A lot had already been priced into Nvidia's (NVDA) shares going into the company's fiscal third-quarter report. But apparently not enough.
Look for analysts to dial their estimates sharply higher after the latest results, and perhaps also wonder just how big Nvidia's long-term addressable market is.
The GPU giant reported third-quarter revenue of $2 billion (up 54% annually) and EPS of 83 cents, trouncing consensus analyst estimates of $1.69 billion and 60 cents. It also guided for fourth-quarter revenue of $2.1 billion (plus or minus 2%), well above a $1.69 billion consensus and implying 50% annual growth at the midpoint.
Shares are up 14.6% after hours to $77.65 as of the time of this article, surging to new highs and now up 136% on the year. With a current market cap of $42 billion, only four U.S.-traded chipmakers -- Intel (INTC) , Qualcomm (QCOM) , Broadcom (BRCM) and Texas Instruments (TXN) -- are worth more than Nvidia. GPU rival AMD (AMD) is up 3.7% to $6.53.
Though Nvidia's report features several eye-popping sales growth figures, none arguably top the 193% increase in Nvidia's data center product sales to $240 million (12% of revenue). That's easily better than the second quarter's 110% growth and the first quarter's 63%.
While healthy demand for Nvidia's GRID server GPUs (used to power cloud gaming services and server-hosted virtual PCs) played a role, the largest driver was soaring sales of the company's Tesla server accelerator GPU cards. Nvidia also benefited from initial sales of its Tesla-powered DGX-1 deep learning system.
This year might just be seen as the year the tech sector got artificial intelligence religion, and perhaps no company is benefiting more from it than Nvidia.
With the help of new GPUs based on the company's cutting-edge Pascal architecture, adoption of Tesla cards to process algorithms for a very popular branch of A.I. known as deep learning has taken off.
Internet giants such as Facebook (FB) are extensively using Tesla cards for their deep learning work, and the likes of Amazon (AMZN) , Microsoft (MSFT) and Alibaba (BABA) provide third parties with access to Tesla-based cloud computing services for their own AI work.
But while the data center numbers will get more attention, one can't ignore the fact Nvidia's gaming GPU revenue rose 63% to $1.24 billion (62% of revenue). That's far better than the second quarter's 18% growth, comes in the face of a still-sluggish PC market and suggests this year's launch of Pascal-based PC GPUs such as the GTX 1080 and Titan X has been a smash hit.
Interest in 4K-resolution PC gaming has taken off, and with AMD's recently-launched Polaris GPU architecture more focused on the mid-range, Nvidia pretty much has the lucrative high-end market to itself for now. Looking further down the line, adoption of PC-connected virtual reality headsets such as the Oculus Rift could also be a growth driver.
Then there's automotive revenue, which grew 61% to $127 million. That's just slightly below the second quarter's 68%. An 87% increase in sales of Tegra app processors, which are designed into dozens of infotainment systems, drove the growth. Nvidia's Drive PX2 module for autonomous driving systems, which is being tested out by many automakers and is believed to power Tesla Motors' (TSLA) next-gen Autopilot system, could lift automotive sales in future quarters. And Tegra revenue should get a boost from a design win for Nintendo's Switch handheld console, which launches in March.
One weak point: Nvidia's OEM and IP revenue, which covers GPU sales to PC makers and $66 million in quarterly licensing revenue from Intel (soon to disappear), fell 4% to $186 million due to soft PC demand. But that decline pales relative to Nvidia's gaming GPU growth. "Professional visualization" revenue, which covers sales of Quadro workstation GPUs, rose 9% to $207 million.
Outside of its strong sales growth, Nvidia's bottom line benefited from the fact its gross margin rose 270 basis points annually to 59.2%, topping guidance of 58%. Fourth-quarter GM guidance is at 59.2%, plus or minus 2%. Adjusted operating expenses were a little higher than expected, but still only rose a moderate 11% to $478 million.
And in spite of the huge 2016 gains posted by its shares, Nvidia added $2 billion to its buyback authorization (total funds are now at $2.96 billion), and promises to spend $1.25 billion on buybacks and dividends in fiscal 2018, which ends in January. Oh, and the quarterly dividend has been hiked by 22% to 14 cents (0.7% forward yield).
As good as things look for Nvidia right now, it's worth remembering -- particularly with shares now at such lofty levels -- that competition is slowly intensifying in the company's two biggest growth markets.
Intel (INTC) is trying hard to wrest share from Nvidia's Tesla line via its Xeon Phi accelerator cards, and while it faces an uphill battle when it comes to AI workloads that are well-suited for Nvidia's GPUs, it could have more luck when it comes to broader high-performance computing (HPC) workloads.
In addition, Intel recently bought Nervana Systems, a developer of ASICs meant specifically for deep learning. While Nervana's ASICs aren't as flexible as Nvidia's GPUs, they could be quite effective for handling specific algorithms.
Alphabet's (GOOGL) Google, meanwhile, is using an internally-developed ASIC known as the Tensor Processing Unit for much of its AI work. And Microsoft and Baidu (BIDU) have deployed both GPUs and programmable chips (FPGAs) for deep learning work. The latter can't deliver the raw performance of GPUs, but can be re-programmed on the fly to handle new algorithms.
In the PC market, AMD's Polaris GPUs just scored design wins for Apple's new 15-inch MacBook Pros. And the high-end PC market could get more interesting in the first half of 2017, when AMD is set to launch its Vega GPU line.
But for now, Nvidia controls the commanding heights of the PC GPU and AI algorithm-processing markets. And as these markets swell in size, those heights are proving to be very valuable real estate.