In This Article:
Wednesday, April 14, 2021
This article was first featured in Yahoo Finance Tech, a weekly newsletter highlighting our original content on the industry. Get it sent directly to your inbox every Wednesday by 4 p.m. ET. Subscribe
Nvidia's first data-center CPU is a huge hedge against losing its ARM bid
Nvidia (NVDA) is laying the groundwork for a future where the chipmaker could be one of the most dominant players in the competitive data-center market, putting Intel (INTC) and AMD (AMD) directly in its crosshairs.
Nvidia’s latest move comes in the form of its first-ever central processing unit, or CPU, built specifically for data-centers, called Grace. While Grace could help Nvidia compete with AMD and Intel, the new CPU could also soften the blow if Nvidia’s planned $40 billion acquisition of chip designer ARM collapses.
“The disclosure of Grace has a number of dimensions to it at a high level...because the ARM deal may or may not get done. It's a plan B,” said Hans Mosesmann, managing director at Rosenblatt Securities.
Nvidia is known around the world for its best-in-class graphics processing units that power data centers focused on things like artificial intelligence (AI) and machine learning (ML).
It’s never had a data-center CPU of its own, though. That’s problematic because Nvidia has to rely on Intel's or AMD’s CPUs when offering its own data-center GPU systems. With Grace, Nvidia can ditch Intel and AMD, at least at the highest of high-end data-center systems.
An ARM acquisition would also set up Nvidia to build out its own data-center systems without needing Intel or AMD. And in the event it doesn’t happen, Grace is now there for Nvidia to show the company’s customer base that it has another path to not relying on Intel or AMD.
“By having its own CPU combined with the GPU is a pretty powerful story and, I think, it's certainly indicative of the future,” Glenn O’Donnell, vice president & research director at Forrester Research, told Yahoo Finance.
Nvidia’s Grace and ARM moves
Nvidia’s Grace, due out in 2023, is meant for use in the types of heavy-duty data-center applications that major laboratories work on. In fact, both the U.S. Department of Energy’s Los Alamos National Laboratory and the Swiss National Supercomputing Centre (CSCS) plan to build the first two Grace-based supercomputers.
According to Nvidia, when combined with the company’s GPUs, Grace offers 10x faster performance in systems that use Nvidia GPUs and rival CPUs. Nvidia says that Grace systems will likely continue to serve as niche data-center installations. However, as the need to process ever greater amounts of data to train artificial intelligence (AI) and machine learning (ML) systems increases, it’s easy to see how Grace could fit into Nvidia’s future plans.