Amazon Web Services (NASDAQ: AMZN) and Cerebras Systems announced a collaboration to deploy what they describe as the fastest AI inference solutions for generative AI applications through Amazon Bedrock in the coming months.

The partnership combines AWS Trainium-powered servers with Cerebras CS-3 systems and Elastic Fabric Adapter networking in AWS data centers. AWS will be the first cloud provider to offer Cerebras's disaggregated inference solution, available exclusively through Amazon Bedrock.

The solution uses "inference disaggregation," which separates AI inference into two stages: prompt processing ("prefill") and output generation ("decode"). AWS Trainium handles the prefill stage, which is computationally intensive and natively parallel, while Cerebras CS-3 manages the decode stage, which is memory bandwidth intensive and inherently serial.

"Inference is where AI delivers real value to customers, but speed remains a critical bottleneck for demanding workloads like real-time coding assistance and interactive applications," said David Brown, Vice President of Compute & ML Services at AWS.

Andrew Feldman, Founder and CEO of Cerebras Systems, stated that the partnership will "bring the fastest inference to a global customer base" and enable enterprises worldwide to access high-speed inference within their existing AWS environment.

AWS plans to offer open-source large language models and Amazon Nova using Cerebras hardware later this year. The solution will be built on the AWS Nitro System to maintain security and operational consistency standards.

Cerebras claims its CS-3 system delivers thousands of times greater memory bandwidth than the fastest GPU and is used by companies including OpenAI, Cognition, and Mistral for demanding workloads. AWS Trainium is currently used by Anthropic and OpenAI for training and deployment of AI models.