The U.S. construction market is experiencing a stark divergence in 2026, with AI data center development surging while traditional commercial, warehouse, and healthcare construction sectors weaken. Despite soaring demand for AI computing infrastructure, total data center construction capacity declined for the first time in five years, constrained by permitting delays, zoning restrictions, and power procurement challenges. This paradox reveals a market fundamentally reshaping around artificial intelligence infrastructure while legacy sectors struggle.
Electricity demand is the primary driver of data center expansion. U.S. data centers consumed 176 terawatt-hours (TWh) of electricity in 2023—approximately 4.4% of national electricity demand. This represents a dramatic acceleration from 2018, when data centers consumed 76 TWh (1.9% of national consumption). Researchers at Lawrence Berkeley National Laboratory project that electricity demand from U.S. data centers could rise to between 325 and 580 TWh by 2028, equivalent to 6.7-12% of total U.S. electricity consumption.
This trajectory reflects what researchers describe as a structural shift in power demand tied to the expansion of AI compute workloads and large-scale hyperscale facilities.
Despite this explosive demand, total capacity of data centers under construction declined for the first time in five years in 2025. This counterintuitive contraction occurred even as demand for AI computing capacity remained seemingly insatiable and vacancy rates remained low.
The decline reflects a geographic redistribution rather than absolute market weakness. While construction boomed in Chicago and the Dallas area, those gains were offset by declines around Silicon Valley, Northern Virginia, and elsewhere. Legacy markets such as California and Oregon are expected to lose more than half of their relative market share, with Texas set to become the country's leading data center market within the next three years.
Permitting, zoning, and power procurement delays are the primary constraints limiting data center construction despite strong underlying demand. These regulatory and infrastructure barriers have created a bottleneck that prevents developers from meeting market demand.
Data center development is extraordinarily capital-intensive. Benchmarks from Cushman & Wakefield estimate that developing one megawatt of data center critical load capacity in the United States costs between $9.3 million and $15 million, averaging roughly $11.7 million per megawatt across surveyed markets.
This capital intensity means that physical damage is not only disruptive but also slow to reverse. Recovery requires construction lead times, constrained supply chains, and complex commissioning of power and cooling systems—precisely the kinds of frictions that make outages in traditional infrastructure sectors economically costly.
A 500-megawatt AI campus can generate approximately $30 to $40 million per year in tax revenue from electricity consumption alone. System benefit charges and tax revenues tied to these projects represent one of the largest new funding sources for local infrastructure in many communities.
These substantial tax revenues create powerful economic incentives for communities to support data center development, even when environmental concerns exist. If structured correctly, large data centers do not strain the grid but instead help fund the infrastructure upgrades the grid needs.
TeraWulf, a Maryland-based company, is developing Lake Mariner Data—a campus of multiple data centers at the former Somerset Power Plant in Niagara County, New York. The project currently employs 1,000 construction workers.
This project exemplifies the strategic site selection approach: TeraWulf is placing data centers in areas that have ample generation capacity and/or brownfield areas that lost or retired large industrial sites, rather than in big-city locations that already stress the power grid. The Niagara County location benefits from existing industrial infrastructure and adequate electrical generation capacity.
The data center market is undergoing significant geographic consolidation:
Data center projects generate significant temporary employment. TeraWulf's Niagara County facility currently employs 1,000 construction workers.
However, most data center projects do not employ many people once they're up and running. The employment benefits are concentrated in the construction phase, with limited permanent job creation once facilities become operational.
The software engineers and array of professions needed to support data centers once built are expected to absolutely shift the job market. This suggests that while permanent employment at individual facilities may be limited, the broader ecosystem of data center support services will create significant employment opportunities.
Public opposition to data center construction is substantial and growing. A Quinnipiac University poll found that 65% of Americans oppose building an AI data center in their community, with only 24% supporting one being built (survey of 1,397 U.S. adults).
A Harvard/MIT poll presented a more nuanced picture, finding that 40% of people supported building a data center in their area, with 32% opposed when asked about different industrial facilities generally. Notably, more people would rather have an e-commerce warehouse than a data center.
Two-thirds of respondents in the Harvard/MIT poll were worried that a new data center in their region would nudge electricity prices higher. Additional concerns include pollution, noise, and air quality impacts.
The proposed data center in Monterey Park illustrates community opposition intensity. The facility—the size of four football fields—is located close to homes and is expected to consume three times the energy used by the entire city. Residents say the project will raise their electricity bills and increase noise and air pollution.
Maine is set to pass legislation making it the first state to ban the construction of new data centers into 2027. Eleven states have brought similar moratorium legislation.
At the federal level, Senator Bernie Sanders proposed legislation to enact a federal moratorium on data center construction. Representatives Bernie Sanders and Alexandria Ocasio-Cortez recently introduced a bill to pause all new data center construction until federal guardrails and safeguards are instituted for workers, communities, and the environment.
The cost of performing inference at GPT 3.5-level capability fell by more than 280-fold between November 2022 and October 2024, dramatically lowering the marginal cost of deploying AI applications at scale.
By 2030, inference will surpass model training as the dominant AI data-centre workload, representing more than 50% of all AI compute and roughly 30-40% of total global data-centre demand, according to McKinsey & Company.
These dynamics ensure continued strong demand for data center capacity despite regulatory headwinds, as the economics of AI deployment continue to improve.
When a sector begins to shape grid planning, generation investment, and transmission timelines, its continuity becomes a macroeconomic concern rather than a purely private operational matter.
The International Energy Agency has described data centres as an increasingly consequential driver of national electricity systems, particularly in the United States, where they are expected to account for a significant share of incremental electricity demand growth through 2030.
The scale and capital intensity of data center infrastructure is prompting policy discussions about treating AI infrastructure as critical infrastructure—similar to electricity grids, ports, and major telecommunications exchanges.
The GCC (Gulf Cooperation Council) data center construction market is projected to reach USD 5,836.1 million by 2034, exhibiting a growth rate (CAGR) of 17.60% during 2026-2034.
The combination of declining inference costs, increasing workload volumes, and the structural shift in power demand suggests sustained long-term growth in data center construction, despite near-term permitting constraints.
The Cato Institute has proposed consumer-regulated electricity as a policy solution that would balance concerns about data center demand with infrastructure needs. This approach would free up a new market for sophisticated, well-resourced firms constructing data centers to be supplied by new, off-grid utilities able to dynamically adapt and innovate, while ensuring that none of the power required by data centers is subsidized by existing consumers.
Industry representatives argue that communities should embrace data center development possibilities. When structured correctly, large data centers do not strain the grid but help fund the infrastructure upgrades the grid needs. The substantial tax revenues and infrastructure funding generated by these projects represent significant benefits to local communities.
The 2026 construction market reveals a fundamental reshaping of capital allocation toward AI infrastructure. While data center construction faces regulatory and permitting constraints that have reduced total capacity under construction for the first time in five years, the underlying demand remains strong and is projected to accelerate through 2028 and beyond.
This divergence reflects both technological transformation (declining AI inference costs driving deployment) and geographic redistribution (capital flowing from legacy tech hubs to regions with better power infrastructure and regulatory environments). Traditional commercial, warehouse, and healthcare construction sectors are experiencing weakness as capital concentrates in AI infrastructure development.
The tension between strong economic incentives (substantial tax revenues, infrastructure funding, employment) and community opposition (electricity price concerns, environmental impacts) will likely shape regulatory outcomes and project timelines through 2027 and beyond.
