As the artificial-intelligence boom accelerates, data centers have emerged as the unsung heroes powering everything from cloud computing to machine learning. But beneath the surface of this rapid expansion lies a potential economic vulnerability that could ripple through financial markets. In a recent analysis, economist Noah Smith warns that the frenzy to build these facilities might be setting the stage for a crisis, drawing parallels to past bubbles like the dot-com era or the housing meltdown.
Smith’s argument hinges on the financing mechanisms fueling this growth. Data centers require massive upfront capital for construction, energy infrastructure and high-tech equipment, often in the billions per project. Traditional banks have been cautious, leaving a void filled by specialized lenders known as business development companies (BDCs). These entities, which operate outside standard banking regulations, provide high-yield loans to risky ventures, including data-center developers. As detailed in the Noahpinion blog, BDCs have poured funds into this sector, betting on sustained demand from tech giants like Amazon and Google.
The Shadow Banking Parallel and Rising Risks
This shadow-lending ecosystem echoes the subprime mortgage market of the 2000s, where lax oversight led to overleveraged bets. Smith points out that if AI hype cools or if energy costs spike—data centers are notorious power hogs—these projects could falter, leaving BDCs with nonperforming loans. A wave of defaults might then cascade, as BDCs rely on investor capital that could dry up in a panic.
Compounding the issue is the sheer scale of investment. According to a January 2025 forecast from Reuters, citing J.P. Morgan, spending on data centers could add 10 to 20 basis points to U.S. GDP growth through 2026, underscoring their economic weight. Yet this boost comes with strings: many facilities are concentrated in states like Virginia and Texas, where grid strain is already evident. If overbuilding outpaces actual demand, as some insiders fear, asset values could plummet.
Regulatory Gaps and Market Vulnerabilities
Industry experts note that BDCs aren’t subject to the same capital requirements as banks, allowing them to take on riskier debt. A comment thread on the Noahpinion blog raises pointed questions: Why do these entities exist if banks deem the loans too risky? The answer lies in yield chasing—investors flock to BDCs for returns upward of 10%, far above safer bonds. But this creates fragility; a downturn could trigger forced sales, amplifying losses.
Smith doesn’t predict an imminent crash but urges preemptive thinking. He references historical precedents, like the telecom bust of the early 2000s, when fiber-optic overinvestment led to bankruptcies. Today’s data-center surge, driven by AI’s insatiable need for compute power, shares similar traits: speculative builds in anticipation of endless growth.
Balancing Optimism with Cautionary Measures
On the flip side, proponents argue that data centers are foundational to the digital economy, with demand projected to soar. A 2025 outlook from JLL, a real estate services firm, predicts phenomenal sector growth, fueled by hyperscale operators. Innovations in cooling and renewable energy could mitigate risks, ensuring sustainability.
Still, for industry insiders, the lesson is clear: vigilance is key. Regulators might step in with tighter oversight on BDCs, while investors should diversify beyond high-yield data bets. As Smith concludes in his piece, foreseeing a crisis doesn’t mean averting it entirely, but it could soften the blow. With trillions at stake, the data-center boom tests whether we’ve learned from past financial follies—or if history is poised to repeat.