Power shortages, water shortages, labor shortages, and land grabbing! The US data center construction boom faces obstacles

The data center construction boom driven by the artificial intelligence revolution is facing practical obstacles. From power grid capacity and water resources to skilled labor shortages and land competition with residential development, this infrastructure surge led by tech giants like Microsoft, Alphabet, Meta, and Amazon is encountering multiple constraints. Execution risks are rising, potentially dampening market optimism about AI investment returns.

Goldman Sachs analyst Brian Singer’s recent conversation with Mark Monroe, former lead chief engineer of Microsoft’s data center development team, revealed three key bottlenecks: Power remains the most urgent near-term constraint, water resource pressures are pushing the industry toward more energy-intensive cooling technologies, and a shortage of skilled workers could become the next hurdle. Monroe warns that by 2030, the U.S. will need to add over 500,000 manufacturing, construction, operations, and transmission workers to meet data center power demands.

Meanwhile, tech giants are acquiring land across the U.S. at unprecedented prices, directly squeezing residential development. According to the Wall Street Journal, Amazon spent $700 million in November last year to purchase land in Virginia from Stanley Martin, a developer that had bought parts of the same land just a few years earlier for only around $50 million. In Northern Virginia, rural land that once sold for a few thousand dollars per acre now commands over $3 million, making it impossible for residential developers to compete.

Whether this construction boom can continue depends on core assumptions in the current macro narrative and tech stock valuations—that data center investments will translate into measurable productivity gains and support long-term growth. However, supply chain bottlenecks, infrastructure constraints, and community opposition are accumulating, which could cause overly optimistic expectations to fall short.

Power Bottlenecks Are the Most Urgent

Power supply remains the most critical near-term constraint for data center deployment. Monroe notes that while cloud computing and AI inference workloads typically need to be close to end users—leading to crowded markets experiencing power shortages—AI training workloads are less sensitive to location and are migrating to remote areas with ample electricity.

Flexible load management could free up some capacity, but adoption is hindered. A Duke University study shows that if data centers accept an average annual load reduction of 0.25% (99.75% uptime), an additional 76 gigawatts of capacity could be added—equivalent to about 10% of the U.S. peak demand. A 0.5% reduction (99.5% uptime) could add 98 gigawatts. However, Monroe states that industry risk-averse culture hampers the widespread adoption of such measures—repeatedly switching IT equipment unsettles operators and may require stronger financial or regulatory incentives.

Behind-the-Meter onsite generation is becoming an expensive temporary solution. Although only a small percentage of new data centers are applying for onsite generation, Monroe emphasizes that these are typically large facilities with significant power needs. These systems mainly deploy simple-cycle natural gas generators, costing 5 to 20 times more than grid power. Yet, considering the enormous profitability of large AI data centers, deploying onsite generation remains economically feasible to kick-start projects. Monroe says that the ultimate goal for data centers with onsite generation is to connect to the grid within three years, at which point they will either relocate, integrate and sell power back to the grid, or decommission onsite assets.

Water Resource Constraints Come with Energy Costs

Community, regulatory, and chip technology advancements are pushing the industry toward more water-efficient but more energy-intensive cooling methods. Monroe states that as community and regulatory pressures intensify, the industry is shifting from traditional high-water evaporative cooling toward designs that use less water—especially among large-scale operators.

This transition results in significant efficiency losses. Monroe explains that moving to closed-loop and waterless cooling systems could raise the power usage effectiveness (PUE) from a best-in-class 1.08 to between 1.35 and 1.40, meaning energy costs could jump from 8% for evaporative systems to 35-40%. While innovations like direct chip liquid cooling and high-temperature water cooling can enable more efficient heat transfer in certain locations, colocation data centers—due to their diverse customer base and the need to finalize cooling architecture early—may continue to rely on chilled water systems. Monroe notes that although evaporative cooling’s share in overall data center cooling may decline, demand for chilled water units will still grow substantially over the next decade due to overall capacity expansion.

Skilled Worker Shortage Becomes the Next Barrier

Monroe warns that a shortage of skilled workers could become the next major hurdle for data center deployment. Unlike typical industrial buildings, data centers require specialized electrical and mechanical systems, making electricians and pipefitters critical.

Industry groups are partnering with technical universities and colleges to develop training programs to fill this gap and are trying to engage students as early as middle school to make tech careers more attractive. According to Goldman Sachs estimates, by 2030, the U.S. will need to net over 500,000 additional workers in manufacturing, construction, operations, and power transmission to meet all power deployment needs.

Tech Giants’ Land Grab Drives Prices Higher

Data center developers are acquiring land at prices far exceeding those for other uses, directly impacting residential development. According to the Wall Street Journal, five years ago, Stanley Martin CEO Steve Alloy was preparing to develop 516 new homes in Bristow, Virginia, when he noticed that surrounding land was being aggressively purchased by tech giants like Microsoft and Google. By November last year, the company sold some of that land—purchased just a few years earlier for over $50 million—for $700 million to Amazon, marking one of the largest land deals in U.S. history.

Northern Virginia has become the global capital of data centers. The region benefits from open land, expanding power infrastructure, and dense fiber networks laid during the internet boom. Loudoun County is home to the “Data Center Alley,” a cluster of facilities, with major tech companies expanding south along Interstate 95 into Prince William County.

Land prices have soared, making it impossible for residential developers to compete. In Northern Virginia, developers are offering landowners up to $1 million per acre. Some rural land that sold for a few thousand dollars per acre just a few years ago now exceeds $3 million. Near Chicago, in Elk Grove Village—a data center hub—Stream Data Centers bought and demolished a 55-unit residential complex in 2024 for nearly $1 million per building to build three data centers totaling 2.1 million square feet. Along U.S. Highway 67 near Dallas, land that sold for $20,000 to $40,000 per acre three years ago has now surged past $350,000 in some areas. Residential land developer Scott Finfer comments, “There’s no way residential builders can match these numbers.”

Looking ahead, the key question is whether the U.S. can sustain the rapid growth in data center capital expenditures, given that these projects are deeply embedded in macro narratives and tech stock valuations. The investment thesis assumes ongoing construction will translate into measurable productivity improvements and support years of growth. However, execution risks—such as core component availability, grid access, and supply chain bottlenecks—could slow progress and undermine overly optimistic expectations.

Risk Warning and Disclaimer

Market risks exist; investments should be made cautiously. This article does not constitute personal investment advice and does not consider individual user’s specific investment goals, financial situation, or needs. Users should evaluate whether any opinions, viewpoints, or conclusions herein are suitable for their circumstances. Any investment based on this information is at their own risk.