OpenAI is preparing to roll out its first custom artificial intelligence chip next year in collaboration with U.S. semiconductor leader Broadcom, according to a report from the Financial Times citing sources familiar with the matter. The chip is expected to be used internally at OpenAI rather than sold to external clients, one source told FT. Reuters has not independently confirmed the report, and neither OpenAI nor Broadcom responded to requests for comment outside regular business hours.

As the company behind ChatGPT, OpenAI depends heavily on computing resources to develop and run its AI systems. Last year, Reuters reported that OpenAI was working with Broadcom and Taiwan Semiconductor Manufacturing Co. (TSMC) to build its own in-house chip, while continuing to use processors from AMD and Nvidia to keep up with rising infrastructure needs. At that time, the company was exploring ways to broaden its chip supply and reduce expenses. In February, Reuters also noted that OpenAI was pushing forward with plans to reduce reliance on Nvidia by designing its first generation of proprietary AI chips, which it expected to finalize within months before sending them to TSMC for fabrication.

Broadcom CEO Hock Tan said Thursday that the company foresees a significant jump in AI-related revenue by fiscal 2026, citing over $10 billion in new AI infrastructure orders from a recently qualified customer, though he did not name the buyer. Earlier this year, Tan hinted that four new customers were in advanced discussions with Broadcom to create custom chips, alongside its three major existing partners. OpenAI’s initiative mirrors moves by other tech giants such as Google, Amazon, and Meta, which have developed specialized chips to meet the soaring demand for computing power required to train and operate large AI models.