OpenAI and Broadcom said Monday they will co-develop and deploy 10 gigawatts of custom AI accelerators and racks, with deployments targeted to start in the second half of 2026 and complete by the end of 2029. OpenAI will design the accelerators and systems; Broadcom will build and deploy the racks, which use Ethernet for scale-up and scale-out networking. The companies did not disclose financial terms, but said they have worked together for 18 months. Broadcom shares rose nearly 10% on the news.

The deal adds to OpenAI’s recent compute commitments. Over the past three weeks, the company has announced partnerships with Nvidia, Oracle, AMD and Broadcom. OpenAI currently operates just over 2 gigawatts of compute capacity. Sam Altman said the systems should lower costs and improve performance, and that demand for high-quality intelligence will absorb the capacity quickly.

Broadcom CEO Hock Tan said custom chips let companies control their own destiny. Broadcom semiconductor president Charlie Kawwas said OpenAI is not the company’s previously disclosed $10 billion mystery customer. The two firms signed a term sheet and released a podcast discussion alongside the announcement.

Skeptics point to industry estimates that a 1-gigawatt data center costs roughly $50 billion, with about $35 billion of that typically allocated to chips. OpenAI has now announced roughly 33 gigawatts of compute commitments over the past three weeks, across partnerships with Nvidia, Oracle, AMD and Broadcom.