OpenAI and Broadcom plan 10 GW of custom AI accelerators by 2029

OpenAI said Monday it has partnered with chipmaker Broadcom to design and deploy custom artificial intelligence processors, a move the companies said will add roughly 10 gigawatts of AI accelerator capacity beginning in the second half of 2026 and completed by the end of 2029. The two California firms said OpenAI will design the chips and Broadcom will build, fabricate and help deploy the systems, which they expect to scale using Broadcom networking equipment.

Under the agreement, OpenAI-designed racks of AI accelerators and associated network systems will be developed and rolled out across OpenAI facilities and partner data centers, the companies said. Financial terms were not disclosed and OpenAI did not detail how it would fund the program. Broadcom shares rose sharply on the announcement, gaining about 7 percent in early trading and approaching a double-digit intraday increase later in the session.

OpenAI has been lining up multiple infrastructure partnerships in recent weeks. The company has separately announced plans to deploy large volumes of Nvidia and AMD processors and has secured capacity and investments tied to those suppliers. The Broadcom deal is intended to...