SAN FRANCISCO, January 13, 2026 – OpenAI has entered a multibillion-dollar partnership with Cerebras Systems to secure additional high-performance computing capacity for training and running its artificial intelligence models, according to people familiar with the agreement.
The deal, valued in the billions over multiple years, grants OpenAI access to Cerebras’ wafer-scale AI chips and cloud infrastructure. Cerebras’ CS-3 system, featuring the world’s largest single processor with 4 trillion transistors, is optimized for large-scale AI workloads, offering significantly higher memory bandwidth and lower latency than traditional GPU clusters.
The agreement comes as OpenAI faces escalating compute demands for next-generation models while navigating supply constraints on Nvidia hardware due to global competition and export restrictions. OpenAI has been diversifying its compute sources, including partnerships with Microsoft, Oracle, and CoreWeave.
Cerebras, founded in 2016, specializes in massive single-chip designs that reduce data movement bottlenecks common in multi-GPU setups. The company has positioned itself as an alternative to Nvidia’s dominance in AI training, claiming up to 20 times faster performance for certain large language model workloads.
Neither OpenAI nor Cerebras immediately commented on the partnership. Details on the exact capacity allocated or the timeline for deployment were not disclosed.
The move reflects broader industry trends toward multi-vendor compute strategies as AI training costs soar into the billions. OpenAI’s previous reliance on Microsoft Azure has drawn scrutiny, particularly after reports of capacity shortages during peak training periods.
Cerebras has secured contracts with government and research institutions, including the U.S. Department of Energy and pharmaceutical companies, but commercial hyperscaler deals remain relatively rare. The partnership with OpenAI represents a major validation for the startup’s architecture in the competitive AI infrastructure market.
As AI model sizes continue to grow, access to diverse, high-performance compute resources has become a critical competitive factor. The deal could accelerate OpenAI’s development timeline for future models while providing Cerebras with a high-profile commercial reference.
The agreement is subject to standard regulatory review but is expected to proceed without significant delays.
