Overview
- OpenAI has lined up roughly 26 gigawatts of accelerator capacity across Nvidia, AMD, and Broadcom, with first deployments scheduled to begin in the second half of 2026.
- AMD’s agreement covers 6 GW of Instinct GPUs, starting with 1 GW of MI450 systems in 2H26, and includes equity warrants for OpenAI tied to performance milestones.
- Broadcom will provide 10 GW of custom accelerators and rack systems, with rollout beginning in 2H26 and targeted completion by the end of 2029.
- Analysts say fulfilling the chip orders could require hundreds of billions of dollars as OpenAI projects profitability only in 2029, with reports noting vendor-financing loops and a potential Nvidia investment of up to $100 billion.
- Investor focus has shifted to execution and milestones as HSBC lifted AMD’s target to $310 and Bank of America to $300, highlighting MI450 visibility and a Nov. 11 analyst event, while energy availability and long-term power contracts remain pivotal.