The US artificial intelligence infrastructure boom, estimated at more than $10 trillion, is starting to crack due to construction delays, higher financing costs and growing resistance from local communities. The Financial Times describes the trend as a moment when the boom is “starting to crack“.
According to data from the analytics company RegLayer, at least 47 major data center projects in the country have been delayed, canceled or blocked. Analysts at Bernstein estimate that about 35% of planned US projects are realistic. More than 1,000 cities have discussed moratoriums or other restrictions on building AI infrastructure.
Financing is getting more expensive
The sector’s development depends on so-called neoclouds — companies that raise funding, build data centers and rent computing power. But higher interest rates, inflation and a deteriorating bond market have raised doubts about whether the revenues of companies like OpenAI and Anthropic will be enough to cover the costs of computing infrastructure.
Against this backdrop, investors have begun to divest themselves of debt instruments related to AI infrastructure. Shares of cloud provider CoreWeave have fallen by almost 25% in a week. SB Energy, owned by SoftBank, and Firmus, backed by Nvidia, have postponed or frozen their initial public offerings. OpenAI has also postponed its own IPO.
Stargate project also faces delays
The problems also affect specific construction sites. Delays have been reported at several Oracle sites related to the Stargate project, which OpenAI is developing for building data centers. The project is estimated at $500 billion.
In the event of longer delays, contracting authorities may be given the opportunity to abandon contracts if the sites are not completed on time. Contractors, in turn, may request additional financing. Obtaining building permits and connecting to the electricity grid also become more difficult.
Investments remain record highs
Despite the problems, the scale of investments remains unprecedented. According to an estimate by Columbia Business School economist Steen van Nyverburg, US spending on AI infrastructure between 2025 and 2032 could exceed $10 trillion, or about 3.6% of gross domestic product annually.
Google, Amazon, Microsoft and Meta have invested a total of $1.1 trillion since the AI boom began in 2023. The tech giants plan to invest $745 billion in data centers, advanced chips and computing power. According to the Financial Times, some of this investment will only pay off if companies like OpenAI and Anthropic manage to raise enough money to pay for the computing power they use.
Sources: rbc.ru, ft.com, axios.com