AMD and OpenAI announce strategic partnership to deploy 6 gigawatts of AMD GPUs (openai.com)

AMD and OpenAI announced a multi-year, multi-generation strategic partnership in which OpenAI will deploy up to 6 gigawatts of AMD GPU capacity to power its next-generation AI infrastructure. The initial phase is a 1 GW deployment of AMD Instinct MI450 series GPUs slated to begin in 2H 2026, with collaboration extending from rack-scale AI systems to future Instinct generations. The companies will deepen hardware/software co-optimization—building on prior work with the MI300X and MI350X—so AMD chips, system designs and software stacks are tuned to OpenAI’s training and inference workloads. The deal is significant because it injects massive, diversified compute capacity into the AI ecosystem and intensifies AMD’s challenge to dominant GPU suppliers. OpenAI’s commitment (and AMD’s expectation of “tens of billions” in revenue) includes a warrant for up to 160 million AMD shares that vests with deployment and performance milestones, aligning commercial and technical incentives. For practitioners, the partnership promises tighter integration across silicon, racks and software—faster scaling of model training and inference, potential new hardware features optimized for large models, and meaningful supply-chain and data-center power/cooling implications as AI compute grows to gigawatt scale.