Nearly Half of Planned US Data Centers Face Delays or Cancellation as Electrical Equipment Shortages Stall the AI Buildout
Sightline Climate estimates 30-50% of US data centers slated for 2026 will be delayed or canceled, with transformer shortages and Chinese import dependency creating bottlenecks that billions in AI spending cannot solve.
Overview
The race to build AI infrastructure in the United States is hitting a wall that money alone cannot fix. An analysis by Sightline Climate, reported by Bloomberg, estimates that between 30 and 50 percent of US data centers planned for deployment in 2026 will be delayed or canceled. The bottleneck is not computing chips or capital but the mundane electrical equipment needed to actually power these facilities: transformers, switchgear, batteries, and circuit breakers.
The finding arrives as Alphabet, Amazon, Meta, and Microsoft are collectively expected to spend more than $650 billion on AI capacity this year, according to Tom’s Hardware.
What We Know
Data centers representing at least 12 gigawatts of power consumption were announced for completion in the US before the end of 2026. However, only about one-third of that capacity is currently under active construction, according to the Sightline Climate analysis cited by Bloomberg. The remaining projects sit in pre-production stages where delays in any single electrical component can halt an entire build.
“If one piece of your supply chain is delayed, then your whole project can’t deliver,” Andrew Likens, energy and infrastructure lead at Crusoe Energy Systems, told TechRadar. “It is a pretty wild puzzle at the moment.”
The core problem is power delivery infrastructure. Batteries, transformers, and circuit breakers each represent less than 10 percent of a data center’s total construction cost, but it is impossible to bring a facility online without them, as Tom’s Hardware reported. Lead times for high-power transformers have expanded from a typical 24 to 30 months before 2020 to as long as five years today, far exceeding the 18-month deployment cycles that AI companies typically target.
The supply chain for these components depends heavily on imports. Transformer imports from China surged from fewer than 1,500 units in 2022 to more than 8,000 units through October 2025, according to trade data cited by Tom’s Hardware. China accounts for over 40 percent of US battery imports and roughly 30 percent of transformer and switchgear categories. Canada, Mexico, and South Korea have become the largest high-power transformer suppliers, but despite a decade of reshoring initiatives, US domestic manufacturing capacity for electrical equipment remains insufficient.
What We Don’t Know
The full impact of current and proposed tariffs on Chinese electrical equipment imports is unclear. While the US government has worked with industry to reduce Chinese import dependency, TechRadar noted that domestic electrical manufacturing buildout shows minimal progress. It remains to be seen whether tariff pressures will accelerate reshoring efforts or simply compound the existing shortages.
The outlook for future years is also uncertain but concerning. Among data centers slated to open in 2027, only about 6.3 gigawatts of computing infrastructure are under construction versus 21.5 gigawatts announced, according to Bloomberg. For 2028 through 2032, 37 gigawatts of capacity have been planned but only 4.5 gigawatts are actively being built.
Beyond supply chain issues, community opposition is emerging as an additional barrier. Concerns about noise, water consumption, and environmental impact have slowed permitting in some regions, as Tom’s Hardware reported, though the extent to which this will affect the overall buildout timeline remains unquantified.
Analysis
The data center supply crisis exposes a structural mismatch in the AI boom: the software layer moves at the speed of venture capital, but the physical infrastructure layer moves at the speed of industrial manufacturing. Companies can announce hundreds of billions of dollars in AI spending, but they cannot will transformers into existence faster than foundries can produce them.
The heavy reliance on Chinese imports for critical electrical components also introduces geopolitical risk into AI infrastructure planning. With trade tensions ongoing and tariff policies in flux, data center developers face a dual squeeze: they cannot build fast enough with current supply, and the policy environment may further constrain the import channels they depend on.
For hyperscalers, the immediate consequence is that announced capacity will come online later and cost more than planned. For the broader economy, the bottleneck means the AI computing capacity that enterprises and startups are counting on may not materialize on the timelines that current investment projections assume.