the worst single day in market history, Jim Cramer told his “Mad Money” audience that the AI trade may be repeating an old mistake, one where prices stop listening to the businesses behind them, CNBC noted.

The timing makes his warning hard to ignore. Chip stocks spent most of last week sliding, while talk of an AI bubble grew louder on trading desks. For investors holding the market’s favorite names, the real question is whether the pain points to cracks in the companies or simply a crowded trade coming apart.

Also read: Jim Cramer sees something Wall Street may be missing in Broadcom

How the recent AI stock sell-off turned into a 1987 trapdoor

Thirty-nine years ago, on Oct. 19, 1987, the Dow Jones Industrial Average lost 508 points in one session. That 22.6% plunge gave Wall Street its worst one-day percentage drop ever, as reported by TheStreet, on a day that became known as Black Monday.

Much of the blame landed on portfolio insurance. As reported by TheStreet, it was an automated strategy that used stock index futures to hedge large portfolios. When prices slipped, the system sold futures to limit losses.

That works for one fund, but it backfires when dozens of big players are wired to sell at the same moment.

The product had been heavily promoted by smart traders in the years before the crash, and when markets turned, the result was herd-like panic selling that wiped out hundreds of billions in market value in a single session.

No one had to make a fresh judgment about any company for the selling to continue. Each new drop simply triggered more of the same orders.

Cramer’s reading of that day is what ties it to the present. He recalled that the automatic selling wrecked share prices even though the businesses underneath stayed healthy, and he called the collapse a spectacular buying opportunity in hindsight.

That gap between price and value is the thread he now sees running through the AI trade.

Chip and data center stocks fell together

The damage from last week was hard to miss. The Nasdaq Composite and the S&P 500 both posted sharp weekly losses in the stretch ending Oct. 16. Higher interest rates and growing worry over AI debt hung over the whole tech group.

Chip stocks got it worse. A closely watched semiconductor index fell for a seventh straight session on Oct. 16 and logged its steepest weekly loss since April 2025, Reuters reported. These are the companies building the hardware behind AI, and they slid almost in lockstep.

The spark came from the economy rather than from earnings. A hotter consumer inflation report pushed Treasury yields higher and revived concerns about how the AI boom is being financed, Capital.com analyst Daniela Hathorn said, according to TheStreet.

She added that investors are now asking whether future AI profits can justify rich valuations and pricier borrowing at the same time.

CoreWeave became the face of that fear. The AI cloud provider plunged 12% on Oct. 16 and was on pace to lose about a quarter of its value for the week as its debt financing costs climbed, as reported by TheStreet.

Even so, analysts argued the selling reflected financing worries rather than weaker demand, noting that nearly all of CoreWeave’s 2027 capacity is already sold.

Why forced selling may be driving the drop

Cramer suspects the Oct. 16 slide was made worse by a hedge fund being forced to dump stock. No such liquidation has been confirmed.

Still, he argued that some data center names are trading with little connection to their fundamentals, rising in bunches and falling in unison.

More Jim Cramer:

• Jim Cramer says Scott Bessent is fighting a flood with one finger
• Jim Cramer spots something investors may be missing on Wall Street
• Jim Cramer calls one major data center play a no-brainer buy

He has seen this movie before. Leopold Aschenbrenner’s Situational Awareness, which had peaked at roughly $45 billion in assets, slashed its listed stock positions in a matter of weeks, after a sharp momentum reversal set off margin calls and compulsory sales, TheNextWeb noted.

The fund had loaded up on chip, data center, and power names while betting against software, so both sides lost money at once.

That unwind hit stocks that had done nothing wrong. Cramer said the liquidation, along with selling from investors who had copied the fund’s trades, flooded the market with AI infrastructure shares, although that would need to be assessed company by company.

What happened next is the part he wants investors to remember. Once the forced seller was gone, tech stocks ripped higher because there were no large sellers left to stand in the way, and Cramer described the episode as a clearing event.

If last week carried the same fingerprints, the bounce could come just as quickly.

Green Monday brings early relief for tech

Oct. 19 offered a first sign that the panic may be fading. The Nasdaq rose about 1.1% by the closing bell, and the S&P 500 added 0.41%, while the Dow finished roughly flat. That was the Green Monday Cramer had in mind.

Some of last week’s hardest-hit names led the bounce. CoreWeave climbed after analysts flagged the dip as a buying opportunity, and the days ahead bring earnings from Netflix, Tesla, and Meta that could decide whether the rebound sticks.

Before hitting the sell button, Cramer wants investors to ask one question: Is the stock falling because the business is getting weaker, or because of forces outside the company’s control, such as money managers who took on too much risk?

The first deserves caution. The second, as 1987 showed, can become a rare chance to buy.

Related: Jim Cramer sees something Wall Street may be missing in Broadcom

On Wall Street, that’s incredible news. 

More construction leads to more computer chips, and another chance to cash in on the relentless AI boom. For perspective, according to S&P Dow Jones Indices, the Magnificent Seven (mostly AI stocks) accounted for a head-turning 55% of the S&P 500’s total returns between 2023 and 2025. 

But if you’re the one living nearby, your questions are likely to be a lot different. Will all those servers push up electricity costs? Will the project strain local roads and water supplies? Moreover, if AI is supposed to make businesses more efficient, what happens to the jobs people already have?

It’s not tough to see why that enthusiasm might look different from the other side of the fence.

That disconnect just caught the attention of Dario Perkins, a senior economist at TS Lombard. While investors butt heads over AI CapEx, Perkins sees another potential threat: Americans who may not want the consequences in their backyards.

