AMD and OpenAI have signed a multi-year agreement to deploy 6 gigawatts of AMD graphics processing units (GPUs) for OpenAI’s AI infrastructure.
The plan envisions staged deployments over several years, beginning with an initial one-gigawatt phase powered by AMD Instinct MI450 chips, set to commence in the second half of 2026. The deal covers multiple generations of AMD Instinct GPUs and continues a technical collaboration already underway with the MI300X and MI350X series.
The six gigawatts of compute power roughly equate to the energy needs of about 5 million American households, or about three times the electricity produced by the Hoover Dam.
OpenAI will work with AMD as its primary strategic partner for large-scale deployments. The two companies will share technical know-how to optimize their respective roadmaps. OpenAI has already contributed to the design of earlier AMD chip generations, notably the MI300X.
OpenAI could acquire 10% of AMD
AMD has issued a warrant to OpenAI giving it the option to acquire up to 160 million ordinary shares. This option, representing roughly 10% of the equity, is structured as a staged exercise.
The first tranche will be exercised during the initial deployment of one gigawatt. The subsequent tranches will depend on the expansion of deployments up to 6 gigawatts, as well as the achievement of AMD stock price targets and the attainment of technical and commercial milestones for OpenAI. One of the later tranches is tied to an AMD stock price of $600 per share (the stock closed at $164.67 on October 4, 2025).
According to AMD CEO Lisa Su, quoted by Bloomberg, this structure aims to align the interests of both parties: “We wanted to ensure that OpenAI is motivated for AMD to succeed. The more OpenAI deploys, the more revenue we generate, and they share in the gains.”
As the second player in the AI accelerator chip market, well behind Nvidia, AMD expects the deal to generate tens of billions of dollars in annual revenue. Company leaders anticipate that revenue recognition will begin in 2026, with it speeding up in 2027. AMD’s AI GPU revenue is projected at $6.55 billion for 2025, compared with $115 billion for Nvidia’s datacenter division in its most recent fiscal year.
