OpenAI is preparing to design and produce its own artificial intelligence chip in partnership with Broadcom Inc., with the first units expected to ship in 2026, the Financial Times reported, citing people familiar with the matter. The chips will initially be used internally, but could position OpenAI as a competitor to Nvidia Corp. in the AI accelerator market.

Broadcom Chief Executive Officer Hock Tan confirmed the company is working with potential customers on AI accelerators, a segment currently dominated by Nvidia. Speaking to investors on Thursday, he said, “Last quarter, one of these prospects released production orders to Broadcom,” without naming the client. He added that the company now anticipates fiscal 2026 AI revenue to improve “significantly” compared with prior forecasts.

According to Reuters, OpenAI is finalizing the design of its first in-house chip and plans to send it for fabrication at Taiwan Semiconductor Manufacturing Company (TSMC) in the coming months.

The move follows similar initiatives by Google, Amazon and Meta, which have developed custom chips to handle surging AI workloads. As demand for computing power to train and operate large-scale AI models grows, companies are increasingly turning to bespoke hardware to optimise performance and reduce reliance on third-party suppliers.