OpenAI and AMD have inked a multibillion-dollar supply and equity agreement poised to reshape how large AI models are built and powered. Under the deal, OpenAI will deploy up to 6 gigawatts of AMD’s Instinct GPU systems over the coming years, with the first 1 gigawatt tranche arriving in the second half of 2026 using the forthcoming MI450 series, says Tom’s Hardware.
The structure of the agreement is as significant as the scale. AMD has granted OpenAI warrants to acquire up to 160 million shares, roughly 10% of AMD’s outstanding equity, at a nominal strike price of one cent per share. These warrants vest in stages tied to deployment milestones and share-price thresholds as more GPU capacity is installed.
For AMD, the deal could usher in tens of billions in new revenue and strengthen its position in the AI compute landscape long dominated by Nvidia. OpenAI, meanwhile, gains a strategic partner in hardware, reducing its reliance on a single supplier chain and securing long-term access to scale.
That said, the partnership is additive, not exclusive. OpenAI continues existing relationships with Nvidia and expects its AI compute strategy to remain diversified. The real test will lie in execution: coordinating deployment, optimizing software and hardware co-design, and meeting the performance and financial milestones built into the stock warrants.
In essence, the AMD-OpenAI pact underscores a moment when AI’s growth demands are colliding with semiconductor economics. It’s not just about buying chips; it’s about forging long-term alignment between AI innovators and compute providers in the race to higher scale and performance.