And growing opposition could turn into a bigger financial problem.

Read more: Palo Alto CEO wants AI labs out of the ‘fear-mongering business’

Americans are souring on the data centers powering the AI boom

Wall Street has spent the bulk of the past three years worrying if the AI boom has gotten too expensive. Dario Perkins thinks investors might be missing a different kind of problem.

“AI boosters on social media live in a bubble,” Perkins wrote in a Monday research note as reported by Business Insider. His blunt remark serves as a reminder about Americans unhappy with the industry’s expansion: “And they vote.”

That could sound like an unusually political risk for a Wall Street forecast, but the public-opinion numbers are tough to dismiss.

In a Pew Research Center survey released in September, 60% of Americans said they’d be uncomfortable with a new data center operating in their area. Just 7% said they’d be very or extremely comfortable.

More Artificial Intelligence:

• AI’s biggest business opportunity may not look like AI
• AI racks are outgrowing old power: 2 chipmakers have a strategy
• BofA splits software into winners and losers as AI agents arrive

However, the sentiment has soured even more, as in January, 30% thought data centers were mostly bad for the quality of life of nearby residents. By August, that share reached 49%. The percentage who said data centers were bad for home energy costs rose from 38% to 50%.

And the concern goes beyond electricity bills.

In a separate Pew survey, 71% of Americans expected AI to reduce U.S. jobs over the next two decades, up from 64% in 2024. 

That said, polling alone doesn’t stop construction. But in some communities, objections have already turned into delays and cancellations.

Local opposition threatens a massive bet on AI infrastructure

The backlash is showing up in development plans already.

Goldman Sachs says 86 U.S. data-center projects have been canceled or extensively delayed since January, with 76 linked directly to local opposition. So clearly, things have gotten past the point of angry comments at town meetings. It’s a potential problem for companies counting on those facilities to open on schedule.

And it’s easy to speculate on why neighbors are paying attention.

Consider a 100-megawatt data center running at full capacity around the clock. It would consume 876 million kilowatt-hours a year, about as much electricity as 83,000 average American households, based on federal consumption figures. 

Still, power costs aren’t hypothetical. PJM, the grid operator serving parts of 13 states and Washington, D.C., said through its independent market monitor that existing and expected data-center demand accounted for 9% of wholesale electricity costs through July 2026.

For Perkins, this is exactly where a neighborhood fight becomes Wall Street’s problem.

Building a data center entails years of spending before a customer starts paying for its computing power.

Delay an opening, and the financing costs don’t disappear. Perkins estimates debt across parts of the AI ecosystem could approach $2 trillion by the end of 2027. S&P Global has separately warned that complex AI financing deals might leave some obligations less visible to investors.

There is a catch to the backlash narrative, though. SemiAnalysis found that many local moratoriums have barely affected projects already underway; roughly 2.3 gigawatts of the 21 gigawatts in affected areas are likely to face delays.

AI construction isn’t stopping. But companies that can secure power, permits, and affordable financing may be the ones best positioned to weather the resistance.

Electricity costs and borrowing risks could decide AI’s next chapter

Perkins comments leave Big Tech with some difficult choices.

Companies could shoulder more of the costs of power infrastructure, while utilities introduce special rates for large data-center customers. Developers might also seek locations with fewer electricity constraints or secure dedicated power supplies.

That said, none of these solutions comes for free, and many of the Big Tech giants are under immense pressure already. 

According to FactSet, Microsoft (MSFT), Alphabet (GOOGL), Amazon (AMZN), Meta Platforms (META), and Oracle (ORCL) are expected to drop more than $690 billion on capital expenditures in fiscal 2026, relying a ton on external financing as their AI investments strain free cash flow. 

For investors, the stakes extend well beyond Nvidia’s next chip orders. Rising borrowing costs, mounting project delays, and changes to spending plans at Microsoft, Amazon, Alphabet, and Meta could signal early trouble.

Equally important is whether cloud-related sales and demand for paid AI services grow quickly enough to justify the enormous investments already underway.

A slowdown wouldn’t necessarily mean AI has failed. It could simply mean the financial and physical costs of expansion are becoming harder to ignore.

Wall Street is betting that AI will transform the American economy.

Perkins’ warning suggests ordinary Americans may have more say in how quickly that future arrives than investors realize.

Related: Bank of America sees a costly catch in America’s strong economy

Artificial intelligence may expand that edge, according to KeyBanc.

The firm reiterated its Overweight rating and $350 price target on Amazon, forecasting Amazon Web Services revenue growth of 38% year over year, slightly ahead of the 37% consensus estimate. KeyBanc sees Amazon Bedrock and the company’s custom artificial intelligence chips as important drivers, according to Investing.com.

The call became more interesting on Oct. 19.

In order to enable developers to run code created for the well-known PyTorch artificial intelligence framework on Amazon’s Trainium chips with nearly no rewriting, the company released TorchNeuron. The software ecosystem that discourages developers from switching hardware is one of Nvidia’s (NVDA) main advantages. Amazon’s stock increased by roughly 5% following the announcement, according to Reuters.

For investors, this means that KeyBanc’s $350 call is about more than just the expansion of the cloud business.

Amazon is attempting to profit from AI in three ways: by renting out processing capacity, by selling model access via Bedrock, and by using AI to increase the value of its massive retail business.

KeyBanc sees AWS keeping its AI momentum

KeyBanc’s prediction has a solid foundation thanks to Amazon’s most recent performance.

