OpenAI taps AMD GPUs
OpenAI’s multi‑year infrastructure tie will deploy AMD Instinct GPUs at scale — the briefing flags a 6 GW deployment that funnels tens of billions in revenue toward AMD and intensifies the hardware race. That underlines how compute supply and vendor economics now sit at the center of AI strategy conversations. (markets.businessinsider.com)
AMD and OpenAI publicly filed the pact on October 6, 2025, outlining a multi‑year, multi‑generation agreement that starts with a 1‑gigawatt deployment of AMD Instinct MI450 Series GPUs scheduled for the second half of 2026. (ir.amd.com) AMD issued OpenAI a warrant to buy up to 160 million AMD shares at an exercise price of $0.01 per share, with vesting tied to delivery milestones that begin after the initial 1 GW tranche and accelerate as purchases scale to the full 6 GW. (ir.amd.com) The SEC filing also shows full vesting of the 160 million shares is contingent on meeting specified company stock‑price thresholds that escalate to $600 per share for the final tranche. (ir.amd.com) Market reaction included a one‑day surge in AMD equity of roughly 23.7% on the announcement, reflecting investor reassessment of AMD’s role in AI compute supply chains. (cnbc.com) Independent back‑of‑the‑envelope analysis equates 6 GW to roughly 3.5 million–5 million AMD GPUs and implies simultaneous demand for large volumes of EPYC CPUs and HBM stacks to populate those racks. (fierce-network.com) AMD has been positioning a rack‑scale reference design called “Helios” (double‑wide, liquid‑cooled, built to Meta’s Open Rack Wide standard) for volume deployment alongside the MI450 rollout, and AMD has announced manufacturing and OEM partnerships — including a Celestica collaboration — to commercialize Helios in 2026. (convergedigest.com)
