Construction delays hit 40% of AI data centers as power infrastructure and labor shortages stall expansion

Data Center Construction Delays Threaten AI Expansion Timeline

Satellite imagery from geospatial analytics company SynMax reveals that nearly 40% of US AI data centers slated for completion in 2026 are at risk of missing their deadlines by more than three months 1

1

. The Financial Times analysis, which cross-checked satellite imagery against public statements and permit documents compiled by industry research group IIR Energy, shows major projects from Microsoft, Oracle, and OpenAI facing significant setbacks 5

5

. These data center construction delays emerge as tech companies pour hundreds of billions of dollars into building facilities that require as much electricity as hundreds of thousands of US homes, raising concerns about how quickly vast AI spending can generate returns.

Labor Shortages and Supply Chain Bottlenecks Hamper Progress

Interviews with more than a dozen industry executives highlighted chronic labor shortages as a critical constraint, with construction teams lacking enough specialist workers such as electricians and pipe fitters to meet demand across multiple projects 1

1

. Two construction executives working on OpenAI-linked projects confirmed insufficient specialist workers to handle the buildout as companies race to construct increasingly large and complex facilities . Remote locations are pushing labor costs up as much as 30%, while workers move between projects seeking better pay. Doug O'Laughlin, president of SemiAnalysis, noted that the concentration of projects in some regions intensifies competition, with OpenAI effectively competing with itself as workers shift between sites 5

5

. Supply chain bottlenecks compound these challenges, with jet engines for power generation facing delivery timelines stretching from 2028 to 2030 for orders placed in 2025 2

2

.

Power Infrastructure Emerges as Critical Bottleneck

The substantial power demand requirements represent a massive energy bottleneck as utility companies struggle to build enough generation capacity and expand power infrastructure necessary to deliver electricity 1

1

. Hyperscalers are racing to build facilities drawing at least 1 gigawatt of electricity—roughly a nuclear reactor's output—but strained grid capacity and equipment shortages are causing significant delays 5

5

. In Santa Clara, California, home of Nvidia, nearly 100 megawatts of newly constructed datacenter capacity sits empty awaiting power connections that may not arrive for years 3

3

. Tariffs on imported Chinese equipment such as transformers have worsened the situation for Silicon Valley's AI ambitions 1

1

. The grid connection bottleneck has pushed wait times to 12 to 15 years in Britain, where demand for connections leapt 460% in the first six months of 2025 4

4

.

OpenAI Projects Show Significant Delays Despite Company Denials

A 1,200-acre, 1.4-gigawatt campus in Shackelford County, Texas, being developed for Oracle to equip for OpenAI, illustrates the scale of delays facing AI infrastructure 2

2

. Expected to host 10 buildings with delivery in the second half of 2026, satellite imagery from early April shows only six plots cleared for construction with just one showing development signs 5

5

. SynMax estimates one building could possibly be delivered by December, but a more realistic timeline pushes completion to late 2027 2

2

. Another OpenAI-linked project, a 1.2-gigawatt site in Milam County, Texas, shows only one building under construction as seen from space 2

2

. Despite these indicators, OpenAI maintained that "our historic data center build-out is on schedule and we will accelerate from here," while Oracle stated each data center is "moving forward on time" 2

2

.

High Energy Costs Drive AI Workloads Overseas

One in five UK firms have already moved AI workloads abroad due to high energy costs, according to the "Land, Power, Compute" report published by CUDO Compute, which surveyed 700 organizations across the US and Europe 3

3

. Energy costs are particularly acute in Britain, which has some of the highest prices in the developed world, with a third of UK organizations saying energy costs limit their ability to scale 3

3

. Among AI-first companies, nearly 32% are considering moving workloads overseas due to power costs, threatening the UK government's ambitions to make Britain a global AI leader 3

3

. Nearly one in five survey respondents have already scaled back AI workloads due to energy costs, with more than a fifth claiming energy bills account for more than a third of their AI infrastructure budgets 3

3

. The US is rated as the most attractive location for new AI cluster provisioning, with 75% of respondents viewing it positively, followed by India, while Eastern Europe scores highly at 58% 3

3

.

Powered Land and Zombie Projects Reshape Real Estate Market

The AI gold rush has spawned a whole new industry around data center development, upended land valuations, and created a logjam in the lengthy queue for grid connections 4

4

. Across Britain, owners of industrial sites, speculative investors, property developers, and even farmers are positioning their land to cash in on billions in planned spending, with construction analytics group Barbour ABI tracking plans for 119 data centers on sites ranging from disused car plants to former hotels 4

4

. The concept of "powered land"—plots with existing power generation or high-voltage grid connections—has become highly valuable, exemplified by the Wilton International site in Teesside, where land left dormant by chemical industry decline now attracts data center interest 4

4

. However, many sites have no power, leading to an explosion in grid connection applications that Britain's National Energy System Operator identified as including 140 data centers representing about 50 gigawatts of capacity 4

4

. So-called "zombie projects"—requests from landowners without power, planning permission, or potential end users—are clogging the system and delaying viable projects 4

4

.

Community Resistance and Permitting Issues Add Regulatory Friction

Growing resistance from communities across the US to data centers adds another layer of complexity, with Virginia, known as "the data center capital of the world," seeing public opinion turn sharply against new development 1

1

. In a recent poll, a majority of Virginians described concerns about land use, environmental impacts, and negative effects on home electricity bills, as data center development can pressure utility companies to raise rates for all local and regional customers 1

1

. Maine legislators recently became the first to pass an 18-month moratorium on approvals for new data centers requiring more than 20 megawatts of power, though Governor Janet Mills must still decide whether to veto the legislation 1

1

. Permitting issues extend beyond community opposition, with some AI hyperscalers turning to on-site generators like turbines as alternative power sources, but these require EPA permits adding regulatory friction 2

2

. The Trump administration responded to energy cost concerns in March 2026 by announcing a Ratepayer Protection Pledge from major tech companies, though the agreement lacks meaningful legal enforcement or practical implementation 1

1

.