Cerebras Systems was founded in 2016 by Andrew Feldman, Gary Lauterbach, and others in Sunnyvale, California. The company takes a radical approach to AI hardware: instead of using clusters of small chips, Cerebras builds a single chip the size of an entire silicon wafer.
Cerebras raised over $720 million and filed for an IPO in 2024. Investors include Alpha Wave Global, Eclipse Ventures, and Abu Dhabi’s G42. The company’s wafer-scale approach has attracted significant attention from both the AI research community and sovereign AI initiatives.
The Cerebras WSE-3 (Wafer-Scale Engine 3) contains 4 trillion transistors and 900,000 AI-optimized cores on a single chip — roughly 50 times larger than the biggest GPU. This architecture eliminates the communication bottleneck between chips that limits traditional GPU cluster performance.
Cerebras’s CS-3 system powers the company’s AI inference service, which can run models like LLaMA 3.1 70B at over 2,000 tokens per second — roughly 20 times faster than GPU-based alternatives. The company also partners with organizations building large-Scale AI training clusters, including projects in the Middle East and Europe.
With around 500 employees, Cerebras represents a bold bet that purpose-built silicon can outperform general-purpose GPUs for AI workloads. Their technology serves pharmaceutical companies, national labs, and AI research groups.
