Broadcom and OpenAI have signed a deal to develop their own AI chips worth several billion dollars

OpenAI has announced a strategic partnership with Broadcom to jointly develop and deploy 10 gigawatts of specialized AI chips and server systems over the next four years, one of the largest contracts in the artificial intelligence space to meet the company's growing computing needs, The Wall Street Journal reports.

The new chips will be developed jointly with Broadcom. They will begin production in the second half of 2026. The deal is valued at several billion dollars, although the parties did not disclose the exact financial terms.

Broadcom has been working with OpenAI on its own chips for 18 months, according to the sources. The collaboration also includes server racks, networking equipment and Ethernet solutions that will be used in both OpenAI's own data centers and partner facilities.

This deal increases the total computing power that OpenAI has ordered from Broadcom, NVIDIA, and AMD to 26 gigawatts—more than double what New York City needs to cover its summer peak electricity consumption.

According to WSJ estimates, OpenAI will have to spend hundreds of billions of dollars to finance such a scale. At the same time, the company expects to generate $13 billion in revenue in 2025.

OpenAI, now valued at $500 billion and with over 800 million weekly active users, has repeatedly said its CEOs, Sam Altman and Greg Brockman, lack the computing resources to develop new AI models.

According to sources, OpenAI plans to create 250 gigawatts of new computing power by 2033 — a cost that could exceed $10 trillion.