Amazon Collaborates with Cerebras to Integrate AI Chips, Boosting Inference Efficiency for Chatbots and Other Apps

U.S. tech giant Amazon and AI chip startup Cerebras Systems have reached a collaboration to integrate their chips into a new service, accelerating the inference process for AI applications like chatbots and programming tools. The service is expected to launch in the second half of this year.

Cerebras chips will be deployed in Amazon Web Services (AWS) data centers, connected to Amazon's self-developed Trainium3 chips, splitting inference tasks: Trainium3 converts natural language requests into tokens, while Cerebras generates answers. This partnership allows users to easily access Cerebras' computing power.

The AI industry is shifting from model training to inference. Enterprises need chips with higher response speeds. Amazon says its service will offer better cost-performance than GPUs on the market, and Trainium3 is only months away from production-level use, ahead of Nvidia's integration plan with Groq.