AWS and Cerebras Partner to Deliver Fastest AI Inference on Bedrock

Amazon Web Services and Cerebras Systems announced a partnership to provide what they describe as the fastest AI inference available in the cloud, according to a press release. The collaboration will deploy Cerebras CS-3 systems in AWS data centers and make them accessible through Amazon Bedrock in the coming months.

The joint solution combines AWS Trainium processors, optimized for prefill computation, with Cerebras CS-3 hardware, optimized for decode operations. These components are connected through Amazon’s Elastic Fabric Adapter networking, enabling a disaggregated inference architecture that separates the two stages of AI inference—prefill and decode—for greater efficiency and speed.

As detailed by Cerebras, this configuration allows Trainium to handle compute-intensive prefill tasks while the CS-3 focuses on generating output tokens. The companies claim this setup will deliver up to five times more high-speed token capacity within the same hardware footprint compared to traditional GPU-based systems.

The service, available via Amazon Bedrock, will support leading open-source large language models as well as Amazon’s own Nova models. AWS plans to roll out the integrated Trainium and CS-3 inference capability globally later this year.