While the headlines focus on multi-gigawatt hyperscale projects, enterprise-focused developers are seeing a dramatic shift in customer expectations and deployment models.

Learn how operators are evolving to meet growing enterprise AI demands, navigate power constraints, and adapt their infrastructure strategies in a rapidly changing environment.

Jim Grice, Partner - Akerman (moderator)
Jeffrey Moerdler, Member - Mintz
Brett Lindsey, CEO - Ark Data Centres
Scott Willis, CEO - Dartpoints
Doug Recker, President - Duos Edge AI
Jim Buie, CEO - Valor C3 Data Centers

As hyperscale projects capture the spotlight, experts at the recent Metro Connect conference emphasised a rising need for facilities in the 10-50MW range—demand they say is underserved.

Brett Lindsey, CEO of Ark Data Centres, said that while 250kW used to be a strong anchor tenant, most enterprise customers are now requesting deployments of 500kW to 1MW and above.

Scott Willis, CEO of edge colocation operator DartPoints, described this shift as “the opportunity” for mid-market platforms: “We're perfectly positioned to capture that 10 to 50 megawatt demand gap.”

“Its always easy to talk about the hypers[calers], but if you dive in to each one of the conversations that we're having, and you see the energy and the enthusiasm in that 10 to 50 area, there's a real opportunity to take advantage of that in the market, and there's a real need.”

Jeffrey Moerdler, a member at the law firm Mintz, recalling a time when 5- 25MW builds were considered large: “Today, a 100MW data centre is almost small by hyperscale standards.”

“We're in the brave new world where the hyperscalers and the AI users have pushed the size of the deal.”

He pointed out that M&A activity has consolidated many mid-sized players, leaving gaps that regional developers are now filling.

“The formerly public companies and the wannabe public companies have scooped up many of the mid-sized and larger providers, creating a big gap in the industry,” Moerdler said. “It used to be that if you wanted to get 250 kW to a megawatt in any primary or secondary market, you had a few options in that market. Today, you're lucky if you have two options in any market.”

Rather than simply seeing a repatriation of workloads from the cloud, mid-sized data centre operators are witnessing a surge in new enterprise demand, particularly around AI and high-performance computing (HPC).

Panellists at Metro Connect were quick to clarify that while some companies are bringing select applications out of public cloud environments, the larger driver is fresh deployments that require dedicated, secure, and often high-density infrastructure.

“Customers are asking for things that don't work well in the cloud," said Lindsey. “Especially with AI, there's real discomfort about placing their most valuable intellectual property in an environment they don't fully control."

Willis noted that while hyperscale projects often capture attention, “the real energy is in the 10 to 50MW range,” where enterprises are deploying AI inference workloads that require significant compute density but not the hyperscale footprint.

These evolving needs are forcing developers to rethink facility designs. Traditional enterprise data halls, historically supporting racks at 5-8kW, must now accommodate densities of 30kW, 50kW, or more, often incorporating liquid cooling technologies.

Jim Buie, CEO of Valor C3 Data Centres highlighted the need for adaptability: “You have to build facilities that flex between air and liquid cooling, because one size no longer fits all.”

Mintz’s Moerdler pointed out a parallel benefit: in dedicated facilities, enterprises can work directly with operators to tailor configurations, sidestepping the anonymity and rigidity often associated with cloud providers. As AI adoption grows, he suggested this close collaboration could become a decisive advantage for regional data centre operators.

The availability of power is rapidly becoming one of the most critical factors shaping data centre strategy. While mid-sized operators are not requesting gigawatt-scale connections, the panel agreed that power constraints are a growing challenge, even for 10 to 50MW builds.

Power is a challenge, no question,” Willis said. “In some markets, we’re dealing with substation constraints. In others, it's a race against time — can you align your project delivery with the utility’s timelines?”

Doug Recker, president of Duos Edge AI, which builds modular edge data centres, described how his team are addressing the issue by deploying natural gas turbines as a temporary bridge until utilities can deliver permanent capacity.

“We acquired 850MW of gas turbine assets because enterprises can't afford to wait two or three years for power,” Recker revealed.

Valors’s Buie highlighted that for enterprise customers signing 10-year contracts, predictability around power costs has become a top priority. Operators, therefore, must increasingly offer behind-the-metre solutions or work with energy providers to guarantee long-term stability.

Moerdler pointed out that power availability is also reshaping site selection: “Today, it’s not just about finding land and fibre. If you can’t secure power quickly, the project won't happen.”

“People are going anywhere they can to get capacity and segment their load. Citibank has been in South Dakota with a lot of its data processing, data centre operations for 20 or 30 years. Not everything has to be in the primary or secondary markets, and about a third of the deals I'm working on are behind the metre, off the grid, with a whole bunch of other non-traditional power sources.”

Deals like the ones Moerdler is helping shape are driving new opportunities for developers willing to build in secondary and tertiary markets, where land is cheaper and creative energy solutions can be deployed.

As hyperscaler expansion strains traditional hubs like northern Virginia, the enterprise sector is leading a more flexible, decentralised wave of growth—one that demands power readiness as much as fibre connectivity.

With enterprise workloads becoming more demanding and diverse, the traditional approach to data centre design is evolving rapidly. Operators can no longer rely on a one size fits all model; instead, facilities must be engineered for flexibility from day one.

“Our customers are asking for a broader range of deployments, from legacy air-cooled systems to liquid-cooled high-density racks,” Lindsey explained. “The goal is to support everything from low-density office IT to HPC clusters requiring direct-to-chip cooling, all under one roof.”

Dartpoints’ Willis emphasised the importance of designing for a wide catchment of workloads: “You can't just build for hyperscalers. You need to create environments that can attract a broad cross-section of enterprise needs, whether it’s universities deploying AI research clusters or healthcare institutions rolling out new data-intensive applications.”

Buie noted that hybrid designs are becoming the norm, blending air and liquid cooling strategies to allow for easy adaptation as customer needs shift.

“Liquid was once a bad word in the data centre industry,” he said. “Today, it’s increasingly a necessity, and operators need to ensure their facilities can evolve over time without major retrofits.”

Moerdler added a critical design consideration: floor loading. High-density deployments require structural planning to prevent overloading traditional data centre floors.

“If you're building for 30-50kW racks, your floors need to handle significantly higher weights. These things must be baked into designs from the beginning.”

The bottom line: adaptability is now a critical competitive advantage. Facilities that offer greater flexibility, without sacrificing efficiency or reliability, will be best positioned to meet the demands of the AI-powered enterprise era.

While capital is pouring into hyperscale-focused platforms, the Metro Connect panellists agreed that there is still significant investor appetite for mid-market data centre opportunities, provided investors are properly educated.

“There is plenty of capital available,” Willis said. “But the KPIs that investors care about in the mid-market are different, and you have to take them through a different education journey.”

Lindsey noted that investors left out of the billion-dollar hyperscale projects are looking for ways to gain exposure to digital infrastructure, and mid-sized enterprise facilities offer an attractive proposition.

“You don't need to be raising a billion dollars at a time to play in this space,” he said. “There are investors who want something they can get their arms around.”

Ultimately, the panel agreed that capital constraints are not the issue—rather, the challenge lies in clearly articulating the different risk-return profiles, development timelines, and tenant dynamics that define the enterprise segment.