Cerebras Systems, based in Sunnyvale, CA, is pioneering a new class of computer systems to accelerate AI by orders of magnitude. The company was founded in 2015 and is led by CEO Andrew Feldman, a seasoned Silicon Valley entrepreneur who previously co-founded and led SeaMicro (acquired by AMD for $357 million) and had roles at companies like Force10 Networks and Riverstone Networks.

At the core of Cerebras’ solution is the Wafer Scale Engine (WSE), the largest chip ever built with 2.6 trillion transistors. The WSE powers systems like the CS-2, enabling breakthroughs across industries.

Major customers include TotalEnergies in energy, GlaxoSmithKline, and AstraZeneca in pharma, the Mayo Clinic for medical AI, and national labs like Argonne and Lawrence Livermore. Cerebras has around 335 employees but a major industry impact, with customer commitments approaching $1 billion.

With over 400,000 AI cores, massive memory bandwidth, and ultra-fast interconnects, the CS-2 delivers cluster-scale performance in a single system. This allows for cutting training times from months to minutes on complex workloads like seismic modeling, drug discovery, and COVID research.

Key verticals are pharmaceuticals/healthcare, energy/industrial, scientific research, and AI/ML. Cerebras removes current hardware constraints, allowing researchers to explore novel architectures and drive previously impossible innovations.

As AI complexity grows, wafer-scale computing solutions like those developed by Cerebras Systems may be increasingly important in accelerating AI workloads and enabling new innovations across various industries.

Credit: Jeffrey Cooper & Perplexity.ai