Advanced Micro Devices Inc.’s biggest rival in graphics processing units is teaming up with its biggest rival in central processing units. Does that spell trouble for the chip company?
At first, Wall Street seemed to think so. AMD shares fell as much as 5.9% earlier Thursday after Nvidia Corp. and Intel Corp. announced a collaboration to design and manufacture chips. But by the end of the session, AMD’s stock had largely clawed back to finish the day down only 0.8%.
AMD gets the most buzz for being second to Nvidia in the red-hot market for GPUs that are powering the artificial-intelligence revolution. The company also competes with Intel in the sizable market for central processing units that go into laptops and servers.
“Having two major competitors combining their efforts is not exactly a positive outcome” for AMD, Jack Gold, principal analyst at J.Gold Associates, said in a note to clients on Thursday. Not only does AMD face the prospect of a stronger Intel, but it separately risks losing market share to chip startups down the road, he added.
The partnership between Nvidia and Intel “means that there is going to be optimization between CPU and GPU environments more so than in the past,” Gold told MarketWatch. He explained that enterprise customers that are already using Intel’s products will likely look to Nvidia rather than AMD.
But Intel still faces challenges, and AMD has been making progress in its own right.
Gold said that AMD should still have a significant advantage in the AI space “because Nvidia is delivery-limited,” meaning it can’t supply as many chips as the market needs.
“If someone needs something and you can’t get it from one vendor, you’re going to go try and get it from someone else,” Gold said, pointing to AMD.
AMD’s chips have also shown “solid performance” at a lower average price point than Nvidia’s, “making them a popular alternative to Nvidia’s high prices,” Brian Mulberry, a senior portfolio manager at Zacks Investment Management, told MarketWatch.
And even though Intel is benefitting from its partnership with Nvidia, “it does not automatically fix the fact that they have had significant delays in the new manufacturing plants being built,” Mulberry said in emailed comments.
Over the long term, if capital-expenditure levels stay consistent, Mulberry sees “plenty of business for everyone looking to build out AI infrastructure.”
At the same time, the chip market is expanding with the rise of AI inferencing and the shift to edge computing, where Gold said Nvidia may not have as much of an advantage.
Edge devices, which process and store data locally rather than via the cloud, still need AI accelerators but mostly run on CPUs, Gold said. Nvidia’s relative weakness with CPUs, in Gold’s view, is one of the reasons for its collaboration with Intel.
Nvidia said in the announcement on Thursday that it and Intel are focused “on seamlessly connecting Nvidia and Intel architectures” using its interconnect technology called NVLink, which enables scaling to handle intensive AI and high-performance computing workloads. NVLink allows for multiple GPUs to connect to each other or to CPUs for faster and efficient communication between chips in data centers.
Intel will develop custom CPUs based on its x86 computer chip architecture for data centers, which Nvidia said will be integrated with its AI infrastructure. The chip pioneer also plans to build x86 system-on-chips that can work with Nvidia’s RTX GPU chiplets in personal computers.
The collaboration was good news for Intel and its investors following an agreement in August to sell a 10% stake in the company to the U.S. government. Intel shares finished Thursday up 22.7%.
At a press conference on Thursday, Nvidia Chief Executive Jensen Huang said the Trump administration was not involved with its partnership with Intel.
Comments