Data Centers: What's Next for Infrastructure?
Explore key trends reshaping data center infrastructure and energy needs in today's evolving landscape.
The power grid wasn't built for this. When utilities designed transmission infrastructure decades ago, they planned for factories, hospitals, and office parks—not facilities that can draw 100+ megawatts from a single campus and run 24 hours a day, 365 days a year without interruption. Now those utilities are scrambling, and the gap between what data centers need and what the grid can actually deliver is becoming one of the defining infrastructure problems of our era.
This isn't a distant problem. It's happening in substations right now, in utility commission hearings, and in the boardrooms of every major hyperscaler on the planet.
The Grid Wasn't Ready — And Neither Was Anyone Else
Data center development accelerated faster than almost any infrastructure category in history. The AI compute boom didn't just increase demand linearly—it broke the model entirely. Training a large language model requires sustained, massive energy draws. Inference—actually running these models at scale—requires the same. Multiply that across hundreds of facilities, and you understand why utilities are issuing queue interconnection backlogs stretching three to five years in some markets.
The core tension is simple: data centers need power now, and the grid operates on timelines measured in years, sometimes decades.
Northern Virginia—the world's largest data center market by capacity—has become the clearest case study in what happens when demand outpaces infrastructure. Dominion Energy has publicly acknowledged the strain. New data center projects in Loudoun County face unprecedented scrutiny not because of zoning objections, but because there literally isn't enough substation capacity to support them. Local officials who once competed aggressively for data center investment are now asking harder questions about grid resilience for existing residents and businesses.
This is what a real infrastructure constraint looks like. It's not a talking point—it's a queue.
Energy Needs Are Reshaping the Entire Supply Chain
A hyperscale facility today—think 200MW to 500MW campus—consumes roughly as much electricity as a mid-sized American city. That number is climbing. As AI workloads intensify and GPU density per rack increases from 10kW to 30kW, then to 100kW with liquid cooling, the energy footprint per square foot of data center space keeps rising even as operators chase efficiency gains.
Power Usage Effectiveness (PUE) ratios have improved dramatically over the past decade—Google and Meta regularly report PUEs below 1.1, meaning nearly all energy drawn goes to compute rather than cooling. But efficiency gains can't outrun raw demand growth. You can optimize a 100MW facility to perform like a 95MW one. You can't optimize away the fact that you need 500 of them.
Renewable energy procurement has moved from a marketing checkbox to a genuine site selection variable—and in some cases, a regulatory prerequisite. Microsoft, Amazon, and Google have all made aggressive renewable commitments, but fulfilling them requires available clean generation capacity near where the data centers actually sit. That's driving co-location of solar farms and battery storage with data center campuses, and it's pushing development into new geographic markets where land is cheap, power is available, and the grid has headroom.
The Midwest and Southeast are seeing this play out in real time. States like Indiana, Georgia, and Ohio are attracting serious data center investment partly because their utilities have capacity—and partly because land acquisition costs in saturated markets like Northern Virginia or Silicon Valley have become prohibitive.
The Challenges That Don't Make the Press Releases
Infrastructure limitations get discussed. The regulatory dimension gets far less attention, and it deserves more.
Interconnection reform at the federal level—specifically FERC's Order 2023—is attempting to overhaul how generation projects get connected to the grid. The intent is to speed up the queue and bring more clean energy online faster. The reality is that implementation is uneven, utilities are still adapting, and a faster interconnection process still takes years. For data center operators who need power on 18-to-24-month development timelines, this is a structural mismatch.
Water rights are another constraint that surprises people outside the industry. Traditional air-cooled and evaporative cooling systems use substantial amounts of water—a single large data center can consume millions of gallons annually. In water-stressed regions across the American West, this is already triggering local opposition and regulatory review. Liquid cooling adoption (direct-to-chip and immersion cooling) addresses part of this, but the transition is expensive and not yet universal.
The operators who navigate these challenges successfully aren't the ones with the best technology—they're the ones who invested early in government relations, utility partnerships, and community engagement.
Zoning and local permitting have also grown more contentious. Communities that once offered tax abatements and fast-track permitting to attract data centers are reassessing the bargain. Data centers generate enormous property tax revenue but relatively few jobs per megawatt of power consumed. When residents start experiencing reliability issues on the grid, that calculus shifts quickly.
Future-Proofing: Where Smart Capital Is Moving
The operators and investors paying attention right now are making a few specific bets that are worth understanding.
Behind-the-meter power generation is gaining serious traction. Rather than waiting years for utility interconnection, some developers are pursuing on-site generation—natural gas peakers, fuel cells, small modular reactors (SMRs) in the longer term, and combinations of solar plus battery storage that can carry critical load through grid interruptions. This isn't a complete solution, but it changes the math on site selection and timeline.
Battery storage, specifically grid-scale BESS deployments, is becoming standard infrastructure at serious data center campuses—both as backup power and as a mechanism to participate in grid services markets, which can meaningfully offset operating costs. A 100MW data center with 4-hour battery storage isn't just resilient; it's also a grid asset that utilities want to work with.
Geographically, the smart money is looking beyond the traditional clusters. The Carolinas, the upper Midwest, and parts of Texas (despite ERCOT's volatility concerns) are attracting serious due diligence. International markets—particularly in Northern Europe where renewable power is abundant and fiber connectivity is strong—are absorbing overflow demand from U.S. operators.
On the technology side, the shift to liquid cooling isn't optional for next-generation AI infrastructure—it's mandatory. Rack densities required by NVIDIA's H100 and Blackwell architectures can't be managed with air cooling at scale. Facilities that can't retrofit or weren't designed for liquid cooling are already becoming functionally obsolete for the most valuable workloads. Investors evaluating data center assets should treat cooling infrastructure with the same rigor they apply to power capacity.
What Comes Next — And Who Needs to Act
The data center infrastructure challenge isn't going to resolve itself through incremental improvements. It requires deliberate action from multiple directions simultaneously: utilities investing in transmission and substation upgrades, developers pursuing creative power procurement strategies, policymakers accelerating interconnection reform, and communities engaging early rather than reacting late.
For landowners and developers watching these trends, the opportunity is specific: sites with existing power access—particularly those adjacent to transmission infrastructure or with direct utility relationships—carry a premium that wasn't there five years ago and will only grow. A 200-acre parcel with a nearby substation and favorable zoning isn't just land anymore. It's a data center site.
For investors, the infrastructure layer beneath the data center matters as much as the facility itself. Power purchase agreements, transmission rights, grid interconnection status—these are the variables that determine whether a project actually gets built on schedule and whether it performs as underwritten.
The leaders Inskeep referenced who are "looking for what's next" are asking the right question. The answer isn't a single technology or a single market. It's the recognition that data center infrastructure is now critical national infrastructure—with all the complexity, long timelines, and stakeholder politics that designation implies. The organizations that internalize that reality and plan accordingly will capture the opportunity. The ones waiting for the market to simplify are going to be waiting a long time.
Call to Action
Ready to explore the future of data centers and infrastructure? Visit InfraSale Marketplace to discover opportunities and insights.