Data Centers See Enterprise AI Demand Creating New Opportunities

FORT LAUDERDALE, Fla., March 24, 2025 — While hyperscale data centers capture headlines with gigawatt-sized projects, mid-market data center operators are finding substantial growth opportunities serving enterprise artificial enterprise demands, according to industry leaders at the MetroConnect conference here on Feb. 24.

"There's a real gap when you talk about demand in that 10 to 50 megawatt range," said Scott Willis of DartPoints. "We're all excited about it because there's a real opportunity to take advantage of that in the market."

The panel, moderated by Jim Grice of Akerman, a law firm focused on how operators are adapting to power constraints and meeting the demands of evolving workloads driven by AI that require increased cooling density.

Brett Lindsey, CEO of ark data centers, noted how demand has scaled up significantly. "Most of us historically would have viewed a 250 kilowatt customer as kind of an anchor," Lindsey said. "I think that's kind of the bottom of what we're seeing now. Most customers are looking for 500 kilowatts, 1 megawatt and beyond."

Contrary to speculation about companies repatriating workloads from cloud providers, panelists indicated this trend remains limited. Instead, they're seeing enterprises developing new AI applications that require different infrastructure than what cloud platforms can efficiently provide.

"Enterprise customers are coming up with new things that they need to put in place that don't work well in the cloud," Lindsey said. "Especially as they're thinking about AI, they're uncomfortable with the idea of having their most secure and precious IP sitting somewhere that isn't in a cage that no one can get into."

Unlike hyperscale facilities that can be purpose-built for specific customers, mid-market operators face the challenge of designing facilities flexible enough to accommodate various workloads while supporting advanced cooling technologies needed for AI computing.

"You've got to design that facility with a catcher's mitt that's big enough to catch all the workloads," Willis explained. "I've got to be able to design and configure that environment so that I'm compelling enough to capture a broad enough part of the workload, whether it's higher ed[ucation], healthcare or manufacturing."

Jim Buie, CEO of ValorC3 Data Centers, emphasized the need for adaptable designs that can support both traditional and AI computing. "It's got to be a hybrid of air and liquid cooling. You have to be adaptable to the enterprise needs," he said.

The panel also addressed how power constraints are affecting their segment differently than hyperscale operators. While they aren't seeking gigawatt-sized power contracts, these operators still face challenges securing 10-50 megawatts in established markets.

Doug Recker, President of Duos Edge AI, described acquiring gas turbines to provide temporary power until utilities can build permanent infrastructure. "The main reason we did this is because of the focus on enterprises," Recker said. "Our niche is the 10 to 30 megawatt data center where they need 2 megawatts of power until the utility builds out. Some utilities, you're lucky it's six months, some are two years."

According to Jeffrey Moerdler, a member at Mintz, a law firm, the industry has experienced significant consolidation, creating market opportunities. "It used to be if you wanted to get 250 kilowatts to a megawatt in any primary or secondary market, you had a few options," Moerdler said. "Today you're lucky if you have two options in any market."

As to whether enterprise customers still prioritize sustainability goals amidst the AI rush, panelists indicated it varies widely by customer, with public companies facing more scrutiny. "Speed to market triumphs over sustainability in my experience, for the time being," Moerdler concluded.