OpenAI and Broadcom announced a multi-year partnership under which the two companies will jointly design and deploy 10 gigawatts of custom AI accelerators. The collaboration aims to create new AI processing units designed by OpenAI and manufactured and implemented by Broadcom.
According to the official statement, the first installations of these chips are expected to begin in the second half of 2026, with the entire project scheduled for completion by the end of 2029. It remains unclear whether this timeline means full independence from external suppliers by that point. The partnership also covers the development of network infrastructure based on Broadcom’s Ethernet technology, which will enable the scaling of massive AI clusters.
OpenAI’s goal is to transfer its expertise from building models like GPT-4 and GPT-5 from software to hardware, improving energy efficiency and performance when training next-generation models. The company also seeks to reduce its dependency on external GPU providers — primarily Nvidia and AMD.
“Partnering with Broadcom is a critical step in building the infrastructure needed to unlock AI’s potential and deliver real benefits for people and businesses,” said Sam Altman, co-founder and CEO of OpenAI. “Developing our own accelerators adds to the broader ecosystem of partners all building the capacity required to push the frontier of AI to provide benefits to all humanity.”
Following the announcement, Broadcom’s (AVGO) stock surged by up to 9%, reaching an all-time high and adding tens of billions of dollars to its market capitalization in a single day.
However, despite its ambitious plans, OpenAI and Broadcom will face tough competition — Nvidia’s dominance in the AI accelerator market is clear, and it benefits from years of experience in the field. Developing hardware that can match or exceed Nvidia’s performance will require enormous financial and engineering investment.
If successful, OpenAI could become one of the first companies in the world to fully control both the software and hardware layers of its AI infrastructure — effectively becoming the “Apple of AI”, designing an integrated ecosystem from the ground up.
