< Back to all clusters
[TECHNOLOGY] · United States · 20 sources

started · updated

Cerebras Systems unveils CS-4 AI accelerator to challenge GPU dominance

Cerebras Systems has unveiled the CS-4, a new rack-scale AI accelerator designed to compete with NVIDIA in the AI inference market. The system is the first product of the Cerebras Nexus platform architecture and is built using three overclocked Wafer Scale Engine 3 Turbo (WSE-3T) chips manufactured with TSMC's 5-nanometer process.

According to the company, the CS-4 delivers up to 30 times faster inference than GPU-based solutions and up to 10 times more throughput per watt than the previous CS-3 model. The modular design features 50% fewer components, which Cerebras claims will reduce data center deployment time from days to hours. The system is designed to support massive frontier models with over 50 trillion parameters by reducing wafer-to-wafer latency to two microseconds.

Financially, Cerebras reported a second-quarter revenue of $180 million, though it recorded a net loss of $450.5 million, largely due to $386.6 million in stock-compensation expenses. Despite the loss, the company raised its full-year core revenue guidance to between $880 million and $890 million. The CS-4 is expected to be available to customers in the third quarter of 2026.

Entities

Andrew Feldman · Cerebras Systems · Dylan Patel · Nexus · Nvidia · OpenAI · Sean Lie · TSMC

Claims

What the coverage asserts, and how many sources carry each claim.

Sources

about 7 hours ago
1 day ago
about 9 hours ago