OpenAI and Broadcom announce deployment of 10 GW of custom AI accelerators
OpenAI and Broadcom announce a partnership to develop and deploy 10 GW of OpenAI-designed custom AI accelerators and Ethernet systems, with deployment planned from the second half of 2026 through the end of 2029.
What happened
Event details
OpenAI is designing the accelerators and systems, while Broadcom is helping develop and deploy them and providing Ethernet, PCIe and optical interconnects for scale-up and scale-out. The racks will be installed in OpenAI facilities and partner data centers. This is a multiyear infrastructure partnership and future deployment commitment, not evidence that 10 GW of capacity or a specific chip was delivered on the announcement date.
Assessment
Why it matters
The 10 GW partnership elevated OpenAI's custom-accelerator effort from a chip plan to a multiyear, gigawatt-scale systems roadmap and established the context for later Jalapeño silicon and benchmark results.
Availability
Access notes
Deployment is targeted to begin in the second half of 2026 and finish by the end of 2029; no 10 GW of operating capacity existed on the announcement date.