AMD and OpenAI announce strategic partnership to deploy 6 gigawatts of AMD GPUs

Source: OpenAI Date: 2025-10-06 URL: https://openai.com/index/openai-amd-strategic-partnership

Summary

OpenAI and AMD announced a strategic partnership in October 2025 to deploy 6 gigawatts of AMD GPU capacity — a massive infrastructure commitment reflecting OpenAI’s Stargate-era compute expansion and AMD’s push to capture AI training and inference market share from Nvidia. The partnership positioned AMD’s Instinct MI-series accelerators as viable alternatives (or complements) to Nvidia’s H100/H200/GB200 family in OpenAI’s production infrastructure, with 6GW representing a scale commitment that would require years of datacenter buildout to realize.

Implications

Supply chain diversification at scale. OpenAI’s AMD partnership followed the Broadcom custom chip collaboration (October 2025) and the Oracle Stargate expansion — a pattern of multi-vendor hardware strategy designed to reduce Nvidia dependency and negotiate better pricing through competition. At 6GW, this represented a multi-year commitment requiring significant AMD roadmap delivery on MI400-series and beyond.

Thread: Stargate infrastructure build-out. Sits in the infrastructure expansion sequence with the January 2025 Stargate announcement, the Oracle 4.5GW partnership (July 2025), the Michigan expansion (October 2025), and the Foxconn manufacturing collaboration (November 2025). Each announcement added hardware, geography, or vendor diversity to the compute stack.

Watch: Whether AMD’s software ecosystem (ROCm, inference stack) was mature enough for OpenAI’s production workloads, and how the Nvidia/AMD/Broadcom custom chip mix evolved through 2026.