AWS revenue surged 37% to $42.2 billion in the second quarter, its fastest growth rate in 18 quarters. AWS operating income climbed to $16.6 billion from $10.2 billion a year earlier, according to Amazon.

CEO Andy Jassy said the company’s artificial intelligence business had exceeded a $25 billion annual revenue run rate and was growing at a triple-digit percentage rate. Amazon’s chips business separately exceeded a $25 billion annual run rate and was also growing at triple-digit rates.

These numbers contribute to KeyBanc’s prediction that AWS growth would stay at around 38% rather than rapidly declining as the company grows.

Amazon is using both strategies to assault the market.

Bedrock offers multiple artificial intelligence model providers to corporate customers. Amazon reported that hundreds of thousands of customers use Bedrock and that second-quarter spending on the platform exceeded all previous quarters.

Trainium comes next.

For clients concerned about the cost of training and operating large AI models, Amazon’s bespoke AI processors are intended to offer an alternative to Nvidia graphics processing units.

Software has always been the problem.

Nvidia’s CUDA environment has been the focus of years of development. Even if the competitor chip is less expensive, shifting workloads to a new CPU may entail significant technical effort.

TorchNeuron goes straight after the obstacle.

Reuters reported that more than 30 early users are already testing the tool across robotics, voice AI and other workloads, with Amazon saying some customers have achieved meaningful cost reductions by shifting work to Trainium.

Amazon has another AI business hiding in plain sight

KeyBanc’s argument doesn’t end with AWS.

The firm expects Amazon management to emphasize its artificial intelligence assistants and the commercial potential of Alexa for Shopping, according to Investing.com.

More Amazon:

• Amazon is coming for your quick Walmart run
• Amazon’s Alexa shopping tool faces its first big October test
• Amazon pulls a best-selling pain reliever again, and customers need to check their medicine cabinets

Since cloud computing and chips get the majority of Wall Street’s AI attention, it is easy to undervalue that retail viewpoint.

Earlier this year, Amazon created Alexa for retail by merging Alexa+ with Rufus, its retail assistant. Consumers may view pricing history, compare items, receive suggestions, and automate transactions.

Amazon says active users of its AI shopping tools nearly doubled year over year during the second quarter, while interactions increased more than fivefold. More importantly for investors, U.S. customers using Alexa for Shopping spend more than 40% more per order on average than shoppers who do not use it.

This transforms AI from a characteristic of technology into something that investors can quantify much more easily.

Retail sales may increase if an AI assistant helps Amazon consumers find more items and create larger baskets. Amazon may also boost advertising income if companies pay to participate in such AI-driven talks.

Amazon Ads says more than 300 million customers used its earlier shopping assistant Rufus in 2025. The company has since combined Rufus with Alexa+ and introduced advertising formats specifically designed for conversational shopping.

Google and Microsoft may sell AI infrastructure. Nvidia can sell the processors that power it.

Amazon can offer the infrastructure and immediately present artificial intelligence to customers ready to make a purchase.

Wall Street is getting more bullish on Amazon’s AI math

KeyBanc is no longer alone at $350.

On Oct. 19, Morgan Stanley named Amazon its top pick, raised its price target to $350 from $335, and maintained an Overweight rating, according to Investing.com. The firm sees roughly 40% upside to its target.

When it comes to AWS expansion, Morgan Stanley has grown even more assertive than KeyBanc.

The firm now expects AWS revenue growth of 47% in 2027 and 42% in 2028, helped in part by greater availability of Nvidia graphics processors. It also sees Amazon generating roughly $16.50 in earnings per share in 2028.

Those estimates underline an important change in the Amazon investment story.

After consumers spent many quarters optimizing their cloud expenses, investors spent years wondering whether AWS’s growth would get back on track.

The question of how much more profit Amazon can make as the need for AI in AWS, chips, advertising, and commerce increases is gradually taking shape.

The solution will be expensive.

In contrast to a $18.2 billion inflow a year earlier, Amazon’s trailing 12-month free cash flow swung to an outflow of $7.6 billion in the second quarter. A $66.1 billion year-over-year increase in property and equipment spending, mostly for artificial intelligence infrastructure, was the main cause of Amazon’s decline.

Investors should be aware of this tension.

AWS is expanding quickly, but so is the amount needed to sustain that expansion.

Amazon is hardly alone. S&P Global warned this week that major hyperscalers are increasingly using complex financing structures to support AI infrastructure, while the industry’s capital requirements continue to climb, Reuters reported.

Amazon’s $350 case depends on turning AI spending into profit

KeyBanc still regards Amazon as one of the most alluring mega-cap companies.

Despite minor adjustments for foreign currency impacts, the company’s profits predictions are still 4% and 3% over consensus, respectively, and the shares are valued at around 23 times its 2027 earnings estimate and 18 times its 2028 earnings. India’s Investing.com

That valuation argument becomes much more compelling if AWS continues to grow at close to 40%.

If capital expenditures increase faster than the returns from such investments, it loses some of its appeal.

I find it noteworthy that Amazon now has many ways to benefit from its AI investment.

Through AWS, it may rent computer resources from Nvidia and Trainium. Bedrock may charge businesses for using it to access models. Using bespoke silicon may reduce the cost of its infrastructure. Additionally, it may convince customers to add additional things to their shopping carts by using AI assistants.

The recent TorchNeuron revelation strengthens one especially significant component of that equation.

When it comes to creating the fastest microprocessor in the world, Amazon doesn’t need to outperform Nvidia.

