OpenAI just broadened its hardware playbook. In a five-year deal, the AI leader will use AMD processors across its data-center footprint — a move that directly challenges Nvidia’s long run as the dominant supplier of AI chips. The partnership includes a staged deployment that starts with AMD’s Instinct MI450 GPUs and could scale up to six gigawatts of compute, shifting the balance in a market that’s racing to keep up with exploding demand for model training and inference.
Big numbers, early rollout
The agreement commits AMD to supplying up to six gigawatts of processors over five years, with the first gigawatt of Instinct MI450 GPUs expected in the second half of 2026. AMD says the deal could translate into “tens of billions of dollars in revenue,” though it didn’t disclose exact terms. Investors reacted quickly: AMD’s shares jumped roughly 24% in pre-market trading after the announcement.
Not a one-vendor strategy anymore
This comes after OpenAI announced a strategic relationship with Nvidia last month — a separate plan that would allow OpenAI to deploy at least 10 gigawatts of Nvidia GPUs and could include an investment of up to $100 billion from Nvidia. In the new messaging, Nvidia is described as a “preferred strategic compute and networking partner,” while AMD is called a “core strategic compute partner.”
Part of the AMD deal gives OpenAI an option to receive up to 160 million AMD shares — about a 10% stake — at a nominal price as certain milestones are met. That kind of equity optionality ties the two companies together beyond a typical buyer-seller relationship.
Why OpenAI is diversifying
OpenAI’s models—particularly large language models—require huge, reliable compute pools. Depending on a single supplier creates supply-chain and price risks, especially as data centers scale. By signing multiple hardware agreements, OpenAI buys flexibility and negotiating leverage, which helps keep costs, availability and innovation moving in step with its needs.
AMD’s chips will compete directly with Nvidia’s H100 and next-generation accelerators; AMD pitches comparable performance and improved cost or energy tradeoffs in some workloads, which can matter a lot at hyperscale.
A bigger ecosystem moment
These deals follow a shift earlier this year when OpenAI moved away from an exclusive compute arrangement with Microsoft. That exclusivity was relaxed in January, and recent steps — including a non-binding memorandum of understanding with Microsoft in September — indicate OpenAI is reshaping how it partners across cloud and hardware vendors as it scales and explores corporate options like an eventual IPO.
What leaders are saying
OpenAI CEO Sam Altman framed the deal as essential to building the compute capacity needed for AI’s next chapter, calling AMD’s high-performance chips an accelerator for progress. AMD CEO Lisa Su described the partnership as a way to deliver AI compute “at massive scale,” saying it’s a win for both companies and the broader AI ecosystem.
Why it matters
This partnership is the most concrete sign yet that Nvidia’s effective monopoly on AI compute could loosen. For OpenAI, it’s a practical hedge and a way to speed up expansion. For AMD, it’s a major validation of years of investment in AI-focused silicon. For the industry, it signals more competition, and potentially better pricing and innovation as providers battle for hyperscale customers.
Takeaway: The AI hardware market is entering a new phase where multi-vendor strategies and equity-linked partnerships are becoming the norm. That will affect pricing, supply resilience, and how quickly new generations of chips get adopted.
Do you think AMD can meaningfully chip away at Nvidia’s dominance, or will Nvidia remain the clear leader? Share your view in the comments.
