Cerebras says its fast inference suits agentic AI workflows

Cerebras chief strategy officer Andy Hock says inference speed will decide which platforms win agentic AI, since one request can spawn a family of coordinating agents.

He says Cerebras' Wafer-Scale Engine, more than 50 times bigger than any prior chip, delivers 10 to 15 times faster inference than other platforms, per the company's claim.

Hock says enterprises are shifting from "token maxing" toward value per token, and that Cerebras' users find the most value in its fastest tokens.