Anxiety over America’s build-out of data centers has reached Wall Street in recent days, kindling fears that the artificial intelligence boom is unsustainable. A better way to view the sell-off: as a chance to recalibrate financial risks and to highlight the unnecessary burdens slowing development.
Share prices for some of the most highly leveraged U.S. tech companies have taken a hit amid fears that inflation and high interest rates could undermine planned projects. CoreWeave, which has been borrowing heavily to construct data center capacity for renting out to AI developers, led the sell-off, losing 36 percent of its market capitalization in the past month.
The fear is that massive capital expenditures won’t deliver hoped-for payoffs. Some investors are buying insurance-like derivatives to protect themselves in case companies behind the build-out default on their heavy debts.
This is the market working as it should. Investors are pricing in risks — such as higher rates, poorly managed balance sheets and higher-than-expected costs — that could make some projects financially infeasible. That may disappoint AI boosters, but these market signals are important. Speculation is typically fueled by easy access to capital, so higher rates insulate the broader economy from overexuberant speculation.
The bigger threat is misguided government policy. The Financial Times reports that at least 47 large data center projects have either been delayed, withdrawn or blocked this year, many because of resistance from local or state officials. More than 1,000 communities in the United States have sought to impose moratoriums or other restrictions on AI infrastructure.
Much of that resistance is based on wildly overstated fears, such as how much water data centers use, and generalized nervousness about how AI will change the nature of work. Other fears, such as rising electricity prices, can be ameliorated with agreements clarifying that tech companies must pay for their own energy needs. The sector’s hunger for power is already helping drive overdue grid modernization.
But the political movement to slow down or block the growth of data centers could have adverse consequences for the entire economy. By some estimates, AI-related goods accounted for as much as three-quarters of all global trade growth in the first quarter of 2026 and upward of a third of U.S. economic growth.
Financial risk does not diminish the structural utility of technological advances. Many railroads failed, but the track remained. Ditto for companies in the 1990s that laid fiber-optic cable, which continues to generate value.
If CoreWeave has borrowed too much, creditors and shareholders will bear the consequences. If Microsoft or Meta wants another gigawatt of electricity, they should pay the cost of producing and transmitting it. But neither investors nor consumers benefit when zoning boards, permitting regimes and utility bottlenecks prevent otherwise economically viable projects from being built.
The post Let markets, not government, decide whether there’s an AI bubble appeared first on Washington Post.