It must make Trainium so excellent and user-friendly that AWS clients have a reliable alternative.

This might reduce Amazon’s reliance on Nvidia while providing cloud users with an additional incentive to remain on AWS.

Therefore, KeyBanc’s $350 goal hinges on more than just excitement for AI.

Amazon must demonstrate that its massive expenditures are generating higher profitability, quicker cloud growth, and more valued customers.

So far, Wall Street increasingly thinks it can.

More Investing, Markets, and Tech News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Talk of resumption of Iran war rattles markets
• Microsoft wants to keep more of its AI billions from Nvidia

Anyone who owns an S&P 500 index fund already owns both, along with other stocks that comprise the Magnificent Seven. Collectively, those seven companies account for nearly 35% of the benchmark. 

Truth be told, for the better part of the past three years, that concentration was a good thing. Big Tech powered the market higher, driving record gains in retirement and brokerage accounts.

For perspective, Fidelity Investments reported that average 401(k) balances reached a record $155,800 in Q2 2026, up 13% year over year, backed by market gains and continued savings.

Also read: 40-year market veteran shares S&P 500 outlook for 2027

But something has changed in 2026.

Wall Street veteran Ed Yardeni, who has spent years favoring major technology stocks, says the S&P 500’s other 493 companies are now outperforming the celebrated seven.

And he thinks the shift has a lot to do with who’s paying for the AI boom, and who’s positioned to profit from it.

The Magnificent Seven are no longer running away from the S&P 493 

For years, especially since the AI boom took off following the release of OpenAI’s ChatGPT, betting on America’s biggest tech companies has been one of the easiest ways to outperform the broader stock market.

The Magnificent Seven, which includes Apple (AAPL), Microsoft (MSFT), Nvidia (NVDA), Amazon (AMZN), Alphabet (GOOGL), Meta Platforms (META), and Tesla (TSLA), were almost unstoppable.

In 2023, the group surged 76%, compared with just 11% for the other 493 companies in the S&P 500, according to J.P. Morgan Asset Management.

The dominance persisted in 2024, when the seven gained 48% versus 14% for the rest. In 2025, they won again, returning 23% compared with 13%.

That’s a tremendous run, particularly given how much these stocks influence the retirement accounts and index funds owned by millions of Americans.

But 2026 is telling a different story.

In his October 19 market commentary, veteran Wall Street strategist Ed Yardeni reported that his “Impressive 493” had gained 14.1% year-to-date, outperforming the Magnificent Seven’s 10.2%, based on his ETF comparison.

Yardeni foresees a shakeup coming.

In December 2025, after roughly 15 years of favoring major technology stocks, he recommended underweighting the Magnificent Seven, anticipating stronger competition and better opportunities elsewhere in the market, as reported by Reuters.

Interestingly, the shift isn’t happening because Big Tech’s profits have fallen off a cliff.

J.P. Morgan Private Bank expects Magnificent Seven earnings to grow 59% in 2026, far exceeding the 23% projected for the remaining S&P 500 companies. So, even with explosive earnings growth, Yardeni believes investors have become less enthusiastic about the usual suspects.

Yardeni sees AI’s biggest companies facing an expensive new problem

Yardeni argues that the bigger an AI company gets, the fiercer the competition it faces, often from rivals that are just as big.

According to him, AI has basically turned the Magnificent Seven’s once-protected businesses into a costly battleground. Microsoft, Amazon, Alphabet, and Meta are pouring billions into chips and data centers, raising questions about how much of that spending will eventually translate into profits and free cash flow.

Yardeni Research explained the opportunity for other companies back in December:

“The winners may not be among the Mag-7 at all but the S&P 500’s Impressive 493 and the economy at large.”

Meanwhile, lenders are beginning to scrutinize the AI spending boom more closely.

In a recent CNBC interview, Yardeni talked at length about concerns surrounding AI infrastructure provider CoreWeave (CRWV) and how they have prompted bond investors to reassess the risks of financing these projects.

“Well, the bond market is finally paying attention to credit quality in the bond market for AI credits, which is good. I mean, it’s good if the bond market gets ahead of the credit rating agencies and really assesses the riskiness.”

Yardeni considers that a healthy development. More expensive financing might discourage riskier projects and help prevent another technology bubble.

The other 493 companies, meanwhile, may have an advantage. Businesses across healthcare, finance, and manufacturing can adopt AI to improve productivity and profit margins without spending billions on the underlying infrastructure.

Some, including Broadcom (AVGO) and Advanced Micro Devices (AMD), are themselves major AI suppliers.

Yardeni’s market outlook raises a new question for investors

That said, Yardeni hasn’t abandoned his long-term bull case for stocks. He’s just less certain how much investors will pay for future earnings.

In his CNBC interview, he warned that higher oil prices, stubborn inflation, and rising bond yields could pressure stock valuations, even as corporate profits remain strong.

He now targets 7,900 for the S&P 500 by year-end 2026, with 8,400 more realistic by mid-2027. He’s still looking for 10,000 by the end of the decade, although those forecasts are far from guaranteed.

For everyday investors, the bigger question may be how much Big Tech they already own.

Traditional S&P 500 index funds give the largest companies the biggest weight, leaving investors heavily exposed to the Magnificent Seven. Equal-weight funds or strategies excluding those seven offer broader exposure, although they carry their own risks.

Yardeni’s outlook also gives investors something to watch as they look to reassess their portfolios.

