AMD and OpenAI Announce Strategic Partnership to Deploy 6GW of AMD Instinct GPUs

Posted on Edited on
AMD and OpenAI ink a 6 GW Instinct GPU pact with MI450 rollout in H2 2026 and a warrant for up to 160 million AMD shares tied to AI infrastructure milestones.

Quick Report

AMD and OpenAI announced a multi-generation partnership that will roll out up to 6 gigawatts of AMD Instinct GPU capacity, starting with a 1 GW deployment of forthcoming MI450 accelerators in the second half of 2026. The deal includes a warrant granting OpenAI the option to earn up to 160 million AMD shares as deployments scale, aligning incentives around both infrastructure build-out and AMD share-price performance.

Under the definitive agreement, OpenAI designates AMD as a core strategic compute partner across current and future Instinct roadmaps, extending their collaboration beyond MI300X and MI350X systems. The partnership spans rack-scale AI solutions, joint product roadmap tuning, and software optimization aimed at accelerating large language model training at hyperscale. AMD leadership says the arrangement could generate tens of billions of dollars in revenue while OpenAI gains diversified supply to complement its existing NVIDIA-heavy footprint.

Key points called out in AMD's release:

  • First 1 GW build begins H2 2026 using Instinct MI450 GPUs, with subsequent phases scaling to 6 GW across future families.
  • OpenAI receives a warrant for up to 160 million AMD common shares, vesting in tranches tied to deployment milestones, AMD share-price targets, and OpenAI's technical/commercial progress.
  • AMD CFO Jean Hu projects the deal will be highly accretive to non-GAAP EPS, underscoring the revenue scale expected from the build-out.
  • Quotes from Dr. Lisa Su, Sam Altman, Greg Brockman, and Jean Hu emphasize the mutual goal of expanding global AI compute capacity.

This pact gives OpenAI a formal second source for advanced accelerators just as the company ramps its ambitious AI factory roadmap. AMD gains long-term visibility into multi-gigawatt demand for its Helios platform and Instinct GPU families, plus stock-aligned incentives that tether OpenAI's purchases to AMD's market valuation. If AMD can execute on MI450 performance per watt targets and continue improving ROCm software, the company could convert this agreement into meaningful share gains against NVIDIA in the hyperscale AI segment.

For OpenAI, diversifying suppliers mitigates risk amid intense competition for high-end GPUs and power delivery. The 6 GW figure implies millions of accelerators over multiple years, driving substantial pressure on data center power, cooling, and supply chains. Success hinges on AMD delivering competitive silicon, robust software tooling, and reliable manufacturing scale through partners like TSMC.

Written using GitHub Copilot GPT-5 mini in agentic mode instructed to follow current codebase style and conventions for writing articles.

Source(s)

  • TPU
  • AMD Newsroom Post