by Gina Narcisi, CRN

AI is putting a renewed limelight on the data center after years of cloud-first strategies for many enterprises, according to Cisco executives and channel partners.

"Private data centers are back," declared Cisco CEO Chuck Robbins during his keynote to customers and partners at Cisco Live 2025 earlier this month. "We could talk forever about all the pressure we felt on the transition to cloud. I think that was around the same time someone said: 'Why would you ever buy an ethernet port again?' And now most of you are looking at a very balanced hybrid approach," he told customers and partners.

Cloud has provided efficiency and value for a variety of workloads, but the demanding nature of AI workloads will require businesses to build their own private infrastructure to go with their public infrastructure, Robbins said.

An architectural shift is underway as AI reshapes network and computing infrastructure, according to a recent survey of more than 8,000 senior IT and business leaders conducted by Sandpiper Research & Insights on behalf of Cisco. The survey found that 71 percent of respondents said their data centers can't yet meet today's AI demands, and 88 percent plan to expand capacity – on-prem, in the cloud, or both.

Power and compute are two of the biggest constraints to AI infrastructure, said Jeetu Patel, Cisco's president and chief product officer.

"We don't have enough power. And in fact, what's happening is data centers are being built where the power resides," he told reporters ahead of Cisco live 2025.

For Cisco Gold partner WWT, the solution provider giant's data center business was up 60 percent year over year last year after struggling for the last several years, according to Neil Anderson, vice president of cloud, infrastructure, and AI solutions for WWT. 

This "cloud repatriation" is happening because cloud migrations didn't always deliver promised benefits for certain workloads, customers often faced higher-than-expected costs, and the fact that sensitive AI data often belongs in controlled environments, Anderson said.

"We call it cloud smart, not cloud first. Put the right apps in the cloud, put the right apps in your data center," he said. "We do see some repatriation of applications back into the data center and now there's massive investments back into the data center."

AI applications in particular lend themselves to running in private data centers because they often consist of a business's intellectual property or sensitive data, Anderson said.

"Most of our customers are saying: 'I'm going to build this in my data center. It may take me a while because of power and cooling requirements, but that's what my strategy is going to be,'" he said.

Read full article