OpenAI will start mass producing its own AI chips in 2026

OpenAI is preparing to launch mass production of its own artificial intelligence chips as early as next year. This will allow the company to provide users with massive computing power and reduce its reliance on NVIDIA, the Financial Times reports.

The OpenAI custom chip was reportedly developed in conjunction with US chip maker Broadcom, whose CEO recently announced a new customer with a $10 billion order. Although the customer was not named, Times sources confirmed that it is OpenAI. The chips are planned to be used only internally, with no sales to external customers.

The chips are to be used only internally, with no sales to external customers.

Back in 2023, Reuters reported that OpenAI was exploring the possibility of building its own AI chips after Altman himself pointed to GPU shortages as the cause of the company’s API performance and stability problems. At the time, it was also reported that OpenAI was working with both Broadcom and TSMC to build its own product. The Financial Times doesn’t specify whether the collaboration with TSMC is still ongoing at this time.

After the release of GPT-5, Altman announced OpenAI’s plans to expand its data processing capacity due to “growing demand.” The company intends to double the number of servers over the next five months while prioritizing paid ChatGPT users. The proprietary chips will help address a possible GPU shortage and could reduce the company’s costs.

At the same time, OpenAI is also planning to expand its processing capacity to include more GPUs.

Financial Times notes that custom AI chips, which OpenAI is developing under the name XPUs, could take a significant market share in the future. For now, the leader remains NVIDIA, which recently reported revenue growth of 56% in the second quarter compared to last year, and the company hasn’t even shipped H20 chips to China.