Interestingly, Jim Cramer recently raised a similar concern about investors being overly exposed to the tech stocks. I covered the October 12 episode of Mad Money, where Cramer pointed out to a caller that owning several familiar companies doesn’t necessarily mean having a diversified portfolio.

He pointed out that stocks tied to the same investment themes can fall together, even when their businesses differ. 

Cloud revenue, AI spending commitments, corporate borrowing costs, and Treasury yields will help determine whether the technology boom remains sustainable. Rising earnings estimates and improving market breadth outside Big Tech would strengthen his rotation thesis.

Related: Bank of America sees a costly catch in America’s strong economy

Microsoft left while reviewing its data center portfolio and its share of power generated on site, a source told Semafor. Microsoft declined to comment.

That detail is important because Monarch is designed to make its own electricity. Nscale, which owns the campus, expects up to 2 gigawatts of generation there by the first half of 2028, according to its IPO filing. A tenant there is betting on a private power plant, not just renting servers.

Now the company that replaced Microsoft wants more. Anthropic, the private company behind the Claude chatbot, is in talks to buy computing power from a second Monarch building. The deal could be worth about $45 billion, Bloomberg reported on Oct. 19.

Anthropic’s current Monarch agreements are worth up to about $44.6 billion, the Nscale filing shows. The new building would match the first in size and double Anthropic’s footprint at Monarch. Anthropic also holds first dibs on a planned third building, Bloomberg noted.

Bloomberg’s sources said a deal could close soon in the coming days, though no decision has been made. I read Microsoft’s exit less as a verdict on the site than as a choice about who carries the risk.

Also read: Musk called Anthropic ‘evil’; SpaceX’s deal with it nearly doubled

Microsoft left the building but kept the tenant

Microsoft never walked away from Anthropic itself. In November 2025, it committed to invest up to $5 billion in the AI lab. The same announcement had Anthropic agreeing to buy $30 billion of computing capacity on Microsoft’s Azure cloud.

That stake is already paying off. Microsoft booked a $3.2 billion gain on it in the quarter ended in June, TechCrunch reported.

Nor did Microsoft drop Nscale, which the filing shows is expanding its work in Portugal. Microsoft simply stepped back from this one campus.

I’d argue that Microsoft got the cleaner side of this trade. It shares in Anthropic’s growth without committing to years of payments at Monarch. The heavier load now sits with the landlord.

Nscale’s IPO pitch now leans on a single customer

Nscale, a London company launched in 2024, has applied to list on the New York Stock Exchange under the ticker NSCL. It booked $140.6 million in revenue in the first half of 2026, its filing shows. That is a thin base for a company promising tens of billions in computing power.

The filing also puts Nscale’s total contracted value at about $103 billion as of Aug. 31, 2026. That backlog is the heart of its IPO pitch.

By my math, Anthropic’s agreements make up as much as 43% of that backlog. A second deal of similar size would push that share past half.

Nscale’s filing flags the danger itself, warning that a handful of customers drive much of its revenue. Meanwhile, some investors have urged Nscale to lower its target of about a $30 billion valuation, Bloomberg reported.

A second Anthropic contract would help answer that pushback, which may explain the timing. Still, I see a mixed blessing for IPO buyers, who would largely be betting on one customer paying for years.

Main Street already owns the safer side of this deal

You can’t buy Anthropic or Nscale shares today, since neither trades publicly. Even so, anyone with an S&P 500 index fund already owns a slice of Microsoft and Nvidia Corp. (NVDA), and both sit on what I see as the safer side of this deal.

Microsoft owns the Anthropic stake that is already showing gains. Nvidia pledged up to $10 billion to Anthropic in the same November 2025 announcement. It also joined Nscale’s latest funding round, Bloomberg reported, and its Vera Rubin chips would power the new building.

That is the key difference for everyday investors. Index fund holders benefit if Anthropic keeps growing, without betting the house on one landlord. Anyone buying Nscale at its IPO would be making a narrower bet, tied heavily to a single tenant.

If that IPO tempts you, read the final prospectus first. Check how much of the backlog Anthropic represents once a second deal is counted. Nscale hasn’t started marketing its shares and is still answering SEC questions, Bloomberg reported, so there is time to do that homework.

For Claude users, the payoff will take time. Anthropic said earlier this year that heavy demand had strained its systems, yet Monarch only aims to come online in 2027, Semafor reported. Steadier service from this deal is a 2027 story at the earliest.

More Artificial Intelligence:

• AI’s biggest business opportunity may not look like AI
• AI racks are outgrowing old power: 2 chipmakers have a strategy
• BofA splits software into winners and losers as AI agents arrive

Tech giants are handing AI’s power risk to startups

Step back, and Monarch looks like a template. Tech giants secure AI capacity through stakes and cloud contracts, while younger firms such as Nscale borrow, build, and own the power plants.

That split has a cost. Investors have grown wary of the borrowing behind the data center boom, Bloomberg reported, turning against chipmaker shares and cloud companies’ bonds.

The setup works as long as AI labs keep growing into their contracts. If they stumble, the losses land first on the companies with the shortest track records and fewest customers.

Whether this second deal closes, and how Nscale’s IPO prices afterward, will show who Wall Street thinks should hold that risk. The campus Microsoft passed on may end up revealing where the AI boom’s weakest link is.

Related: Anthropic’s $2 trillion IPO dream hits a new reality

The employee had initially suspected the request might be a phishing attempt. The call removed that suspicion, and the employee transferred roughly $25 million.

The attackers had not cracked a password or exploited a software vulnerability. They had built convincing digital replicas of people the employee recognized. Standard fraud controls were not designed to detect the difference between a genuine video call and a fabricated one.

