OpenAI has been on a roll over the past few months, announcing several major partnerships and deals related to AI infrastructure. Last month, the company revealed an agreement to deploy at least 10 gigawatts of NVIDIA AI infrastructure, valued at around $100 billion. Earlier this month, OpenAI announced another deal to deploy 6 gigawatts of AMD Instinct GPUs across its AI infrastructure.
Today, OpenAI announced a new partnership with silicon leader Broadcom to produce 10 gigawatts of custom AI accelerators. As previously rumored, OpenAI will design these new AI accelerators and systems, which will be developed and deployed in collaboration with Broadcom. The company stated that it will leverage the knowledge gained from developing frontier models and products to design its own chips and systems.
These upcoming custom AI accelerator racks will be powered by Ethernet and other connectivity solutions from Broadcom. Interestingly, the official blog post also mentions that these new AI accelerators will be deployed across OpenAI’s facilities and partner data centers. This raises the possibility that Microsoft may use them within its own infrastructure to power OpenAI models.
Sam Altman, co-founder and CEO of OpenAI, said the following regarding the Broadcom partnership:
“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. 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 is expected to start deploying these OpenAI-designed racks of AI accelerators and network systems in the second half of 2026, and the whole 10 GW capacity is expected to be deployed by the end of 2029.
2 Comments
Load the comments and join the conversation!
Read the comments, ask the editors questions, show respect and join the conversation.