Cerebras Systems

Cerebras Systems (NASDAQ: CBRS) is the pioneer of wafer-scale AI computing, completing the largest U.S. tech IPO of 2026 in May at approximately $60 billion valuation and raising $4.8 billion in the offering.

The company reported $510 million in 2025 revenue with a 47% net margin — making it one of the most profitable semiconductor startups in history. Founded in 2016 by CEO Andrew Feldman and a team of supercomputer architects, Cerebras invented the Wafer-Scale Engine (WSE), a processor that occupies an entire 300mm semiconductor wafer rather than a small die.

The WSE-3 contains 4 trillion transistors, 900,000 AI cores, and delivers 2,625 times more memory bandwidth than NVIDIA B200 while being 58 times physically larger, achieving 125 petaflops of AI compute on a single chip. This architecture eliminates the inter-chip communication bottlenecks that limit GPU cluster performance: training large models on GPUs requires dozens to thousands of chips communicating across high-speed interconnects, whereas the WSE executes the same computation on a single die with no network overhead.

OpenAI signed a landmark 750-megawatt compute deal with Cerebras, making it one of OpenAI primary inference infrastructure partners. Cerebras Inference delivers up to 15 times faster token generation than leading GPU-based solutions, with real-world benchmarks showing 2,100+ tokens per second on Llama-3 70B.

The CS-3 AI Supercomputer is available as an on-premise appliance for organisations requiring dedicated AI compute with data sovereignty, while Cerebras Inference provides a pay-per-token cloud API for developers. Target customers include frontier AI labs, financial services firms, defence and intelligence agencies, and enterprises requiring maximum inference throughput for real-time AI applications.

Cerebras holds patents on wafer-scale computing and cooling that represent a multi-year technology moat over conventional GPU-based approaches.