Nvidia Isn't Enough — OpenAI Turns to AMD for GPUs, Too

Don’t miss out on our latest stories. Add PCMag as a preferred source on Google.

OpenAI needs so many GPUs that it’s now turning to Nvidia’s biggest rival, AMD, for more chips.

On Monday, OpenAI announced a deal to source up to 6 gigawatts of computing power from AMD. Financial terms, including how much OpenAI will spend, were not disclosed. However, the agreement involves OpenAI purchasing up to 160 million shares of AMD stock at $0.01 per share — or a 10% stake — contingent upon specific milestones being achieved.

In exchange, OpenAI will deploy 1 gigawatt worth of AMD’s Instinct MI450 enterprise GPUs starting in the second half of 2026. AMD also told The Wall Street Journal it expects each gigawatt of compute to cost tens of billions of dollars. 

This Tweet is currently unavailable. It might be loading or has been removed.

The deal is surprising since OpenAI has long touted its partnership with Nvidia — the leading provider of AI GPUs. Two weeks ago, OpenAI also announced a similar agreement with Nvidia for at least 10 gigawatts of AI data center capacity. But in that case, Nvidia is going to invest up to $100 billion in OpenAI. In return, the San Francisco lab is adopting Nvidia’s next-generation GPUs, including the Vera Rubin architecture. 

OpenAI potentially owning a 10% stake in AMD might also raise concerns about whether the deal poses a threat to its partnership with Nvidia. However, CEO Sam Altman said in a tweet: “This is all incremental to our work with Nvidia (and we plan to increase our Nvidia purchasing over time). The world needs much more compute…”

This Tweet is currently unavailable. It might be loading or has been removed.

IT industry analyst Patrick Moorhead added that the deal is about OpenAI incentivizing AMD to “scale” its AI GPUs to the point they can be deployed widely across data centers. 

Altman literally needs millions of additional GPUs for his plan to power next-generation AI, which requires building numerous power-hungry data centers. OpenAI’s CEO even envisions the tech industry possessing enough capacity to “produce a gigawatt of new AI infrastructure every week.” It’s a tall task considering OpenAI is already building a data center about 60 football fields in size in Texas that’ll offer only about 1.2 gigawatts of compute. In contrast, the Hoover Dam generates about 2 gigawatts of electricity. 

Last month, OpenAI also signed a $300 billion contract to buy computing power from Oracle. The eye-popping figures are naturally causing analysts and investors to wonder if OpenAI can actually come up with the cash, and if the investments will pay off. On social media, some users accuse the companies of propping up their stocks and valuations by announcing such deals when the funding remains unclear. According to The Information, OpenAI generated $4.3 billion in revenue during the first half of this year, but incurred $2.5 billion in costs.