OpenAI has announced a landmark collaboration with Broadcom to co-develop 10 gigawatts of custom AI accelerators and networking systems. Under the multi-year agreement, OpenAI will design the accelerators, while Broadcom will handle the build, deployment, and network integration—starting in the second half of 2026 and completing by end of 2029.

This move adds to OpenAI’s already rapid expansion of compute capacity and infrastructure commitments. Earlier this month, OpenAI inked chip supply deals with AMD and Nvidia and entered large scale data center partnerships. The Broadcom deal is particularly strategic: by designing its own chips, OpenAI seeks more control over performance, cost, and integration with its models. 

Analysts caution that despite the ambitious scale, it won’t immediately threaten Nvidia’s dominance in AI accelerators. The financial terms remain undisclosed, though deploying tens of gigawatts of compute could cost tens of billions.  Still, the Broadcom tie-up signifies OpenAI’s determination to vertically integrate infrastructure and double down on its leadership in generative AI.