OpenAI, AMD sign 6 GW AI compute deal
OpenAI taps AMD in multibillion AI chip pact
OpenAI will source 6 GW of AI compute from AMD starting in 2026, with warrants for up to 160M shares—boosting AMD’s challenge to Nvidia and diversifying OpenAI’s supply.
Overview
OpenAI and AMD announced a sweeping partnership that will deliver six gigawatts of compute capacity to power inference for OpenAI’s models, positioning AMD as a stronger rival to Nvidia and further diversifying OpenAI’s chip supply.
Deal details
Deployment timeline
AMD plans to begin deploying the first gigawatt of its Instinct MI450 GPUs in the second half of 2026, kicking off a multi‑year delivery schedule to reach 6 GW.
Equity component
OpenAI will receive warrants for up to 160 million shares of AMD common stock, tied to performance milestones over the next five years. While no specific price tag was disclosed, the agreement is described as multibillion‑dollar.
Market reaction
Investors cheered the news, sending AMD shares up more than 34%.
Context: OpenAI’s broader infrastructure push
The pact adds to a string of infrastructure moves by OpenAI. Last month, the company announced a $100 billion contract with Nvidia to provide 10 gigawatts of training compute and a $300 billion agreement with Oracle to secure capacity through 2030. Together, the deals underscore OpenAI’s drive to lock in massive compute while avoiding dependence on a single provider.
Analyst viewpoints
Analysts say the AMD tie‑up reflects a deliberate push to source chips from multiple vendors. Jack Gold of J. Gold Associates noted OpenAI is likely to keep expanding its roster of hardware partners. For AMD, the agreement bolsters its credibility as a viable alternative to Nvidia, which still controls more than 70% of the market for GPUs that power generative AI, according to Gartner’s Gaurav Gupta.
Why it matters
- For OpenAI: Diversifies chip supply, secures large‑scale inference capacity, and reduces single‑vendor dependence.
- For AMD: Validates MI‑series GPUs in hyperscale AI workloads and elevates AMD as a stronger competitor to Nvidia.
- For AI buyers: Greater competition could moderate infrastructure prices over time.
- For competitors: Large multi‑year allocations may tighten near‑term supply for other model developers, including Anthropic.
Risks and challenges
The strategy carries financial risk. OpenAI’s heavy infrastructure spending has outpaced dependable enterprise revenue, and the company has been losing billions annually. If the AI market cools, the fallout could be felt widely across vendors and customers, analysts cautioned.
What’s next
AMD is slated to roll out its first tranche of MI450 GPUs in 2026, with deliveries scaling over multiple years. The partnership elevates AMD’s standing in AI chips while giving OpenAI another crucial channel for compute—intensifying the race to secure the resources behind cutting‑edge AI.
Key numbers at a glance
- Compute capacity: 6 GW (inference)
- Initial deployment: 1 GW in H2 2026
- GPU: AMD Instinct MI450
- Equity: Warrants for up to 160M AMD shares over 5 years
- Market move: AMD shares +34% on the news
