When a stock goes into an earnings report trading at 46 times its expected earnings for this fiscal year, markets are prone to demand perfection. It's not enough to beat sales/EPS estimates and issue strong guidance; every major business unit probably needs to meet or beat expectations, and there can't be any financial line item that gives investors pause.
That's the bind that Nvidia Corp. (NVDA) finds itself in post-earnings. There's nothing wrong with the GPU giant's latest numbers at first glance, and while some minor issues -- some related to production transitions -- pop up at second glance, there's isn't anything to suggest the company's core franchises are facing major challenges. But in the absence of perfection, a selloff has arrived.
With cryptocurrency miners providing a healthy sales boost, Nvidia reported July quarter (fiscal second quarter) revenue of $2.23 billion (up 56% annually) and adjusted EPS of $1.01 (up 91%), topping consensus analyst estimates of $1.96 billion and $0.81. It also guided for October quarter revenue of $2.35 billion, plus or minus 2%, above a $2.14 billion consensus.
But with shares up around 70% since mid-April going into earnings, Nvidia is down 7% to $153.33 as of the time of this article. There appear to be two or three culprits.
The largest of these: Nvidia's Datacenter revenue, which covers products featuring its Tesla server GPUs, came in at $416 million, below a $423 million consensus. Though the segment's sales still rose 175% annually (close to the April quarter's 186% growth), they only grew 2% sequentially.
On the earnings call, management hinted that some Tesla GPU clients delayed making purchases ahead of the arrival of Nvidia's Tesla V100, its next-gen flagship server product and the first GPU based on the company's new Volta architecture. Though announced on May 10th to much fanfare, customers didn't start receiving V100s until late July. In the interim, some of them appear to have held off on buying cards or systems featuring Nvidia's prior flagship GPU, the Tesla P100, which is based on the Pascal architecture launched last year.
More of What's Trending on TheStreet:
But as Nvidia's guidance appears to signal, the V100 ramp could make Nvidia's Datacenter growth slowdown very temporary. CEO Jen-Hsun Huang asserts the V100 is now in "high-volume production," and is getting a strong reception from both server OEMs and cloud service providers.
With the V100 providing about a 40% gain relative to the P100 in raw performance, and (thanks to specialized cores and other features) often much larger gains in the speed at which neural networks can be trained, cloud giants and enterprises that have gotten AI religion are expected to be major buyers. In time, Intel Corp. (INTC) and others could provide stiffer competition in this space, but for now, Nvidia, aided by both its hardware specs and developer ecosystem, remains quite dominant.
Nvidia also claims Tesla momentum remains strong within the high-performance computing (HPC) and virtual PC markets, and that it's seeing strong traction for its AI-focused DGX-1 server, which launched last year and contains 8 high-end Tesla GPUs. Huang disclosed the DGX-1 has been shipped to over 300 clients (another 1,000-plus are "in the pipeline"), and noted Facebook Inc. (FB) has built a supercomputers containing 128 DGX-1 systems. A new version of the DGX-1 that replaces the 8 P100 GPUs found in the original model with 8 V100 GPUs is due in Q3.
Secondly, Nvidia's Automotive division, which supplies Tegra app processors for infotainment systems and Drive PX boards for autonomous driving systems, also fell slightly short of expectations, with revenue of $142 million (up 19% annually) missing a $146 million consensus. But this segment is still expected to grow sharply as Drive PX partnerships with the likes of Audi, Mercedes-Benz, Toyota, Volkswagen and Volvo bear fruit in the coming years. For now, production volumes for Drive PX hardware depend heavily on Tesla Inc.'s (TSLA) use of the platform in its second-gen Autopilot system.
The third concern: Though Nvidia beat consensus revenue estimates by $270 million, its adjusted gross margin only hit a guidance range midpoint of 58.6%. In addition, the midpoint of its October quarter GM guidance is only at 58.8%.
This is a fairly minor concern, though, as GM still rose by 50 basis points annually in spite of the loss of $66 million in quarterly licensing revenue from Intel that carried a 100% gross margin. Huang suggested initial costs related to ramping the V100/Volta limited margin growth, while adding that he expects pricing for Volta server GPUs to be "quite favorable" relative to Pascal products.
Meanwhile, Nvidia's mainstay gaming product business continues firing on all cylinders: Revenue rose 52% annually to $1.19 billion, easily beating a $1.05 billion consensus. As is the case for AMD Inc. (AMD) , Ethereum mining activity is boosting Nvidia's PC GPU sales. But the company is also benefiting from its continued dominance of a high-end gaming GPU market that's growing nicely with the help of strong interest in 4K gaming and e-sports, and from Tegra sales to Nintendo for its popular Switch handheld console. The Switch led total Tegra revenue to rise 101%.
Going forward, AMD's launch of the first PC GPUs based on its new Vega architecture (they ship on Monday) will spell slightly tougher high-end competition. But Nvidia's high-end offerings still have a big power consumption edge on AMD's, and the most powerful Vega product (the RX Vega 64) can't match the performance of Nvidia's two most powerful gaming GPUs (the GeForce GTX 1080 Ti and Titan Xp). All of this helps explain why Nvidia, judging by Huang's remarks, appears comfortable holding off on launching Volta PC GPUs until early 2018.
Miner demand also propped up sales of Nvidia GPUs that aren't classified as "gaming" products: In spite of the lost Intel payments, Nvidia's "OEM and IP" reporting segment saw revenue grow 54% annually to $251 million, well above a $106 million consensus. And the company's professional visualization business, which features its Quadro workstation GPUs, saw revenue grow 10% to $235 million, above a $215 million consensus.
As Nvidia sells off following its July quarter report, it's worth remembering that EPS estimates have risen meaningfully in response to the report. The fiscal 2018 (ends in January 2018) adjusted EPS consensus has risen to $3.92 from $3.56, and the fiscal 2019 consensus has risen to $4.44 from $4.17.
The fiscal 2019 hike, of course, still leaves Nvidia trading 35 times what it's expected to earn during its next fiscal year. And while further estimate hikes are likely if the company keeps executing the way it has been lately, such multiples still guarantee a lot of volatility when even minor speed bumps are encountered.