AWS and Cerebras to Launch $23.1B AI Chip Inference Service in H2 2026
Amazon Web Services will embed Cerebras Systems' AI chips alongside its Trainium3 processors to accelerate inference tasks for chatbots and coding tools. Cerebras, valued at $23.1 billion with a prior $10 billion OpenAI deal, expects the service to launch in the second half of 2026.
1. Partnership Details
Amazon Web Services and Cerebras Systems will integrate Cerebras’ AI chips with AWS Trainium3 in a new inference service scheduled for the second half of 2026. Cerebras is valued at $23.1 billion and previously secured a $10 billion chips supply deal with OpenAI.
2. Technical Integration
The service will split inference into a 'prefill' stage handled by AWS Trainium3 and a 'decode' stage on Cerebras chips, connected via custom Amazon networking technology to streamline AI tasks like chatbots and coding tools.
3. Competitive Landscape
The collaboration positions AWS and Cerebras against Nvidia’s upcoming GPU-plus-Groq offering, with AWS projecting Trainium3 (and future Trainium4) to lead in price-performance versus merchant GPUs.