US data-center expansion delayed by transformer, switchgear, and battery shortages from China despite AI investment commitments in 2026

Alphabet, Amazon, Meta and Microsoft are pledging more than $650 billion in 2026 to expand artificial intelligence (AI) capacity. But nearly half of U.S. data-center projects this year are being delayed or canceled, largely due to shortages of key components—especially transformers, switchgear and batteries—areas where supply is dominated by China.

Power-component shortages slow AI buildouts

Bloomberg data cited in the report indicates that about 12 gigawatts of data-center capacity were expected to be operating in the United States by 2026. However, only about a third of that amount is currently under construction.

The central constraint is the lack of critical electrical infrastructure components needed not only to power AI data centers, but also to expand grid capacity. Electrical infrastructure accounts for less than 10% of total data-center costs, yet it is described as essential: delays anywhere along the power chain can halt the entire project.

“If one link in the supply chain is delayed, the whole project cannot be completed.”

The report attributes this view to a Crusoe employee involved in building OpenAI’s large data center in Texas.

Transformers are the most severe bottleneck

The most serious bottleneck is high-voltage transformers. Lead times for such transformers were 24–30 months before 2020, but have since stretched to as long as five years. That timing is particularly problematic for AI data centers, where deployment cycles are under 18 months.

Philippe Piron, CEO of GE Vernova’s Electrification division, is quoted saying: “AI companies want something in under 18 months, but the market cannot deliver.”

Shifting suppliers, but China remains central

To address the shortage, companies are turning to global markets. Canada, Mexico and South Korea have become major suppliers of high-voltage transformers for U.S. AI data centers.

Still, dependence on China extends beyond transformers. China accounts for more than 40% of U.S. imported batteries, while its share in some transformer and switchgear segments is near 30%.

Benjamin Boucher, a senior analyst at Wood Mackenzie, said: “With insufficient domestic manufacturing capacity, people are almost forced to seek export markets.”

Trade policy tensions complicate procurement

The report highlights a tension with U.S. trade policy. In December, the Trump administration stated that “there will be only one winner, and that may be the United States or China.” While the administration wants the U.S. to win, its America First approach includes tariff barriers intended to curb imports, placing data-center companies in a difficult position.

Over the past decade, U.S. efforts to bring production home have not produced meaningful domestic capacity growth, leaving companies reliant on China despite tariffs and potential national-security risks.

Implications for AI investment and capacity

The report argues that if shortages of transformers, switchgear and batteries are not resolved, even trillions of dollars in AI investment may not translate into real AI capacity. Deployment, it says, will depend on the availability of power infrastructure rather than capital or computing hardware.

In this framing, the AI competition between the U.S. and China is intensifying, yet the U.S. still relies on China for components viewed as “small” relative to chips and computing—components that can determine whether data-center projects can be completed on time.