Financial institutions are now weighing what verification should look like when a voice, face, or digital identity can be replicated with AI tools.

World, a digital identity company co-founded by Sam Altman, brought a suite of products for banking and finance to Money20/20 on Oct. 18, positioning human verification as part of the answer.

Also read: Michael Burry just made another bold move on Wall Street

Where AI-driven impersonation finds its opening in banking

Most financial security systems check whether someone holds valid credentials. A correct password, a registered device, or a one-time code confirms that someone controls an account. Those checks do not independently confirm that the person behind the credentials is who they claim to be.

The gap persists because credentials are proxies. A password, a registered device or a one-time code confirms that someone controls an account. It cannot confirm whether the person using those credentials is the account’s legitimate owner.

The gap creates two distinct exposures. For individual customers, a fraudster who takes over an account can initiate transfers that the system processes as legitimate.

For enterprises, stolen credentials can be used to access confidential financial data or authorize payments without triggering standard controls. The 2024 Arup transfer cleared existing security systems without flagging an alert.

Deloitte estimated that AI-enabled fraud losses in the United States could reach $40 billion by 2027, up from roughly $12.3 billion in 2023. The estimate reflects how AI tools have lowered the cost and technical barrier to producing convincing impersonations.

The push to add a human layer on top of existing security

Financial institutions have begun exploring verification approaches that go beyond confirming credentials and check whether a specific, verified human is present. The shift reflects a broader recognition that the credential layer was not built to catch impersonation at the level AI tools now make possible.

At Money20/20 on Oct. 18, World co-founder Alex Blania announced World ID for banking and finance, a suite that combines an identity wallet for storing verified human proofs, a video verification tool for confirming whether a meeting participant matches a previously enrolled identity, and a framework for delegating verified proofs to AI agents acting on a person’s behalf.

More than 18 million people hold a Proof of Human credential through the World network.

Ajay Patel, head of World ID at Tools for Humanity, told TheStreet that the standard for high-stakes interactions needed to shift. “Knowing that an account is authenticated is important; having confidence in the human behind it should become part of the foundation, rather than an exception reserved for when something looks suspicious,” he said.

World is not the only company working in this space. Digital identity verification has drawn increased investment as financial institutions contend with fraud that existing controls were not designed to catch.

Bain Capital, VanEck Funds, and OpenExchange are among the organizations exploring how proof-of-human technology applies to banking use cases.

Why proving someone is human is only part of the answer

Yuannan Yang, a security engineer at CertiK, said the threat has grown beyond what identity verification alone addresses. “The biggest threat goes beyond deepfakes or voice cloning,” Yang told TheStreet. “What’s more concerning is how AI is making social engineering more convincing, personalized, and scalable.”

The 2024 Arup case illustrated that point. The employee who authorized the $25 million transfer had been alert to phishing before joining the call.

What overcame that judgment was a fabricated social environment. The video appeared to include colleagues the employee recognized. Confirming that the participants were genuine, verified identities would have caught the fraud. Yet confirming only that they were live humans would not have been enough on its own.

More Wall Street:

• Nvidia and Micron are about to lord their successes all over Wall Street
• Nvidia, Palantir, and Alphabet send waves through Wall Street
• Wall Street elites will pocket windfall from record $46 billion profits

Proof-of-human verification addresses one layer of a larger problem, he said. “Even if a bank successfully verifies that a real person is initiating a transaction, that person could still be acting under fraudulent instructions or being manipulated by a scammer.”

A full response to AI-driven fraud requires combining identity verification, device authentication, transaction monitoring, and independent authorization within a risk-based framework, Yang said.

The goal is not only to establish that a real person is present but to confirm the person’s identity and that the action they are authorizing is legitimate.

What banks will need to see before adopting at scale

Banks evaluating proof-of-human technology will consider whether the security benefit justifies the integration work. The approach needs to fit into existing identity, authorization, and compliance systems without requiring institutions to replace them or build new centralized biometric databases.

Banks will also weigh the user experience. Verification steps that slow legitimate customers down can reduce adoption of digital services. Any proof-of-human check will need to deliver stronger assurance without adding meaningful friction beyond what users already expect from standard authentication.

Regulatory attention to the problem has been growing. In October 2025 remarks reported by American Banker, Federal Reserve Governor Michael Barr cited the rapid growth of AI-driven impersonation as a threat to banks of all sizes.

The sector has not previously required proof-of-human checks at scale, and no institution has yet produced performance data from a broad deployment.

Adoption will depend on whether early pilots demonstrate that the technology reduces fraud losses without adding friction that legitimate customers notice.

Related: Nvidia, Palantir, and Alphabet send waves through Wall Street

Personal computers reached that point. So did the cloud.

Quantum computing hasn’t quite gotten there yet.

Developers can already rent time on quantum computers through IBM (IBM), Amazon (AMZN) or Microsoft (MSFT). But access is only part of the equation.

Each platform has its own programming tools and billing rules, which can make moving between systems more complicated and costly.

That means the skills and budget you build on one system don’t carry over cleanly to the next.

It matters for your money even if you never write a line of code.

Quantum computing companies topped $1 billion in revenue for the first time in 2025, according to McKinsey.

The consulting firm projects that figure could hit $4.4 billion by 2028. But it can only happen if businesses can actually put these machines to work.

Now a small company in Casper, Wyo., says it can fix the friction.

