Agentic workflows are making speed a condition for whole classes of AI applications to exist, and the fastest tokens will determine which platforms win the next phase of AI.
Cerebras Systems Inc., which raised $5.5 billion in its May 2026 initial public offering, built its business on making processors much faster, according to Andy Hock (pictured), chief strategy officer of Cerebras.
“We didn’t build it big because big looks cool. We built it big because AI wants big chips,” Hock said. “AI needs a lot of compute, but it needs that compute to be close together, and it also needs that compute to be close to memory.”
Hock spoke with theCUBE Research’s John Furrier at the AI Data Pipeline Forum, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed agentic workflows and the importance of inference speed. (* Disclosure below.)
Why agentic workflows make speed a requirement
Cerebras’ Wafer-Scale Engine is more than 50 times as large as previous conventional processors, and the company claims its inference performance is 10 to 15 times faster than that of competing GPU-based platforms, Hock noted. In agentic workflows, that means more agents working together.
“One question from you or from me into an agentic system doesn’t just spawn one intelligent agent going down and doing some work or answering a question for us, but it spawns a family of daughter agents that are all doing different work and all coordinating together,” Hock said. “That’s where speed really makes the difference.”
Cerebras reported its first-quarter 2026 results in June. Over the past two months, foundation labs and enterprises have moved away from “token maxing” and toward getting the most value out of each token, Hock noted.
“They’re making decisions based on that value maximization paradigm on models, on datasets and on infrastructure,” Hock said. “On the infrastructure side, to bring it home to Cerebras, what we’re seeing is that our end users are actually finding the greatest value out of the fastest tokens.”
Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of the AI Data Pipeline Forum:
(* Disclosure: ScaleFlux sponsored this segment of theCUBE. Neither ScaleFlux nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)