AMD and OpenAI have unveiled a far-reaching partnership to deploy up to 6 gigawatts of AI compute built on AMD’s Instinct GPUs, signaling one of the most ambitious AI infrastructure buildouts to date. The rollout begins in the second half of 2026 with a 1-gigawatt deployment of next‑generation Instinct MI450 GPUs, followed by multi‑generational expansions across the Instinct lineup.
Key details at a glance:
– Scale: Up to 6 gigawatts of AI compute capacity
– First phase: 1 gigawatt of AMD Instinct MI450 GPUs starting in 2H 2026
– MI450 highlights: Up to 40 PFLOPs of AI performance, 432 GB of HBM4 memory, and 19.6 TB/s memory bandwidth
– Additional hardware: Broader deployments will also include Instinct MI300 and MI350 series
– Collaboration scope: Rack‑scale AI solutions and multi‑generation hardware and software co‑optimization
– Strategic alignment: OpenAI receives a warrant for up to 160 million AMD shares, vesting as deployment and performance milestones are met
– Financial outlook: AMD expects the agreement to drive tens of billions in revenue and be accretive to non‑GAAP EPS
This agreement deepens a multi‑year collaboration that started with the MI300X and continued with the MI350X series. By aligning product roadmaps and sharing technical expertise, the companies aim to accelerate advances in generative AI while scaling infrastructure efficiently across future Instinct GPU generations.
The MI450 sits at the heart of the first deployment wave, pairing cutting‑edge HBM4 capacity with massive bandwidth to feed AI workloads at scale. With up to 40 PFLOPs of AI compute per accelerator and rack‑scale design options, the platform is engineered for high‑throughput training and inference across large models and emerging AI applications.
To further align incentives, AMD has issued OpenAI a warrant for up to 160 million shares of AMD common stock. The first tranche vests with the initial 1‑gigawatt rollout, with additional tranches tied to expanding purchases up to 6 gigawatts. Vesting is also linked to AMD share‑price targets and to OpenAI meeting the technical and commercial milestones required for deployments at scale.
Leaders from both companies framed the deal as a pivotal step for the industry. AMD emphasized its high‑performance computing pedigree and the opportunity to advance the broader AI ecosystem, while OpenAI highlighted the need for deep, stack‑level collaboration to bring powerful AI capabilities to more people, faster. AMD’s finance chief noted the agreement is expected to be a strong growth driver and add meaningfully to earnings.
Bottom line: AMD and OpenAI are combining next‑gen GPUs, rack‑scale systems, and a multi‑year roadmap to deliver AI compute at unprecedented scale. With the first MI450 gigawatt landing in late 2026 and larger phases to follow, this partnership is set to shape the next era of high‑performance, generative AI infrastructure.
