OpenAI to Begin Mass Production of AI Chips with Broadcom

OpenAI is preparing to launch mass production of its own artificial intelligence chips next year in collaboration with US semiconductor giant Broadcom, in a move aimed at easing its reliance on Nvidia and meeting the surging demand for computing power.

According to people familiar with the matter, the custom chips—co-designed by OpenAI and Broadcom—are expected to begin shipping in 2026. The development comes after more than a year of joint work between the two companies, with Broadcom chief executive Hock Tan recently confirming a $10bn commitment from a new, undisclosed client.

The deal marks a significant step in Broadcom’s fast-growing custom AI chip business, which analysts expect to expand at a faster pace than Nvidia’s from 2026 onward.

OpenAI intends to deploy the chips internally to support its AI models including ChatGPT, rather than selling them to external customers. The decision reflects a strategy similar to that of tech giants google, Amazon, and Meta, which have also developed proprietary chips to run AI workloads.

Sam Altman chief executive of OpenAI, has repeatedly underscored the urgent need for greater computing resources to serve both enterprise and consumer demand. OpenAI has historically been one of Nvidia’s largest customers but has increasingly sought alternatives amid industry-wide shortages and rising costs.

On a recent earnings call, Tan said Broadcom’s fourth major custom AI chip client would begin receiving shipments “pretty strongly” from next year, without naming OpenAI directly. Both companies declined to comment.

The announcement highlights the growing role of specialised AI chips in the booming artificial intelligence infrastructure market, which has propelled Broadcom’s shares more than 44% higher this year.

Advertisment