Postquant Labs launches a one-language quantum tool

Postquant Labs launched its Quip Virtual Machine, or QVM, on Tuesday, Oct. 20, according to a company statement.

The QVM allows developers to write an optimization program once and run it on supported quantum and classical computing hardware, without having to rewrite it for each system.

That includes annealers, the type of machine D-Wave Quantum (QBTS) builds, and gate-based systems like those from IBM and IonQ (IONQ).

Classical computers check the answers. Developers can also publish programs and get paid each time someone else runs them, much like an app store.

“We created the QVM because quantum computing is still fractured between competing instruction sets and assemblers, providing few easy onramps to outsiders,” Postquant Labs CEO Colton Dillion said in the statement.

“Our virtual machine has three major goals: to unify and clarify pricing of quantum operations and memory across architectures, to distribute hybrid computation across available machines on our network, and to provide an easier abstraction for new quantum programmers that doesn’t require them to think directly in quantum gates or qubit layouts,” he said.

“This initial release brings together annealers, gate-based systems, and classical computers under one constraint-based programming language, and marks our first steps toward a unified representation for all of computing,” Dillion added.

Every vendor runs its own meter

Here’s the reality: renting quantum time means signing up for one vendor’s meter. Those meters don’t look alike.

More Quantum Computing:

• Bank of America makes fresh 33% call on surging quantum stock 
• IBM CEO sends blunt message on quantum computing
• Nvidia once rattled IonQ stock. Now it plans to install IonQ tech

IBM’s prices apply only to IBM hardware. Postquant Labs claims Microsoft’s Azure Quantum typically requires about $1,000 in spending before a first circuit can run.

Programming is fragmented, too. Each vendor’s software stack works differently, so code written for one machine usually has to be reworked for another.

Here’s what IBM and Amazon actually charge

I ran the published list prices to see what a first project actually costs. The math adds up fast.

Here’s how the numbers stack up, according to IBM’s pricing page and Amazon’s Braket pricing page:

Quantum computing prices at a glance

• IBM pay-as-you-go: $96 a minute, or $5,760 an hour
• IBM Flex plan: $72 a minute, with a 400-minute minimum of $28,800
• IBM Premium plan: $48 a minute, with a 5,200-minute minimum of $249,600 a year
• Amazon Braket: $0.30 per task, plus a per-shot fee that varies by machine
• Amazon’s own 10,000-shot example: $9.30

None of those prices is outrageous for a corporate research lab. The catch is that each one covers a single platform, so comparing them takes real work.

Each job gets a price ceiling upfront

Postquant Labs’ answer is metered pricing with a ceiling. Each job’s cost depends on how many qubits it uses, how long it runs and how much entanglement it needs.

Every job has a base price plus an optional priority fee. What you see is the maximum cost before anything runs.

Postquant Labs Chief Technology Officer Richard Carback calls those metered units “gas.” He argues that predictable pricing is what turns quantum into a business tool.

“Using a pay-as-you-go ‘gas-based’ cost model is the missing piece that demonstrates quantum computing is commercially viable and practical,” Carback said.

“When you can predict exactly what a computation will cost before you run it, quantum tests that began as a research experiment become a real tool for engineering because you can quantify exactly how much value it can bring to your products. This is the shift we will be delivering,” he added.

There’s a catch. The test version comes with free compute credits, but the company hasn’t published what paid jobs will cost.

The 1,930-race demo comes with caveats

Portfolio optimization is the math behind picking an investment mix that balances return and risk. It’s one of the jobs quantum computers are supposed to do well.

At the Quantum.Tech World event, the QVM ran 1,930 races between a quantum solver and a classical solver, according to Postquant Labs’ demo write-up.

About 100 visitors created the problems. Results within 0.5% of each other were counted as ties.

The company claims its quantum solver matched or beat the classical one in every race. When the gap topped 0.5%, quantum built the better portfolio by a median of 2.56%.

In my analysis, those numbers need more context. The company ran the demo itself, and the release doesn’t name the classical solver or explain how “better” was scored.

Speed wasn’t the edge, either. Both solvers typically answered in 0.1 to 0.3 seconds.

Universities kick the tires on the QVM

Researchers at Monash University, the University of Notre Dame and Purdue University are assessing the QVM. Their pitch is about saving time, not raw power.

“Oftentimes, when we put a student on a project, it takes them the whole semester just to learn how to program using the latest quantum paradigms,” Monash University professor Terrence Mak said.

“As we evaluate Quip’s virtual machine, we believe it can speed up student training and give them more time to do real research instead of spending cycles learning a specific programming language or how to program quantum circuits,” Mak added.

“Working with Quip’s QVM can help researchers and students spend less time learning specific quantum programming tools and more time applying quantum methods to practical research problems, including our work on the Lagrangian operator,” University of Notre Dame doctoral candidate Shravan Kumar Sharma said.

SiC Systems wants faster answers

SiC Systems is exploring whether the QVM can handle industrial problems. These include artificial intelligence-driven manufacturing and the control of chemical and biological processes.

“At SiC Systems, we face a variety of combinatorial optimization problems, and time is of the essence. We want to explore as many solutions as possible in as short a time as possible,” SiC Systems Chief Science Officer Soheil Mansouri said.

“This is particularly important in our physics-informed agent architecture to design, control and operate complex chemical and biological manufacturing processes. Quip Network’s QVM can help us meet that requirement and better serve our customers,” he added.

Every job also passes three layers of checks. Quantum results are compared against classical simulations, and classical providers benchmark them independently.

A tamper-evident record then ties each result to the machine that produced it.

