OpenAI will roll out its first in-house artificial intelligence (AI) chip in 2026 through a partnership with US semiconductor giant Broadcom, Reuters reported based on a report by the Financial Times, citing people familiar with the matter.
The chip will power OpenAI’s own AI models, including ChatGPT, as the company looks to cut costs and reduce heavy reliance on Nvidia, whose dominance in AI hardware has driven up both demand and prices.
Broadcom CEO Hock Tan said on Sept 4 that the company had secured over US$10 billion in AI infrastructure orders from a new client, widely believed to be OpenAI. He added that Broadcom’s AI revenues will “improve significantly” in fiscal 2026 as custom chip demand accelerates.
OpenAI’s push into custom silicon follows moves by Google, Amazon and Meta, all of which have built proprietary chips to handle massive AI workloads. The company is also working with TSMC on chip fabrication, while continuing to use AMD and Nvidia hardware to support current infrastructure needs.
By producing its own chip, OpenAI aims to gain greater control over its supply chain and secure the computing power needed to train increasingly complex AI models.
