Cerebras bets the fastest tokens will win AI’s next phase

Cerebras Systems, a company that raised $5.5 billion in its May 2026 initial public offering, is betting that the fastest processors will determine the winners of the next phase of AI. Andy Hock, Cerebras' chief strategy officer, explained that the company built its processors to be extremely powerful because AI requires a significant amount of compute power.

Hock emphasized the importance of speed during an interview on theCUBE Research’s John Furrier show at the AI Data Pipeline Forum, stating that the Cerebras Wafer-Scale Engine is more than 50 times larger than any previous chip and offers 10 to 15 times faster inference than other platforms.

In the context of agentic workflows, which involve multiple agents working together to accomplish a task, the speed of the processors becomes crucial. Hock explained that a single query to an agentic system doesn't just result in one intelligent agent processing the information; instead, it spawns a "family of daughter agents" that all work concurrently and coordinate their efforts. This interdependence highlights the significance of speed in AI applications.

As foundation labs and enterprises shift away from "token maxing" and focus on maximizing the value derived from each token, the need for faster processors becomes increasingly apparent. Hock noted that Cerebras' customers are finding the greatest value in the company's fast tokens, which allows them to achieve more with each token processed.

The video interview, part of SiliconANGLE's and theCUBE's coverage of the AI Data Pipeline Forum, can be viewed in its entirety. TheCUBE is sponsored by ScaleFlux, which has editorial control over the content.