Should IonQ and D-Wave investors care?

If you own quantum stocks, the hardware gets the headlines. Software that cuts the semester students spend learning to program could matter just as much.

Easier tools widen the pool of customers who can use the machines. That’s the bull case for IonQ and D-Wave Quantum.

The realistic case is that both companies are still posting operating losses, according to The Motley Fool. Quantum stocks also routinely swing by double digits.

Postquant Labs’ launch is a test release from a small company, and its demo numbers are self-reported. Treat it as a sign of where the industry is heading, not a reason to chase a stock.

The machines keep getting better. The bigger question is whether anyone besides a physicist can afford to use them.

Related: Cathie Wood buys $11.8 million of tumbling tech stock

He spoke on the sidelines of the IMF and World Bank annual meetings in Bangkok. His reason: “the gap between demand and supply for semiconductors is still very wide.”

The timing took some nerve. Shares of SK Hynix Inc. (SKHY) fell about 10% in Seoul on Tuesday, October 13, Korea JoongAng Daily reported.

Fears of rate hikes and Chinese competition drove that selloff, Seeking Alpha noted. Korea JoongAng Daily tied the rate worries to U.S. inflation data due before the Federal Reserve’s October meeting.

Lee’s case rests on a real shortage. Data centers built for chatbots and other AI tools have soaked up much of the world’s memory, Reuters reported. Chipmakers now favor those bigger spenders over phone and laptop makers, the report added.

Reuters counts SK Hynix and Samsung Electronics Co. (SSNLF) among the three biggest suppliers and their windfall now flows into Seoul’s treasury. This helps explain why the country’s finance minister sounds like a chip bull.

Also read: Chip toolmakers can’t catch a break, even as TSMC spends more

Seoul’s budget now rides on two chipmakers

Together, Samsung and SK Hynix make up about half of the Kospi, South Korea’s main stock index, Seeking Alpha reported. Lee acknowledged that heavy weighting has made the market swing harder this year.

The index shows how hard. It topped 9,000 in June, then briefly sank to its lowest level since late July on October 13, AP reported.

More importantly, the two companies’ profits now prop up the government’s budget.

Samsung expects about 107.4 trillion won ($80 billion) in third-quarter operating profit, according to its earnings guidance.

That is nearly nine times last year’s third quarter, the same release shows. Profits like these turn into corporate tax bills a few months later.

As a result, Seoul now expects record tax revenue this year. Lee told CNBC the chip boom will leave the country in a stronger fiscal position.

The 2027 forecast lines up with Seoul’s tax calendar

That fiscal boost arrives on a delay. Korean companies pay corporate tax in two installments, in March and August, Kyunghyang Shinmun explained in August.

Tax collected in 2027 will therefore reflect chip profits from late 2026 through the middle of 2027. The paper also reported that Seoul’s 2027 budget counts on about 580 trillion won ($430 billion) in tax revenue.

That makes Lee’s timeline convenient for his budget. To be fair, Samsung and SK Hynix themselves expect memory demand to stay solid through 2027. Still, I would treat his forecast as much a policy view as a market call.

Korea has lived through the reverse. National tax revenue fell in both 2023 and 2024, Kyunghyang reported. The same lag helps explain why 2024 stayed weak, since those bills still reflected the 2023 chip slump.

Lee’s own reasoning holds the biggest risk

That history is why one part of Lee’s answer deserves a second look. He said supply is still being expanded. I would watch that line closely, since every new factory that eases the shortage trims the windfall.

Meanwhile, China is adding pressure. ChangXin Memory Technologies, known as CXMT, said it had made progress on a denser memory design, CNBC reported. CXMT has not yet reached mass production, and its yields remain unproven.

Even Korean commentators are bracing for that turn. CNBC highlighted a Maeil Business Newspaper column urging Seoul to prepare, since new capacity could make memory cutthroat by 2029.

More Artificial Intelligence:

• AI’s biggest business opportunity may not look like AI
• AI racks are outgrowing old power: 2 chipmakers have a strategy
• BofA splits software into winners and losers as AI agents arrive

Korea’s chip windfall is America’s gadget bill

While Seoul cashes in on chip taxes, American shoppers help foot the bill. If Lee is right, that won’t change soon. A wide gap between demand and supply keeps memory scarce, and scarce memory keeps devices expensive.

Gartner’s February forecast has PC prices rising about 17% this year from 2025 levels as memory costs soar. It also expects new PCs under $500 to disappear by 2028. That matches what I found over the weekend on pricier memory cards and phones, where relief keeps slipping further out.

For investors, this is no longer a faraway story. SK Hynix began trading on Nasdaq on Friday, July 10, 2026. Its $26.5 billion share sale set a U.S. record for a foreign company, Quartz reported.

Anyone holding those shares now rides Korea’s swings. Even Korea’s top financial regulator said in June that he regretted letting leveraged funds on single-chip stocks launch, Fortune reported.

Oil exporters learned this lesson decades ago. A budget built on one commodity looks brilliant at the top of the cycle and painful at the bottom.

Officials in Seoul have floated a wealth fund for the windfall modeled on Norway’s oil fund, the Korea Herald reported. Whether Seoul saves or spends that money matters most in 2028. That is when its tax bills start reflecting whatever follows the shortage.

More Stocks News

• Lockheed is spending its own money before Washington signs off
• Apple and Alphabet are sending Wall Street different signals
• Microsoft wants to keep more of its AI billions from Nvidia
• Stock Market Today (Oct. 23, 2026): Dow tumbles following tech sell-off