11:38 ET -- Amazon's plan to deploy Cerebras chips shows the shift in AI computing toward inference functions, which allow AI models to respond to user queries, instead of model training. The GPU chips used widely for AI model training aren't as good for inference workloads that require more speed. Cerebras says its chips can process the phase of inference computing in which an AI model spits out a response to a user query up to 25 times faster than Nvidia's GPUs. Next week, Nvidia plans to unveil a new processing system tailored for inference using the technology of chip startup Groq, with which it signed a licensing deal in December. (nicholas.miller@wsj.com)
(END) Dow Jones Newswires
March 13, 2026 11:38 ET (15:38 GMT)
Copyright (c) 2026 Dow Jones & Company, Inc.
Comments