Trillions of dollars are being funneled into data centers to support artificial intelligence, and markets are swinging wildly on the latest updates. Nvidia shed 3 percent in a single session after reporting weaker-than-expected equipment sales to data centers, while Oracle saw its stock climb 43 percent—the biggest single-day gain in three decades—after revealing a $300 billion deal with OpenAI.
Data centers have become the backbone of the digital economy, and according to McKinsey, U.S. demand could triple by 2030. That would require close to $7 trillion in new investment. Industry alliances, such as the $500 billion commitment from OpenAI, SoftBank, and Oracle, underscore how much capital is at stake. Even small interactions with A.I. models consume immense processing power, raising questions about whether the revenue side of the equation can keep up.
Investors are already probing for signs of overheating. Joe Tsai, chairman of Alibaba, warned earlier this year that the data center market is showing bubble-like behavior. The financing structure fuels those fears: more than $9 billion in securities tied to data centers were issued in just the first four months of 2025, and Meta tapped Pimco to sell $26 billion in bonds to fund its expansion. Analysts note that these deals hinge on the stability of tenant leases, which adds risk if technology needs shift faster than expected.
The balance sheets of big technology companies provide some cushion. Firms like Meta and Alphabet have the cash flow to sustain losses while building their own facilities. But that safety net depends on A.I. eventually generating profits. If efficiency breakthroughs—such as those demonstrated by China’s DeepSeek—reduce computing demand, the payoff from billions of dollars of construction could shrink.
Some observers are unconvinced there will ever be meaningful returns. Harris Kupperman of Praetorian Capital has argued that the speed of depreciation and the continuous need for upgrades make data centers an unattractive long-term investment. He went further, suggesting that companies may choose negative returns over the risk of falling behind competitors in the A.I. race.
The boom has expanded well beyond Silicon Valley balance sheets. State budgets are absorbing the cost of tax breaks for developers, with at least 10 states forfeiting more than $100 million annually, according to watchdog group Good Jobs First. In Virginia, 40 gigawatts of additional power capacity—three times the current grid—is being built to support over 50 new projects. Phoenix expects to expand capacity by 500 percent.
The environmental costs are also striking. A 100-megawatt data center consumes about two million liters of water per day for cooling, equivalent to the daily usage of 6,500 households. For communities already facing water stress, the trade-off between economic development and resource allocation is becoming harder to ignore.
For now, Wall Street is betting that data center growth equals A.I. growth. But the same logic that has pushed stocks to record gains could just as easily magnify the risks if returns fall short.