OpenAI has announced a major partnership with Broadcom to design and produce 10 gigawatts of custom AI accelerator hardware. The collaboration will deliver next-generation AI systems to OpenAI’s own data centers and partner facilities between 2026 and 2029, significantly expanding the company’s computing infrastructure.
The agreement marks a key step in OpenAI’s strategy to design its own AI chips and systems, integrating insights from frontier model development directly into hardware for improved efficiency and intelligence.
While financial terms were not disclosed, the partnership follows OpenAI’s recent multi-billion-dollar infrastructure deals with AMD, Nvidia, and Oracle, underscoring its push to secure long-term computing capacity for advanced AI model training.
