October 6, 2025 — OpenAI and semiconductor company AMD (Advanced Micro Devices) today announced a strategic cooperation agreement under which AMD will become a core compute supplier for OpenAI's "next‑generation AI infrastructure." Under the agreement, OpenAI will phase in purchases of high‑performance AI chips (primarily AMD's Instinct series) and has been granted the right to purchase AMD shares in the future upon achievement of milestones. The pact is seen as a new node in the AI competition over compute and strategic positioning.
Key terms and scale
According to the official press release, the AMD–OpenAI cooperation includes the following main terms:
- Total scale: 6 gigawatts (GW) of compute
- OpenAI has committed to deploy a total of approximately 6 GW of AMD GPU compute capacity over the coming years. The first 1 GW is planned for implementation beginning in the second half of 2026, using AMD's Instinct MI450 series chips.
- Warrants
- To more closely align interests, AMD granted OpenAI the right to purchase approximately 160 million shares of AMD common stock (about 10% of AMD's total outstanding shares) at a nominal exercise price (one cent per share), but exercise is conditional on phased compute deployments and stock price targets being met.
- Technical collaboration and roadmap alignment
- The agreement also includes technical cooperation: OpenAI will work with AMD on future GPU architectures, software optimization and system integration, expanding on prior collaboration around the MI300X/MI350X platforms.
- Revenue and market expectations
- Reuters reported that AMD expects this cooperation to generate billions of dollars in annual revenue, with total revenue impact over four years potentially reaching the order of $100 billion.
Following the announcement, AMD's share price rose sharply in public trading, climbing more than 23% in pre‑market trading, surging over 30% at the open, and closing the day up 23.71%.
Motives and significance
OpenAI's compute diversification strategy
In recent years OpenAI has had close ties with NVIDIA for AI infrastructure. NVIDIA recently signed a separate large investment and compute support agreement with OpenAI (including a commitment of at least 10 GW). This new agreement indicates OpenAI is seeking to diversify its compute supply to avoid overreliance on a single vendor.
OpenAI is also pursuing in‑house chip development efforts and is collaborating with Broadcom on AI chip design to seek lower costs and greater customization.
AMD's AI expansion
For AMD, this is a heavyweight strategic move. Although AMD has built capabilities in AI and data center products (notably in GPUs and accelerators), its market position in large‑scale model training and massive compute deployments has not matched NVIDIA's. Cooperation with OpenAI should bolster AMD's credibility in AI chips, expand its ecosystem influence and increase adoption of its technologies.
AMD CEO Lisa Su said this is an important step for AMD to accelerate its entry into the AI field, and that collaboration with OpenAI will drive deeper hardware and software integration.
Challenges, risks and impact
Supply, manufacturing and power consumption
To start a 1‑GW‑scale deployment in 2026, AMD must ensure mass production, performance, stability and a mature supply chain for its next‑generation chips (such as the MI450). Large‑scale compute deployments also impose very high demands on power, cooling and data center infrastructure. As one report noted, OpenAI's new wave of compute expansion may put pressure on the power grid.
Complexity of the warrant arrangement
The warrant exercise design is based on multiple conditions—"deployment milestones + stock price targets"—which means whether OpenAI can actually acquire those shares depends on future execution and market performance. If those conditions are difficult to meet, the warrants may be partially or fully forfeited.
Impact on NVIDIA and the industry landscape
The AMD–OpenAI cooperation will increase competitive pressure on NVIDIA. NVIDIA has long dominated the AI GPU market; this partnership could encourage more vendors to enter the competition. Analysts generally view the AI chip market as entering a phase of "multi‑polar competition," despite NVIDIA's deep technical lead.
Outlook and key watch points
- Deployment timeline: The second half of 2026 is a critical milestone for both parties: whether the MI450 can be delivered on schedule, whether the initial 1 GW can go online, and whether the first tranche of warrants can be vested.
- Potential partners and scale expansion: As compute demand continues to rise, more AI companies and cloud providers may join AMD's ecosystem or engage in partnerships.
- Market and valuation re‑rating: Successful execution could materially raise AMD's valuation in the AI chip market; investors will closely watch revenue growth and margin performance.
- Spillover effects on compute infrastructure: Such agreements not only involve chip transactions but will also drive development of AI data centers, power and cooling infrastructure, and network interconnects.
Conclusion
The high‑profile cooperation between OpenAI and AMD marks a new phase in the contest for AI infrastructure compute. From diversified sourcing to equity linkage, and from technical collaboration to large‑scale deployment, the strategic implications extend well beyond a simple chip sale. The coming years will be decisive for both parties as they seek to fulfill commitments and shape the market landscape.