OpenAI partners with Broadcom custom AI chips alongside Nvidia, AMD

Weeks after Broadcom‘s stock shot up on the disclosure of a $10 billion customer that analysts said was OpenAI, the two companies have made the partnership official.

OpenAI and Broadcom said Monday that they’re jointly building and deploying 10 gigawatts of custom artificial intelligence accelerators as part of a broader effort across the industry to scale AI infrastructure.

Broadcom shares climbed 7% following news of the deal.

They didn’t disclose financial terms.

While the companies have been working together for 18 months, they’re now going public with plans to develop and deploy racks of OpenAI-designed chips starting late next year. OpenAI has announced massive deals in recent weeks with Nvidia, Oracle and Advanced Micro Devices, as it tries to secure the capital and compute needs necessary for its historically ambitious AI buildout plans.

“These things have gotten so complex you need the whole thing,” OpenAI CEO Sam Altman said in a podcast with OpenAI and Broadcom executives that the companies released along with the news.

The systems include networking, memory and compute — all customized for OpenAI’s workloads and built on Broadcom’s Ethernet stack. By designing its own chips, OpenAI can bring compute costs down and stretch its infrastructure dollars further. Industry estimates peg the cost of a 1-gigawatt data center at roughly $50 billion, with $35 billion of that typically allocated to chips — based on Nvidia’s current pricing.

The Broadcom deal provides “a gigantic amount of computing infrastructure to serve the needs of the world to use advanced intelligence,” Altman said. “We can get huge efficiency gains, and that will lead to much better performance, faster models, cheaper models — all of that.”

Broadcom has been one of the biggest beneficiaries of the generative AI boom, as hyperscalers have been snapping up its custom AI chips, which the company calls XPUs. Broadcom doesn’t name its large web-scale customers, but analysts have said dating back to last year that its first three clients were Google, Meta and TikTok parent ByteDance.

Shares of Broadcom are up 40% this year after more than doubling in 2024, and the company’s market cap has surpassed $1.5 trillion.

OpenAI President Greg Brockman said the company used its own models to accelerate chip design and improve efficiency.

“We’ve been able to get massive area reductions,” he said in the podcast. “You take components that humans have already optimized and just pour compute into it, and the model comes out with its own optimizations.”

Broadcom CEO Hock Tan said in the same conversation that OpenAI is the company building “the most-advanced” frontier models.

“You continue to need compute capacity — the best, latest compute capacity — as you progress in a road map towards a better and better frontier model and towards superintelligence,” he said. “If you do your own chips, you control your destiny.”

Altman indicated that 10 gigawatts is just the beginning.

“Even though it’s vastly more than the world has today, we expect that very high-quality intelligence delivered very fast and at a very low price — the world will absorb it super fast and just find incredible new things to use it for,” he said.

OpenAI today operates on just over 2 gigawatts of compute capacity.

Altman said that’s been enough to scale ChatGPT to where it is today, as well as develop and launch video creation service Sora and do a lot of AI research. But demand is soaring.

OpenAI has announced roughly 33 gigawatts of compute commitments over the past three weeks across partnerships with Nvidia, Oracle, AMD and Broadcom.

“If we had 30 gigawatts today with today’s quality of models,” he added, “I think you would still saturate that relatively quickly in terms of what people would do.”

WATCH: China opens antitrust probe into Qualcomm’s Autotalks deals