OpenAI, Broadcom to roll out 10 Gigawatts of AI accelerators

OpenAI and Broadcom have revealed a long-term collaboration to jointly design, build, and deploy up to 10 gigawatts of custom AI accelerators and networking systems.

OpenAI will lead the architecture and design of the accelerators, while Broadcom will provide its networking, Ethernet, PCIe, and optical connectivity solutions to integrate them into large-scale systems.

Sam Altman, OpenAI’s CEO, noted that collaborating with Broadcom is a critical step in building the infrastructure needed to unlock AI’s potential. While Hock Tan, CEO of Broadcom, called the partnership a pivotal moment in the pursuit of artificial general intelligence, noting the importance of co-developing next-generation accelerators and networking systems at scale.

Read also: OpenAI Dev Day 2025: Ushering SMEs into the new age of AI adoption

“Our collaboration with Broadcom will power breakthroughs in AI and bring the technology’s full potential closer to reality. By building our own chip, we can embed what we’ve learned from creating frontier models and products directly into the hardware, unlocking new levels of capability and intelligence,” Greg Brockman, co-founder and president of OpenAI, stated.

By controlling both the hardware and the networking infrastructure, OpenAI aims to better embed the lessons from its cutting-edge AI models directly into the physical systems that run them. The planned accelerator racks will be fully Ethernet-based and are intended for deployment across OpenAI’s own facilities as well as partner data centers.

Under a signed term sheet, the joint effort will begin deploying these systems starting in the second half of 2026, with completion of deployment expected by the end of 2029. This collaboration reinforces Broadcom’s commitment to custom AI accelerators and reaffirms industry bets on Ethernet as a scalable networking strategy for AI data centers.

“Our partnership with OpenAI continues to set new industry benchmarks for the design and deployment of open, scalable, and power-efficient AI clusters,” said Charlie Kawwas, president of the Semiconductor Solutions Group for Broadcom.

“Custom accelerators combine remarkably well with standards-based Ethernet scale-up and scale-out networking solutions to provide cost and performance optimised next-generation AI infrastructure. The racks include Broadcom’s end-to-end portfolio of Ethernet, PCIe, and optical connectivity solutions, reaffirming our AI infrastructure portfolio leadership,” Kawwas stated.