US AI Data Center Boom Put to the Test as Funding Woes, Construction Delays Fuel Concerns

Bernstein: "Only 35% of Planned Projects Are Likely to Be Completed"
More Than 47 Major Projects Delayed or Halted This Year

The boom in artificial intelligence (AI) data center construction in the United States is being put to the test. As financing pressures mount, construction delays continue, and local communities push back, questions are growing over whether announced project plans can actually be realized.

According to the Financial Times (FT) on Sunday, October 18 (local time), regulatory intelligence firm RegLayer counted more than 47 large data center projects in the United States that had been delayed, withdrawn, or suspended this year.

Bernstein estimated this month that only about 35% of planned data center construction in the United States is likely to be completed. An FT investigation found that more than 1,000 communities were pursuing measures to restrict data center construction.

Financing is a key factor in whether projects move forward. Companies known as "neoclouds," which provide cloud services dedicated to AI, must cover enormous facility investment and semiconductor purchasing costs. Rising interest rates and a weakened bond market have made borrowing more difficult, adding to companies' financial burdens. This week, investors began reassessing the industry's financing risks, with CoreWeave-related bond prices falling and its share price dropping sharply.

Construction delays also increase the risk of contract cancellations. If data centers are not completed on time, their customers, including Big Tech companies and AI research labs, may terminate their contracts. Scheduling changes and cost pressures have also emerged in projects involving OpenAI and Anthropic. Some developers now face having to cover cost overruns themselves or raise additional capital.

The gap between planned investment and current project execution capabilities is also a concern. The FT cited estimates that U.S. spending on AI infrastructure could exceed $10 trillion from 2025 to 2032. However, some companies have little to no track record in data center construction, while others have secured only a fraction of the funding needed to operate their planned facilities. Constraints related to grid connections and permits, as well as opposition from local residents, are also making projects more difficult to advance.

AI companies have signed contracts with inexperienced operators to secure computing resources in advance. But analysts say that as market conditions change, speed alone is no longer enough to justify investment.

This content was produced with the assistance of AI translation services.