As agentic AI surges, CPUs and air-cooled infrastructure move to the fore

Agents are turning up the heat for enterprises, turning air-cooled AI infrastructure into a boardroom priority.

While GPUs have largely dominated the AI conversation, CPUs are increasingly coming into focus. Agentic AI relies on constant orchestration, data processing, and decision-making between tasks — work that plays to the CPU’s strengths, not the GPU’s. The math alone is staggering, and agentic AI is only making it more demanding, according to Robert Hormuth (pictured, right), corporate vice president of architecture and strategy for the Data Center Solutions Group at Advanced Micro Devices Inc.

“Agentic AI is very goal-oriented. You ask it to achieve a goal, and it’s going to use every tool, every software, everything in the book to go achieve that goal,” Hormuth said. “One of those might be querying a GPU to get a complex math equation, but the rest of it is planning, checking, kicking off tools, verification and iterating. There’s just a lot of work to do around the complex math that the foundational model might return. There’s just a lot more going on in the world of agentic. That’s one of the big kickers right now for CPU demand.”

Hormuth and David Schmidt (left), vice president of PowerEdge product management at Dell Technologies Inc., spoke with theCUBE’s Dave Vellante and Gemma Allen at Dell Technologies World 2026, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed the growing role of CPUs in agentic AI workloads and the industry shift toward power-efficient, air-cooled AI infrastructure. (* Disclosure below.)

The shift towards air-cooled AI infrastructure

As Dell Technologies rolled out its next-generation PowerEdge servers, it found many enterprise customers are still not prepared to adopt liquid cooling in their data centers. Liquid cooling is typically driven by the extreme power demands of high-density AI racks, according to Schmidt.

“You’re in a rack environment where maybe you have power constraints along the lines of 15 kilowatts a rack, maybe 20 kilowatts a rack. We get that feedback quite a bit,” he said. “It was really important to us to have a wide range of air-cooled systems, and then it’s a simple economic conversation.”

That economic conversation is completely transforming how enterprises think about AI expansion — not as an infrastructure overhaul, but as a capacity decision within existing constraints. For many, air-cooled systems are proving to be the pragmatic path forward, enabling AI deployment without the cost or complexity of rebuilding from the ground up, Schmidt explained.

“You’ll see a 13-to-1 consolidation ratio in our latest generation. That’s 13 just core rack servers down to one,” Schmidt said. “It’s going to free up resources — power resources, rack space — that you can then deploy for these agentic workloads.”

The stakes of that transition extend well beyond rack space, Hormuth noted. Enterprises that delay risk being outpaced by rivals already deploying AI to generate new revenue and unlock new opportunities. The era of AI, he said, is not one that rewards hesitation.

“The era of AI is not waiting, Hormuth said. “This is not one of these market trends that you can sit back and debate for six months or nine months to enter, because your competition is going to go fast.”

Here’s the complete video interview, part of SiliconANGLE’s and theCUBE’s coverage of Dell Technologies World 2026:

(* Disclosure: AMD sponsored this segment of theCUBE. Neither AMD nor other sponsors have editorial control over content on theCUBE or SiliconANGLE.)

Photo: SiliconANGLE