Cerebras launches disaggregated inference: GPU prefill + wafer-scale decode
Understand this story
Cerebras launches a disaggregated inference solution: GPUs handle prompt prefill, while Cerebras wafer-scale engines handle decode. Cerebras says the solution is already running in production deployments of its Inference Cloud, delivering 5x the throughput per megawatt of a pure CS-3 configuration with unchanged generation speed. Early partners including AWS, Fireworks AI, Gimlet Labs, and General Compute can already use the solution in their own deployments, and Cerebras says it is expanding access to more partners. Cerebras plans to scale disaggregated inference to hundreds of megawatts across the Cerebras Inference Cloud and partner deployments by the end of 2027.
AI overview based on source reports · Updated 6 hr ago
Coverage timeline
Follow the coverage to explore different perspectives.
- Cerebras BlogFeaturedCerebras launches disaggregated inference
Cerebras says it has put disaggregated inference into production, with GPUs handling prefill and Cerebras systems handling decode, delivering five times more tokens per megawatt at the same Cerebras speeds in a measured deployment on its Inference Cloud. The company says it plans to deploy the approach across hundreds of megawatts by the end of 2027 and is offering partners a reference architecture plus an end-to-end software stack or a decode-only SDK. Early-access partners named include Amazon Web Services, Fireworks AI, Gimlet Labs and General Compute.
Activity for this story
- Current comparable activity
- 4
- Peak comparable activity
- 5Nov 18 03:00
- Change in the past 24 hours
- –
Trends compare the same fully observed participants, which may cover fewer participants than current activity. Move the pointer or tap the chart to inspect hourly activity; use the left and right arrow keys with a keyboard.