OpenAI Commits to 6 Gigawatts of AMD GPUs in Multi-Year Deal

OpenAI has signed a definitive agreement to deploy 6 gigawatts of AMD GPUs, the two companies announced on October 6, 2025, in one of the largest compute commitments ever disclosed between a chipmaker and a single AI developer. The first 1 gigawatt deployment of AMD Instinct MI450 Series GPUs is scheduled to begin in the second half of 2026.

The deal makes AMD a core strategic compute partner for OpenAI's next-generation AI infrastructure. It spans multiple years and multiple generations of AMD Instinct GPUs, starting with the MI450 series and rack-scale AI solutions, and extending to future generations of the hardware line. AMD and OpenAI will also share technical expertise to align their product roadmaps, formalizing a hardware and software collaboration that began with the MI300X and continued with the MI350X series.

The scale of the commitment matters. Six gigawatts of accelerated computing represents a buildout measured in entire power plants, not server rooms, and it signals that OpenAI intends to diversify its compute supplier base beyond a single vendor as it races to expand training and inference capacity. For AMD, the agreement is a commercial breakthrough: OpenAI, one of the largest buyers of AI compute in the world, is now contractually tied to its Instinct roadmap through multiple product generations.

The financial structure: a warrant for 160 million shares

Alongside the compute agreement, AMD has issued OpenAI a warrant for up to 160 million shares of AMD common stock, structured to vest as specific milestones are achieved. The first tranche vests with the initial 1 gigawatt deployment. Additional tranches vest as OpenAI's purchases scale up to the full 6 gigawatts.

The vesting terms carry double conditions. Beyond deployment milestones, vesting is tied to AMD achieving certain share-price targets and to OpenAI reaching the technical and commercial milestones required to deploy AMD hardware at scale. That structure aligns both companies' incentives: AMD wins only if its stock performs and its silicon meets OpenAI's requirements in production; OpenAI gains equity upside as it commits more capital to AMD's platform.

What the executives said

AMD CEO Dr. Lisa Su framed the deal in ecosystem terms. "We are thrilled to partner with OpenAI to deliver AI compute at massive scale," said Su, who serves as chair and chief executive of AMD. "This partnership brings the best of AMD and OpenAI together to create a true win-win enabling the world's most ambitious AI buildout and advancing the entire AI ecosystem."

Sam Altman, co-founder and CEO of OpenAI, cast the agreement as a capacity play. "This partnership is a major step in building the compute capacity needed to realize AI's full potential," Altman said. "AMD's leadership in high-performance chips will enable us to accelerate progress and bring the benefits of advanced AI to everyone faster."

Greg Brockman, co-founder and President of OpenAI, emphasized the stack-wide nature of the collaboration. "Building the future of AI requires deep collaboration across every layer of the stack," Brockman said. "Working alongside AMD will allow us to scale to deliver AI tools that benefit people everywhere."

AMD's finance chief put a number on the commercial stakes. "Our partnership with OpenAI is expected to deliver tens of billions of dollars in revenue for AMD while accelerating OpenAI's AI infrastructure buildout," said Jean Hu, EVP, CFO and treasurer of AMD. Hu added that the agreement "creates significant strategic alignment and shareholder value for both AMD and OpenAI and is expected to be highly accretive to AMD's non-GAAP earnings-per-share."

Why this matters

The compute layer of AI has become the industry's central bottleneck. Frontier-model training and large-scale inference demand power in the gigawatt range, and developers with credible compute pipelines hold a structural advantage over those without them. OpenAI's decision to anchor 6 gigawatts of its buildout on AMD Instinct hardware — with the MI450 as the first deployment in the second half of 2026 — confirms that AMD's accelerator line has moved from challenger status to a strategic dependency for one of the largest AI labs in the world.

The roadmap collaboration also carries research weight. The partnership between AMD and OpenAI began with the MI300X and continued with the MI350X series before this agreement extended it to the MI450 and beyond. That multi-generation arc means OpenAI engineers are now co-optimizing silicon and software across AMD's roadmap, deepening a hardware-software feedback loop that both companies argue advances the entire AI ecosystem, not just their own products.

For AMD shareholders, the warrant structure converts the partnership into a measurable set of triggers. The first tranche of the 160 million shares vests with the initial 1 gigawatt deployment, and subsequent tranches unlock as purchases scale toward 6 gigawatts — provided AMD's share price hits its targets and OpenAI clears the technical and commercial milestones required for at-scale deployments. Hu's guidance that the deal will be "highly accretive" to non-GAAP earnings-per-share signals AMD expects the volumes, not just the headline, to reach its income statement.

The announcement also reframes the competitive map of AI infrastructure. A single buyer committing to 6 gigawatts across multiple GPU generations, with equity incentives layered on top, is a new contractual template for how hyperscale AI compute gets financed and secured. The first proof point arrives in the second half of 2026, when the initial 1 gigawatt of Instinct MI450 GPUs goes into OpenAI's infrastructure — and when the first tranche of OpenAI's AMD warrant vests.