The Tesla P100 data center accelerator. Image source: NVIDIA.
Mark Hibben, writing for Seeking Alpha, recently gave his take on graphics specialist NVIDIA's (NVDA +3.22%) freshly announced next-generation Tegra processor, codenamed Parker. Parker is primarily targeted at automotive applications, both powering in-vehicle infotainment systems and serving as the main processor in NVIDIA's Drive PX2 self-driving car platform.
Interestingly, Hibben brings up the possibility of NVIDIA building a higher-performance system-on-a-chip, similar to Parker, that integrates NVIDIA's custom-designed Project Denver processor cores along with the company's graphics processors onto a single chip to do duty as a server processor. As interesting as that sounds, here's why that's simply not going to happen.
Straight from the CEO
On NVIDIA's most recent earnings call, an analyst actually asked CEO Jen-Hsun Huang about this very possibility. Here was the question: "Is there a deep learning integrated CPU/GPU play that might open up more [total addressable market] long term for your company that you guys are considering pursuing?"
Huang made it clear he thinks that, over time, data center workloads will increasingly benefit from the strengths of the company's graphics processing units. However, he did concede that a lot of workloads are still CPU-centric and that to run those workloads, one will "still want to have an extraordinary CPU."
He then went on to praise chip giant Intel (INTC +3.01%), stating that he didn't think "anybody would argue that Intel makes the world's best CPUs."
"It's not even close, there's not even a close second," Huang continued.
The point that Huang was clearly trying to make here is that it just wouldn't make sense for NVIDIA to try to beat Intel at its own game. NVIDIA's play in the data center is to build high-performance graphics processors to accelerate workloads that can see large speed-ups from using a graphics processor relative to a typical CPU.
NVIDIA's strength is in its ability to stay focused
NVIDIA's strength, particularly over the last several years, has been its ability to stay focused and make sound, targeted investments in areas in which it can deliver value and ultimately be a market leader.
This doesn't mean that the company doesn't try to expand its total addressable market. NVIDIA tried to compete in the market for mobile processors, but when it was clear that this wasn't a market in which it could participate and deliver value for its shareholders, it exited.
Indeed, NVIDIA's original pitch when it first announced its Project Denver processor core, all the way back in 2011, was that this core could help it break into the market for processors targeting "PCs, data center servers, and supercomputers."
However, it's clear that NVIDIA no longer plans to target those markets.
NVIDIA's strategy of being a relatively focused chipmaker means that it can't quite boast the extremely large total addressable markets that others companies can. However, as long as NVIDIA can maintain strong competitive positions in these markets, and as long as those markets are -- in aggregate -- reasonably sized and growing, this strategy of staying focused could ultimately pay off for the company and its stockholders over the long term.
