- Nvidia is rapidly evolving into a major force in deep learning and virtual reality.
- Meanwhile, AMD lacks a proper deep learning footprint.
- Does Nvidia's early lead in deep learning threaten AMD's ongoing recovery?
Investors have traditionally recognized NVIDIA (NASDAQ:NVDA) as the market leader in GPUs, ahead of rivals AMD (NASDAQ:AMD) and a host of other smaller players. But the new-look Nvidia is much more than just a graphic cards manufacturer. Nvidia is quickly establishing itself as the new market leader in emerging fields such as artificial intelligence/deep learning as well as virtual reality.
And AMD is now at risk of being left behind.
A revamped Nvidia
If investors need fresh evidence that deep learning has become a big part of the new-look Nvidia, they don't have to look any further than the company's Q2 2016 earnings report. Nvidia reported revenue of $1.43B, up 9% sequentially and 24.3% Y/Y, managing to beat the consensus on Wall Street by about $80M. Meanwhile adjusted EPS of $0.53 was a 56% improvement compared to a year ago and 16 cents better than average Wall Street estimates.
Nvidia then guided for third-quarter revenue of $1.68B ±2%, much higher than Wall Street consensus of $1.45B. That revenue implies growth of 29% Y/Y at the midpoint, much faster than the single-digit growth the company has been posting over the past few years.
Heavily contributing to the healthy earnings beat was the sale of Tesla server accelerator cards as well as GRID GPUs for virtual desktops and cloud gaming, with sales rising an eye-popping 110% Y/Y. Nvidia said that demand for deep learning projects by tech titans including Microsoft (NASDAQ:MSFT), Facebook (NASDAQ:FB), and Baidu (NASDAQ:BIDU) remains strong.
Morgan Stanley's analyst Mitch Steves notes:
"We think the NVIDIA story has materially changed to a structural long driven by 1) deep learning, 2) automotive and 3) virtual reality growth. While expectations were originally slated for ~8-12% revenue growth, we think 20-30% is now on the table combined with margin tailwinds from improving yield on production. Data Center out-performance was the primary driver to a thesis changing moment (potential for q/q growth through CY16) given our previous stance for a single leg of growth (Virtual Reality). Finally, the Automotive segment continues to act as a call option on the stock if the Drive PX platform sees robust
growth. Net Net: we are moving to an Outperform rating and move our price target to $72."
So we are now looking at a revamped Nvidia whose new revenue streams have managed to re- ignite growth for the company. Nvidia stock is up a robust 88.9% YTD partly on the strength of the company's rapidly maturing new revenue streams and partly in sympathy with AMD's own blistering 165.5% YTD gain.
The big question on the mind of AMD investors is whether a revamped Nvidia now poses the biggest threat to AMD's own recovery.
Return to profitability the key to AMD's future growth
AMD risks being left behind in the new deep learning craze that promises to have a profound impact on how data will be utilized across diverse industries. Although AMD has played a big role in the development of OpenCL (an open framework for writing programs that execute across heterogeneous platforms such as CPUs, GPUs, FPGAs, and DSPs) it lacks a proprietary deep learning platform such as Nvidia's cuDNN. cuDNN is a CUDA-based library for deep neural learning that Nvidia launched in 2014.
And deep learning is slated to become the next big thing in AI and data center technology. Research firm has predicted that deep learning enterprise software spending will grow from just $109 M in 2015 to $10.4B in 2024.
About a week ago Intel (NASDAQ:INTC) acquired deep learning startup Nervana in a $400M deal. Last year the company bought FPGA-maker Altera, and can now conceivably use FPGAs in its future deep learning projects instead of standard GPUs. A research team from the University of California recently demoed middleware that might make deep learning on FPGAs feasible. The team demonstrated that using FPGAs in deep learning yielded considerably better operational efficiency than using GPUs.
AMD's near-invisibility in deep learning, however, might be a matter of necessity rather than choice. The company has for years been posting losses due to a dramatic slowdown in the PC market as well as intense competition by Nvidia in the mid and high-end GPU market.
But this is now set to change. During the last quarter, AMD posted a remarkable earnings trifecta by:
- Returning to top line growth after years of declines
- Posting a GAAP net profit for the first time in years
- Recording the first operating profit after a long string of losses
A return to profitability and positive free cash flow is absolutely necessary if AMD is to have enough cash for R&D, or even for the acquisition of deep learning companies a la Intel. Luckily, AMD appears to be firmly on the road to recovery, and its recent launch of Zen processor has been a resounding success with Oppenheimer noting that the chip's initial performance metrics have impressed:
"in a manner we haven't seen from the company in a decade."
That, coupled with the company's strong semi-custom chip sales, could be the key to a recovery.
Although deep learning is a very promising field, it's still relatively new and it's going to be several years before it goes completely mainstream. Nvidia's early lead certainly gives it an edge over the likes of AMD and Intel. AMD stock is, however, not likely to be hurt in the near and mid-term by the company's lack of a proper deep learning footprint. AMD's recovery has probably come just in time to allow the company to beef up its R&D muscle and invest in the field.