Cerebras Systems is an American AI hardware company founded in 2015 by Andrew Feldman, Gary Lauterbach, Michael James, Sean Lie, and Jean-Philippe Fricker, headquartered in Sunnyvale, California. Its defining bet is wafer-scale computing: rather than cutting a silicon wafer into many small chips, Cerebras fabricates a single enormous processor across nearly the whole wafer.
That processor is the Wafer-Scale Engine (WSE). The first generation, unveiled in 2019, packed 1.2 trillion transistors and was billed as the largest chip ever built; later generations, through the WSE-3 introduced in 2024, increased the count further. The chip’s huge on-chip memory and bandwidth let an entire neural network sit on one piece of silicon, avoiding much of the communication overhead that slows clusters of conventional GPUs. Cerebras sells these as CS-series systems and positions itself as a fast platform for both AI training and inference.
By 2025 Cerebras had shifted heavily toward high-speed inference, announcing data-center expansion across North America and Europe and powering public APIs for open models such as Meta’s Llama and Mistral. It is one of a small group of companies - alongside Groq, Google’s TPU effort, and others - trying to break Nvidia’s dominance of AI compute with a fundamentally different chip architecture.