Hyperscalers are pouring trillions into AI data centers, but the math behind the buildout is sobering. New research shows these companies must grow their own productivity by a factor of 2.7 by 2030 just to break even on the investment. What happens if they fall short? Bankruptcy and a massive misallocation of capital, according to a Wharton professor who previously served as the SEC's chief economist.

What You Need to Know

Hyperscaler companies plan to invest nearly $1.1 trillion on AI data centers by 2027, yet AI revenues sit at only $150–200 billion this year. To justify the spending, earnings must grow at an extraordinary pace. The risks are amplified by debt financing that could spread losses through pension funds and insurance policies. The outcome will affect not just tech companies but the broader economy.

The Productivity Math

The AI infrastructure buildout is one of the largest capital investments in history. But the revenue gap is stark. Jessica Wachter, a finance professor at Wharton, and her coauthor calculated what growth rate the hyperscalers need to justify their spending. They found that productivity must rise by a factor of 2.7 by 2030 to break even when accounting for cost of capital, a 15% return, and depreciation. That would match the U.S. IT boom of the mid-1990s. The difference is time: the IT boom stretched over a decade. And the AI boom must compress that growth into a few years.

Wachter's Warning

Wachter, who served as the SEC's director of economic and risk analysis, describes the risk plainly. If earnings do not materialize, companies could default on debt payments. That could lead to bankruptcy. She and her coauthor warn that the current buildout may become the largest misallocation of capital in history if no productivity boom arrives. The big tech names involved include Alphabet, Microsoft, Amazon, Meta and Oracle, which partners with OpenAI.

  • The AI companies: Must increase earnings by 2.7 times to break even by 2030.
  • The revenue gap: Current AI revenues are $150–200 billion versus nearly $750 billion in spending this year.
  • The financing risk: Borrowed money ties pension funds and insurers to outcomes.

The Debt Dimension

Then there is the role of borrowing. Hyperscalers have begun using debt to fund data centers, pushing free cash flow toward negative territory. Even Alphabet, known for hoarding cash, reported that its $120 billion in quarterly revenue could not cover its capital spending. The risk quietly spreads beyond the balance sheets of these companies. Debt is woven into pension funds and insurance policies. People may not know they hold this exposure.

Why This Matters

The stakes go beyond one industry. If the productivity boom fails to materialize, the resulting financial losses could ripple through the broader economy. The investments already represent about 3% of GDP. Gary Gensler, the former SEC chair and now an MIT professor, notes that spending has not yet earned commensurate revenues. That pace of investment is unsustainable without a matching revenue stream. The question is whether the technology will eventually deliver. History suggests that even if a correction comes, the underlying AI advances may survive the bust, much like the internet did after the dot-com crash. But the data centers themselves may not.

A Long-Term Bet

No one knows how profitable AI will become. Models may grow more efficient, reducing the need for raw compute. Or demand could slow as customers turn to cheaper alternatives. The hyperscalers are placing a trillion-dollar wager. They need three things to break even: massive AI revenues, broad economic growth driven by AI, and dominance of expensive frontier models over cheaper competitors. If any of those wagers fail, the math unravels.