AI chipmaker Cerebras namedropped by Oracle, alongside Nvidia and AMD - CNBC

AI chipmaker Cerebras has secured a $10 billion agreement with OpenAI to supply 750 megawatts of computing power through 2028, marking a significant expansion of its partnership with the generative AI leader. This deal follows OpenAI's broader strategy to diversify its AI infrastructure, which includes agreements with established chipmakers such as Nvidia, AMD, and Broadcom.

OpenAI's collaboration with Cerebras complements its existing $100 billion commitment to Nvidia for deploying 10 gigawatts of systems and a $38 billion cloud deal with Amazon Web Services. AMD's involvement includes a multiyear plan to deploy six gigawatts of GPUs, with OpenAI receiving a warrant for up to 10% of AMD shares. Meanwhile, Broadcom's partnership focuses on custom AI accelerators, with 10 gigawatts of capacity targeted for deployment by 2029.

Cerebras' wafer-scale chips, capable of 15 times faster performance than GPU-based systems, aim to reduce latency for real-time AI applications. The deal diversifies Cerebras' customer base, which previously relied heavily on UAE-based G42 for 87% of revenue in mid-2024. CEO Andrew Feldman noted the partnership strengthens Cerebras' position as a challenger to Nvidia's dominance in AI infrastructure.

Potential future partners for OpenAI include Amazon, which is in talks for a $10 billion investment, and Intel, which recently unveiled a new data center GPU but remains an AI laggard. The evolving landscape underscores the competitive dynamics in AI infrastructure, with OpenAI prioritizing scalability and technological differentiation as it pursues a $500 billion private market valuation.