Power shortages, water shortages, labor shortages, and land grabs! The U.S. data center construction boom faces obstacles.

```

The data center construction boom driven by the artificial intelligence revolution is running into real-world obstacles. From power grid capacity and water resources to skilled labor, as well as competition for land with residential development, this infrastructure frenzy led by tech giants such as Microsoft, Alphabet, Meta, and Amazon is facing multiple constraints. Execution risks are rising and may drag down the market’s optimistic expectations for AI investment returns.

Goldman Sachs analyst Brian Singer’s recent conversation with Mark Monroe, former Chief Engineer of Microsoft’s advanced data center development team, revealed three key bottlenecks: Power remains the most urgent near-term constraint, water stress is forcing the industry to adopt more energy-intensive cooling technologies, and a shortage of skilled workers may become the next threshold. Monroe warns that, by 2030, the United States will need to add over 500,000 manufacturing, construction, operations, and transmission and distribution workers to meet data center power demand.

Meanwhile, tech giants are buying land at unprecedented prices across the U.S., directly squeezing residential development. According to the Wall Street Journal, Amazon spent $700 million last November to purchase parcels in Virginia that residential developer Stanley Martin acquired just years ago for just over $50 million. In Northern Virginia, rural land that once sold for tens of thousands of dollars per acre is now priced at over $3 million, pricing residential developers out of the market.

Whether this construction frenzy can continue relates to the core assumption behind today's macro narrative and tech stock valuations—namely, that data center investment can translate into measurable productivity gains and support years of growth. But supply chain bottlenecks, infrastructure constraints, and community opposition are accumulating and could cause overly optimistic expectations to fall short.

Power Bottleneck Most Urgent

Power supply remains the most critical near-term constraint facing data center deployment. Monroe points out that while cloud computing and AI inference workloads typically need to be close to end users—which leads to power shortages in crowded markets—AI training workloads are not sensitive to geography and are now migrating to remote areas with abundant power.

Flexible load management could unlock some capacity, but adoption is limited. A Duke University study shows that if data centers accepted a 0.25% annual average load reduction (99.75% uptime), an additional 76 GW of load could be added, equivalent to 10% of total U.S. peak demand; with 0.5% reduction (99.5% uptime), 98 GW could be added. Monroe says, however, that the industry's inherent risk-averse culture is an obstacle to widespread adoption of these solutions—frequently switching IT equipment on and off makes operators uneasy and may require stronger financial or regulatory incentives.

On-site generation (Behind-the-Meter) has become an expensive temporary solution. While only a single-digit percentage of data centers under construction apply on-site generation, Monroe stresses these are usually large-scale data centers, so the power impact is still significant. These solutions mainly deploy natural gas simple cycle generators, at a cost five to twenty times higher than grid electricity. However, considering the huge profitability of large AI data centers, adopting on-site generation to accelerate project launches is still economically feasible. Monroe says the ultimate goal for data centers using on-site generation is to connect to the grid within three years, at which point they will either move workloads to other centers, integrate and sell power back to the grid, or retire on-site generation assets.

Water Constraints Bring Energy Costs

Community, regulatory, and chip technology advancements are pushing the industry toward more water-efficient but more energy-consuming cooling technologies. Monroe points out that as community, regulatory, and technical pressures intensify, the industry is shifting from traditional, high-water-use evaporative cooling methods to designs using less water, especially among hyperscalers.

This transition leads to notable energy efficiency losses. Monroe notes that moving to closed-loop and waterless cooling systems may raise the Power Usage Effectiveness (PUE) from an optimal 1.08 to 1.35–1.40, meaning energy overhead jumps from 8% for evaporative systems to 35%-40%. Although innovative solutions such as direct-to-chip liquid cooling and high-temperature water cooling can enable efficient heat transfer at more locations, colocation data centers—with their diverse customer base and need to choose cooling architecture early—may still stick with chilled water plant designs. Monroe says that while the share of evaporative cooling in total data center cooling may decline, the demand for chilled water plants will still surge dramatically in the next decade due to total data center capacity growth.

Shortage of Skilled Workers Becomes the Next Hurdle

Monroe warns that the shortage of skilled workers may become the next threshold for data center deployment. Data centers differ from typical industrial buildings in their need for specialized electrical and mechanical systems, which makes electricians and plumbers crucial for construction.

Industry organizations are working with technical universities and colleges to develop training programs to fill this gap, and are trying to reach students as early as middle school to make tech trades a more attractive career path. According to Goldman Sachs estimates, by 2030, the U.S. will need to net-increase over 500,000 workers in manufacturing, construction, operations, and power transmission and distribution in order to deploy all the required capacity.

Tech Giants’ Land Rush Pushes Up Prices

Data center developers are buying land at prices far above other uses, directly impacting residential construction. According to the Wall Street Journal, when Stanley Martin CEO Steve Alloy planned to develop 516 new homes in Bristow, Virginia five years ago, he noticed surrounding land was being snapped up by tech giants like Microsoft and Google. By last November, the company sold parcels it had bought for just over $50 million several years earlier to Amazon for $700 million, one of the largest raw land transactions in U.S. history.

Northern Virginia has become the global data center capital. The area offers wide open land, growing power infrastructure, and a dense fiber network from the dot-com bubble. Loudoun County is home to “Data Center Alley,” and the world’s largest tech companies are pushing south into Prince William County along Interstate 95.

Soaring land prices make it impossible for residential developers to compete. In Northern Virginia, developers are sending landowners letters offering as much as $1 million per acre. Rural land that once sold for tens of thousands per acre now fetches over $3 million. Near Chicago’s Elk Grove Village data center hub, Stream Data Centers in 2024 bought and demolished a 55-home subdivision for nearly $1 million per house to build three data centers totaling 2.1 million square feet. Along U.S. Highway 67 near Dallas, land that sold for $20,000–$40,000 per acre three years ago has shot up to over $350,000 in some places. Residential land developer Scott Finfer remarked, "It’s simply not possible for home builders to make those numbers work."

Looking forward, the key question is whether the U.S. can sustain the surge in data center capital spending, given these projects are now deeply embedded in both the macro narrative and tech stock valuations. The investment thesis assumes ongoing construction will translate into measurable productivity gains and years of growth. But, ultimately, execution risk comes down to key inputs and infrastructure such as core components, grid access, and related supply chain bottlenecks, all of which could slow construction and frustrate overly optimistic expectations.

Risk Disclosure and DisclaimerThe market carries risks, and investments should be made cautiously. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are applicable to their specific circumstances. Investments made based on this are at your own risk. ```