OpenAI Partners with Broadcom on AI Chips
OpenAI has entered a partnership with Broadcom to design and build specialized AI chips. The strategic move is likely aimed at reducing OpenAI's reliance on third-party hardware and optimizing performance for its large language models and enterprise-scale workloads.
- The multi-year collaboration aims to deploy 10 gigawatts of custom-designed AI accelerators, with OpenAI designing the chips and systems and Broadcom co-developing and deploying them. - Deployment of the new hardware is scheduled to begin in the second half of 2026 and is expected to be completed by the end of 2029. - This partnership is the result of approximately 18 months of prior collaboration between the two companies on designing a new custom chip optimized for OpenAI's specific workloads. - Beyond manufacturing the chips, Broadcom will provide its end-to-end portfolio of Ethernet, PCIe, and optical connectivity solutions to scale the AI data centers. - While financial terms were not officially disclosed, Broadcom's CEO Hock Tan announced a new custom chip client with orders over $10 billion, and the company's stock value increased by roughly $126 billion after the partnership was revealed. - This move reflects a broader industry trend of major AI companies like Google, with its Tensor Processing Units (TPUs), designing their own silicon to improve efficiency and reduce reliance on third-party vendors. - The custom chips are designed to lower the staggering costs associated with artificial intelligence; training a model like GPT-4 is estimated to have cost more than $100 million. - The deal complements, rather than replaces, OpenAI's existing partnerships with other chipmakers; the company recently affirmed that Nvidia remains its most important partner for both training and inference.
