OpenAI and Broadcom announced a strategic partnership to jointly develop and deploy 10 gigawatts of tailored AI accelerators, integrating OpenAI-designed chips with Broadcom’s Ethernet and connectivity products. The multi-year partnership will expand AI infrastructure throughout OpenAI centers and partner data centers, with deployment beginning in the second half of 2026 and finishing by the end of 2029.
By creating its own accelerators, OpenAI can implement knowledge from its edge AI models directly into hardware to achieve levels of efficiency and intelligence unseen before. Broadcom will provide racks of networking and accelerator systems, offering the scale-out and scale-up performance needed for future AI clusters.
“Partnering with Broadcom is a critical step in building the infrastructure needed to unlock AI’s potential and deliver real benefits for people and businesses,” said Sam Altman, CEO of OpenAI. “Developing our own accelerators adds to the broader ecosystem of partners all building the capacity required to push the frontier of AI to provide benefits to all humanity.”
Broadcom’s partnership with OpenAI marks an inflection point in the quest for artificial general intelligence,” said Broadcom CEO Hock Tan. “We are excited to co-design and deliver 10 gigawatts of next-generation accelerators and network infrastructure to lay the foundation for the future of AI.
Also Read: Cisco Unveils 51.2T 8223 Routing Systems Powered by P200 Chip to Revolutionize AI Networking
The collaboration showcases the increasing significance of bespoke accelerators and Ethernet-based networking for providing scalable and power-conscious AI infrastructure. The project will incorporate insights from leading-edge AI directly into hardware, Greg Brockman, President of OpenAI, says.
Charlie Kawwas, Head of Broadcom’s Semiconductor Solutions Group, highlighted that the partnership provides “new industry standards for designing and deploying open, scalable and power-efficient AI clusters,” pairing accelerators with Broadcom‘s end-to-end Ethernet, PCIe, and optical solutions to maximize performance and cost.
With more than 800 million weekly active users and widespread enterprise usage, OpenAI plans to use this infrastructure to advance its mission of making artificial general intelligence work for all of humanity.