By: Bill Tierney -- The ProLift Rigging Company

The AI boom has triggered an unprecedented flood of capital, with global hyperscale spending projected at approximately $700 billion. Yet, across the United States, the physical reality of building these facilities is colliding with systemic bottlenecks.

The primary issue facing modern data center construction has shifted from silicon chip scarcity to industrial supply chain limits and utility grid capacity. Industry research has found that between 30% and 50% of the 16 gigawatts (GW) of U.S. data center capacity scheduled to come online are facing severe operational delays or outright cancellation.

In this article, ProLift, a full-service industrial rigging, crane, transportation, and warehousing company, examines this tension. As the technological appetite for megawatts outpaces the nation's industrial capacity to deliver them, the structural execution of data center buildouts is undergoing a profound strain.

The Scale of the AI Boom

To maintain a competitive edge in generative AI and large language model (LLM) development, tech giants are investing capital at unprecedented levels. The world’s largest cloud and technology providers are expanding their capital expenditures toward a record-breaking $25.2 billion to build the infrastructure required to meet their computing needs. After adjusting for inflation, total AI investment is now greater than the cost of the Manhattan Project, the International Space Station (1984–2011), the Apollo Program, and the construction of the interstate highway system combined.

Increasing Campus Sizes

This spending is translating into a dramatic transformation of the physical landscape of American data centers. In layman's terms: they’re bigger and more energy-intensive than ever before. A typical AI-focused data center consumes as much electricity as 100,000 households. This is already an enormous number, but the largest hyperscale campuses under construction today will consume 20 times as much.

Managing New Mega Builds

This shift to more and larger facilities has disrupted traditional infrastructure models. Finance, engineering, and construction teams now must account for how to manage the expense, weight, and volume of the equipment required to manage these loads. This gear includes dense server racks, industrial liquid-cooling chillers, and massive backup diesel generators.

Structural engineers have had to redesign the floor loading and spatial layouts of modern facilities. So while constructing the physical shell of a data center is a relatively fast process, the task of energizing these massive structures is an increasingly daunting one — turning the focus from real estate execution to the broader electrical supply chain.

The Delay Problem

Recent project data shows the widening gap between digital ambition and physical reality. According to the Sandstone Group, nearly half of all planned U.S. data centers are slipping behind schedule.

Why is this happening? According to Allianz Research, a combination of prolonged permitting procedures, a shortage of roughly 439,000 skilled construction workers, and critical electrical equipment deficits are major reasons nearly half of the projected capacity planned for the near-term faces delays.

A prime example of this structural friction occurred when Oracle Corporation and OpenAI terminated plans to expand a flagship AI data center campus in Abilene, Texas. Developed by Crusoe Energy, the 1,000-acre site was to be a core component of the highly publicized "Stargate" infrastructure project. The tentative expansion aimed to scale the facility’s capacity from an already massive 1.2 GW up to an even larger 2.0 GW. However, the parties involved ultimately shelved these expansion plans after prolonged negotiations stalled over complex financing structures and long-term utility infrastructure timelines.

This isn’t a unique circumstance. Across major data center hubs, projects that were once fast-tracked are now being paused as developers face delayed timelines provided by utility companies and equipment manufacturers, as well as public opposition.

Why Projects Are Stalling

There are several interconnected forces stalling the construction of planned data centers in the U.S. These include power availability, critical equipment backlogs, supply chain vulnerabilities, and regulatory friction.

Power Constraints and Grid Interconnection Queues

Securing a physical connection to a high-voltage transmission grid has become a significant obstacle for developers. Regional transmission organizations (RTOs) across the country are facing massive backlogs of energy generation and industrial load requests.

In data center-heavy markets, such as Northern Virginia, Phoenix, and Dallas, the wait time to clear an interconnection queue and receive a firm power commitment now routinely stretches between four and seven years. Utility providers simply can’t build new substations or string high-voltage transmission lines fast enough to keep pace with the compressed schedules of tech sector builds.

The Transformer and Equipment Bottleneck

Even if a utility company approves a power allocation, developers face a shortage of the physical machinery necessary to distribute that electricity. Critical components such as large-scale transformers, high-voltage switchgear, and circuit breakers are increasingly sold out globally.

According to industrial supply chain data, the market lead times for critical grid components have reached historic highs:

Generator step-up transformers: Lead times have surpassed 160 weeks (approximately 3.1 years), up from an average of 143 weeks just two years ago.

High-voltage circuit breakers: Lead times have climbed to 125 weeks, compared to a historical baseline of 77 weeks.

Global Supply Chain Dependencies

A heavy reliance on foreign manufacturing exacerbates the equipment crisis. Decades of industrial offshoring have left the United States with limited domestic capacity to manufacture heavy electrical steel and large power transformers. Currently, an estimated 80% of U.S. power transformer demand is met by imports, as is 50% of the distribution transformer supply. With global demand surging simultaneously, U.S. project managers find themselves caught in aggressive bidding wars, driving up procurement costs and introducing massive schedule uncertainties. And the shortage of hundreds of thousands of skilled workers is another compounding factor in these delays.

Permitting and Local Resistance

Beyond the technical and mechanical shortcomings, data center builds are encountering a growing wave of regulatory and social friction. Municipalities that once welcomed data centers for their tax revenues are pulling back due to concerns over water consumption for cooling, localized noise pollution from backup generators, air pollution from numerous gas-powered generators, and the impact of massive industrial structures on residential utility rates. With over 100 active data center moratoria and more under consideration, the administrative path to a breaking-ground permit has grown increasingly difficult.

Operational and Economic Consequences

The systemic stalling of these projects is altering the trajectory of broader technology deployment. The most obvious consequence is a measurable deceleration in the rollout of advanced AI models. Because computing clusters cannot be powered uniformly, tech companies are facing a structural ceiling on how quickly they can scale their next-generation training clusters.

This constraint is also creating a landscape of regional inequality across the industrial real estate sector. Data center development is clustering in locations that possess immediate, unallocated power reserves or regulatory environments designed to accelerate utility hookups, such as West Texas. However, as these primary regions become saturated, developers are forced to push into secondary and tertiary markets that lack the robust transmission infrastructure required to support heavy industrial loads. This increases the complexity and cost of data center builds.

The average cost per megawatt has increased to $11.3 million globally, with AI-focused facilities costing $15 million to over $20 million per MW. U.S. construction spending reached $85.3 billion, nearly doubling since 2024. When a project is delayed by two to three years because of equipment backlogs, the carrying costs of the land, structural steel, and specialized labor erode the long-term financial viability of the entire asset.

Outlook through 2030

The structural tension between digital demand and physical execution will likely remain the defining characteristic of the data center sector for the next few years. Wood Mackenzie projects U.S. data center power capacity to grow from its current baseline of roughly 24 GW up to 110 GW by 2030. This means data centers will command up to 40% of the entire U.S. power equipment market by the end of the decade. In 2020, this number was just 2%.

These backlogs for heavy equipment and utility interconnections are unlikely to be resolved quickly. While technology companies are actively pursuing creative workarounds — including funding private nuclear energy partnerships, deploying on-site natural gas and renewable energy, and investing in domestic manufacturing facilities — these infrastructure solutions require multi-year development timelines.

Software and computing capabilities won’t taper off due to a lack of coding breakthroughs. Technological progress will simply be bound by the physical laws governing construction, logistics, and energy. Heavy industrial execution, the capacity of global supply chains, and strategic expertise will continue to be major determining factors in the success or failure of facility construction.