Wafer Scale Engine

Cerebras' massive AI processor that uses nearly an entire silicon wafer as a single chip, measuring 8.5 inches per side with trillions of transistors.

The Wafer Scale Engine (WSE) is Cerebras Systems' approach to AI chip design. Traditional processors are thumbnail-sized fragments cut from silicon wafers; the WSE instead uses almost the entire 300mm wafer as a single integrated chip. The current generation contains 4 trillion transistors and 900,000 specialized AI cores. This design eliminates the memory bandwidth bottleneck that occurs when shuttling data between separate chips in conventional GPU clusters, with Cerebras claiming 20x faster AI inference than competing multi-chip systems.

Also known as