With its deal to buy top automotive vision processor supplier Mobileye (MBLY) , Intel (INTC) is signaling that cars are just as important to its long-term growth ambitions as the data center, as it looks for ways to cut its dependence on the slowing PC market. The chip giant is also signaling that its strategies for attacking each end market and fending off several hungry rivals will have much in common.
On Monday morning, Intel gave the chip industry a jolt by announcing that it's buying Mobileye for $63.54 per share, or $14.7 billion after factoring in Mobileye's net cash. The price represents a 34% premium to Mobileye's Friday close.
The deal certainly isn't cheap: Intel is paying 20 times Mobileye's 2018 consensus sales estimates of $732.1 million, and 40 times a consensus EPS estimate of $1.58. Sticker shock is doubtlessly the main reason Intel's shares fell 2.1% on Monday; Shares fell another 0.5% to $34.00 on Tuesday morning.
Making the deal more palatable: Since Mobileye is an Israeli company, Intel can mostly finance the acquisition with an offshore cash balance that stood at $13.6 billion at the end of 2016. Overall, Intel had $17.1 billion in cash and (thanks to its $16.7 billion 2016 purchase of programmable chipmaker Altera) $25.3 billion in debt.
Intel, which has already partnered with Mobileye on autonomous driving deals with BMW and major auto parts supplier Delphi (DLPH) , repeatedly stresses the product synergies it envisions from pairing Mobileye's vision processors and related algorithms with Intel's various automotive offerings. "This acquisition essentially merges the intelligent eyes of the autonomous car with the intelligent brain that actually drives the car," said Intel CEO Brian Krzanich in an employee e-mail.
As highlighted by the BMW and Delphi deals, as well as the launch of its Go in-car computing platform, Intel wants to put one of its powerful (and often costly) Xeon server CPUs inside every car supporting a high level of autonomous driving; it argues the horsepower provided by the chips is needed to analyze the estimated 4TB of daily data each self-driving car will put out. Intel is also hoping Xeon CPUs, as well as complementary solutions such as Xeon Phi co-processors that will leverage technology from recently-acquired deep learning chipmaker Nervana Systems, will be used in data centers to analyze driving data and train the algorithms used by self-driving cars to navigate the real world.
Of course, Intel's auto chip offerings go well beyond processors. The company is also selling automakers on its modems (4G today, 5G eventually) and memory chips (both NAND flash and Intel's 3D XPoint next-gen memory), as well as programmable chips (FPGAs) that could conceivably be re-programmed on the fly to handle new autonomous driving algorithms. There are parallels here with how Intel has been rolling out a number of data center offerings meant to complement Xeon CPUs, such as Xeon Phi, 3D XPoint, its Omni-Path interconnect fabric chipsets and high-speed silicon photonics transceivers.
In Mobileye, Intel gets a company with a long list of top-tier automotive clients: Automakers that have used Mobileye's EyeQ processors to handle driver-assistance (ADAS) features such as collision-avoidance and lane departure warnings include Ford (F) , GM (GM) , Nissan, Honda Motors (HMC) and Volvo. It also gets a company with over a decade of experience in developing algorithms that that make sense of what a car's cameras see, and of optimizing them in response to real-world data.
Mobileye also gives Intel a mapping technology platform (known as REM) for its autonomous driving offerings that has the support of BMW, Volkswagen and top auto maps supplier HERE. During a talk at CES, Mobileye investor relations chief Dan Galves suggested his company views itself as a software firm at heart.
While the availability of Level 5 autonomous driving (no human driving activity is ever needed) is at least several years away, the rollout of Level 3 (cars are self-driving in certain environments) and Level 4 (self-driving in most environments) driving solutions in the interim will still require supporting cars to have a lot more chip content than the average car does today.
Mobileye estimates it would obtain $400 to $500 from each Level 4 car using all of its autonomous solutions, or about 10 times what it gets for a single ADAS solution. Delphi estimates the autonomous driving system it's developing with Mobileye and Intel will initially cost about $5,000, but adds prices will drop as adoption grows. Adding to the bill of materials: Since human lives are on the line, automakers prefer to err on the side of caution when it comes to how many sensors and how much processing power autonomous systems contain.
But deep-pocketed rivals are also going after this opportunity with offerings that are each differentiated in their own way. Nvidia's (NVDA) Drive PX2 self-driving platform, which contains two Tegra app processors and two standalone GPUs to analyze data from cameras and sensors, has won the support of Tesla Motors (TSLA) , Volvo, Mercedes-Benz and Audi, as well as part suppliers Bosch and ZF. Nvidia argues the ability of its GPUs to efficiently process deep learning algorithms makes them well-suited to power autonomous driving systems. A successor to Drive PX2 (codenamed Xavier) is in the works.
NXP Semiconductor, a Mobileye rival and the world's biggest auto chip supplier, has launched a self-driving system called Bluebox that pairs a "central computing engine" with complementary chips. With NXP about to be acquired by Qualcomm (QCOM) , it could be just a matter of time before an offering that combines Bluebox with Qualcomm's processors, modems and connectivity chips is rolled out.
Then there's Alphabet's (GOOGL) Waymo self-driving unit, which has collected over two million miles of self-driving data and is now engaging automakers such as Fiat Chrysler and Honda about forming alliances. Whereas Mobileye's self-driving solution relies on cameras, Waymo's makes heavy use of LIDAR sensors. Waymo started off using LIDAR sensors provided by startup Velodyne, which has Ford and Baidu's backing, but has since come up with a home-grown solution that it claims lowers costs by 90%.
While Intel and Mobileye make a good case that there's value in providing an end-to-end autonomous driving platform, it's far from clear that automakers used to working with numerous chipmakers need to go this route. For example, one could choose to combine Intel Xeon CPUs with Waymo's sensor arrays and algorithms, or use Nvidia's Xavier with Intel's memory chips and modems. Audi's new zFAS self-driving computer features both Nvidia and Mobileye parts to process sensor data.
Moreover, the fact that several big-name Mobileye clients are adopting or at least thinking of using Nvidia and Waymo's self-driving solutions shows what a highly competitive environment this is. Intel just greatly expanded its autonomous driving footprint, but still has much to prove about how much its costly purchase gives it a leg up on rivals.
Editors' pick: Originally published March 13.