OpenAI has announced a massive new partnership with Broadcom to co-develop custom AI accelerators and networking systems powerful enough to support 10 gigawatts of computing infrastructure. The collaboration, revealed today, aims to deliver next-generation AI clusters beginning in the second half of 2026, with full deployment expected by the end of 2029.
This isn’t just another chip deal. OpenAI will design its own accelerators and system architecture, while Broadcom handles manufacturing and integration using its Ethernet-based networking and connectivity stack. The approach will allow OpenAI to translate lessons from its large language models and AI products directly into silicon. This could redefine efficiency and performance across the AI landscape.
ALSO READ: AI could be the lifeline lonely Americans need
According to OpenAI CEO Sam Altman, the collaboration is “a critical step in building the infrastructure needed to unlock AI’s potential and deliver real benefits for people and businesses.” He emphasized that designing custom accelerators will help scale the hardware ecosystem necessary to advance frontier AI models.
Broadcom CEO Hock Tan called the collaboration “a pivotal moment in the pursuit of artificial general intelligence,” highlighting how OpenAI’s leadership since the ChatGPT era continues to push the boundaries of computing.
OpenAI co-founder and President Greg Brockman added that the company’s decision to develop its own chips will “embed what we’ve learned from creating frontier models directly into the hardware,” helping close the gap between software and silicon.
With over 800 million weekly active users and growing demand from enterprises and developers, OpenAI’s shift into hardware represents a new phase for the company… one that blurs the line between AI research lab and semiconductor innovator. Whether this custom approach outpaces NVIDIA’s dominance remains to be seen, but it’s clear OpenAI intends to control more of the AI stack from chip to cloud.
