Four Bottlenecks Shake AI Data Center Boom, NYT Says

The delays in U.S. data center construction reveal underlying vulnerabilities in the AI industry's rapid expansion. The reliance on debt-fueled growth, particularly by 'neocloud' operators like CoreWeave, is being tested as interest rates rise and credit weakens. CoreWeave's reported $72 billion debt and speculative grade rating, with bond coupon rates over 15%, illustrate the financial strain on companies heavily invested in AI infrastructure.

Beyond financing, practical constraints such as power shortages and community resistance are also significant hurdles. Microsoft's withdrawal from a 50-megawatt data center lease in Britain due to grid limitations, and widespread local opposition in the U.S. leading to construction moratoriums, demonstrate that the physical infrastructure required for AI development faces more than just financial challenges. These issues can lead to lease cancellations and difficulty securing essential permits.

The evolving nature of AI chip technology further complicates the buildout. As construction is delayed, chip specifications and associated cooling and power supply requirements change. This necessitates design revisions for projects already underway, adding substantial unexpected costs and prolonging completion timelines, creating a cycle where delays lead to more expenses and further delays.