What Does the Future of Data Centers Look Like?

Data centers are no longer just back-office infrastructure; they are the backbone of the global economy. The race to build more of them—faster, smarter, and cleaner—is reshaping how capital flows through the energy and real estate sectors alike.

Demand is not slowing down. Microsoft, Amazon, Google, and Meta collectively committed over $200 billion in capital expenditures in 2024, much of it aimed at expanding data center capacity. AI workloads are the accelerant. A single AI training run for a large language model can consume as much electricity as hundreds of homes use in a year. Multiply that across thousands of clusters, and the infrastructure math gets serious fast.

So what does the next generation of data centers actually look like? The answer involves smarter design, cleaner energy, and a fundamentally different approach to where and how these facilities get built.

The Demand Signal Is Unmistakable

Cloud adoption plateaued—briefly—during the post-pandemic pullback. Then generative AI arrived, and capacity constraints became front-page news. Hyperscalers are now leasing data center space faster than developers can deliver it. In Northern Virginia, the world's largest data center market, vacancy rates have dropped below 1% in some submarkets. Power constraints, not land or capital, have become the primary bottleneck.

The shift from "build it and they will come" to "they're already here, and we can't build fast enough" happened in less than three years.

This imbalance is pushing development into secondary and tertiary markets—places like Omaha, Columbus, San Antonio, and rural areas of the Carolinas—where power is available and land is cheaper. For infrastructure investors and land sellers, that geographic diffusion is significant. Sites that seemed too remote two years ago are now fielding serious inbound interest.

Modern Data Center Design Isn't What It Used to Be

The monolithic, fortress-style data center—500,000 square feet, single campus, locked in for 20 years—is giving way to something more flexible. Modular data center design is becoming the dominant architecture, and for good reason: it lets operators scale incrementally rather than committing massive capital to capacity that might sit idle for years.

Modular builds typically come in standardized units—often 1 to 5 MW blocks—that can be stacked, replicated, and deployed across multiple sites. This approach compresses construction timelines and reduces the financial risk of overbuilding. It also makes edge deployments viable, bringing compute closer to where data is generated rather than routing everything through a centralized hub.

Cooling is the other major design frontier. Traditional air cooling struggles to handle the thermal density of modern GPU clusters. Liquid cooling—direct-to-chip and immersion variants—is moving from experimental to mainstream. Some new facilities are being designed with liquid cooling as the default, not an afterthought. The infrastructure implications are significant: different floor load requirements, different plumbing, and different facility management expertise.

Clean Energy Is No Longer Optional

For most of the industry's history, data center operators cared about power reliability above everything else. Cost came second. Environmental considerations were, at best, a marketing checkbox.

That calculus has changed. Corporate sustainability commitments, investor ESG mandates, and increasingly, regulatory pressure are making clean energy data centers a baseline expectation rather than a differentiator. Microsoft's agreement to restart the Three Mile Island nuclear plant is the most dramatic example—a tech company essentially backstopping the revival of nuclear energy to secure long-term clean power for its facilities.

The most forward-thinking operators are treating energy procurement as a core competency, not a facilities management problem.

Solar-plus-storage combinations are increasingly common in data center power strategies. On-site generation paired with battery storage can help facilities manage grid demand charges, maintain uptime during grid disruptions, and credibly claim around-the-clock clean energy coverage. For developers in markets with strong solar resources—the Southwest, Southeast, and parts of Texas—this creates a genuine competitive advantage in attracting hyperscaler tenants who need to hit their own sustainability targets.

The infrastructure convergence here is worth watching. Solar farms, battery storage facilities, and data centers are increasingly being co-developed on the same land parcels. That creates complex interconnection, zoning, and land use questions—but also significant opportunity for landowners and developers who understand all three asset classes.

Where AI Is Changing the Infrastructure Equation

AI isn't just driving demand for more data centers; it's changing what those facilities need to do internally.

GPU-optimized facilities—sometimes called AI factories—have fundamentally different specifications than traditional enterprise data centers. Power density per rack has jumped from 5-10 kW in conventional deployments to 30, 60, even 100+ kW for AI training clusters. That changes everything: structural requirements, cooling systems, power distribution architecture, generator sizing, and UPS configurations.

Infrastructure investors and developers who treat AI data centers as simply "bigger versions" of what came before are going to make expensive mistakes.

On the operations side, AI is also being deployed within data centers to optimize cooling efficiency, predict equipment failures before they happen, and dynamically route workloads to minimize energy consumption. This is a meaningful leverage point: cooling accounts for roughly 30-40% of a data center's total energy use, so even modest optimization gains translate directly to margin improvement at scale.

The Investment Case, and What the Numbers Actually Mean

Data center investment has attracted significant institutional capital, but the risk-return profile varies considerably depending on where in the stack you're investing.

Hyperscale build-to-suit deals—where a developer constructs a facility to a tenant's specifications under a long-term lease—offer relative safety. The tenant is creditworthy, the lease term is long (often 10-20 years), and the revenue is contracted. The tradeoff is compressed yields, often in the 5-6% cap rate range for stabilized assets in primary markets.

Speculative development in emerging markets carries more risk but also more upside. The developers who built in Columbus and Phoenix five years ago, when those markets looked like long shots, are now sitting on assets that have appreciated dramatically.

For land sellers and smaller investors, the more immediate opportunity is upstream: sites with available power, fiber access, and favorable zoning are genuinely scarce. A site that can support 50+ MW of critical load, with a clear path to interconnection, is a different kind of asset than it was three years ago. Understanding how to position, permit, and transact those sites requires infrastructure-specific knowledge that general commercial real estate brokers often lack.

Real Challenges That Don't Get Enough Attention

The bullish narrative around data centers is well-established. The friction points deserve equal attention.

Permitting and interconnection timelines are brutal. In many markets, getting a new data center from site acquisition to energized facility takes four to seven years—and that's without major setbacks. Grid interconnection queues in PJM, ERCOT, and other major ISOs are backed up for years. Some developers are now building behind-the-meter generation specifically to avoid the interconnection queue, but that introduces its own complexity.

Water is an underappreciated constraint. Traditional air-cooled data centers use significant quantities of water for evaporative cooling. In drought-stressed regions—the Mountain West, parts of the Southwest—this is becoming a genuine community and regulatory issue. Facilities that can operate with near-zero water usage have a meaningful long-term advantage in those markets.

Local opposition is growing. Communities that once competed aggressively to attract data centers are increasingly scrutinizing the deals: large tax incentives, high power consumption, and relatively few permanent jobs. Northern Virginia has seen organized opposition to new data center development in some jurisdictions, and similar dynamics are emerging elsewhere.

What Comes Next

The data center sector is not going to slow down. The demand drivers—AI adoption, cloud migration, digitization of physical infrastructure—are structural, not cyclical. But the sites, designs, and energy strategies that made sense five years ago are being stress-tested by requirements that didn't exist then.

The developers, investors, and landowners who succeed in this next phase will be the ones who understand that a data center is now inseparable from its energy infrastructure—and that the most valuable assets in this space sit at the intersection of power, land, and connectivity. Getting all three right, in the right market, at the right time, is harder than it looks. It's also where the real value gets created.

Explore the InfraSale Marketplace for investment opportunities in data centers and more!

[INTERNAL LINK: data center investment]

[INTERNAL LINK: clean energy strategies]

[INTERNAL LINK: AI in infrastructure]