Infrastructure atlas
Cerebras/Wafer-scale engine

Wafer Scale Engine 3

A 5 nm wafer-scale processor designed to keep model computation on one very large device instead of distributing it across GPU nodes.

ASICwafer-scaletraininginferenceSRAM
Wafer Scale Engine 3 hardware

01 / AI cores

900,000

On one wafer-scale engine

02 / On-chip SRAM

44 GB

Fast local memory

03 / Peak AI

125 PFLOPS

Cerebras published figure

Precision & peak compute

PrecisionPublished peakBasis
AI compute125 PFLOPSWSE-3 published aggregate; precision not specified in this record

Peak FLOPS and TOPS are theoretical hardware figures. Sparse and dense modes, chip and system totals, and form factors are not interchangeable.

Cerebras WSE-3 announcement

Technical specification

Wafer-scale silicon
Process
5 nm
Transistors
4 trillion
Compute cores
900,000 AI-optimized cores
On-chip memory
44 GB SRAM
System & scale
System
CS-3
Scale-out
Up to 2,048 CS-3 systems in published architecture
Model parallelism
Wafer-scale memory and fabric
Best fit
Large-model training and high-throughput inference