1 week ago
54
Five of the world’s largest technology companies are on track to spend close to $1.1 trillion on AI data centers by 2027. The question nobody can confidently answer: will those data centers ever earn it back?
A new analysis from Wharton finance professor Jessica Wachter and collaborator Jonathan Wachter cuts through the hype with a deceptively simple approach. Instead of debating whether AI will reshape civilization or fizzle into an expensive parlor trick, they just ran the numbers on what has to happen for the spending to make financial sense.
The accounting reality check
The Wachters’ March 2026 paper, “What Investment Data Implies about the AI Transition,” focuses on five hyperscalers: Amazon, Alphabet, Microsoft, Meta, and Oracle. Their combined capital expenditures have ballooned from $155 billion in 2022 to $226 billion in 2024, with $381 billion expected in 2025 alone. That’s roughly a doubling every couple of years.
Using a two-sector economic model, the researchers found that justifying this level of investment requires AI-sector productivity to increase by approximately 2.7 times.
S&P Global Ratings paints an even more aggressive picture, estimating that aggregate capex across six firms (adding SpaceX to the mix) could reach $1.3 trillion in 2027, up sharply from a projected $870 billion in 2026. Of that group, only Microsoft is expected to report positive free cash flow in 2027, projected at $33.6 billion. Every other firm in the cohort faces negative free cash flow scenarios.
Debt is filling the gap
Hyperscaler debt issuance surged from $17 billion in 2024 to nearly $194 billion in just the first half of 2026. Goldman Sachs forecasts approximately $400 billion in global investment-grade bond issuance in 2027 to help fund an estimated $1.14 trillion in capex that year. Debt financing is projected to cover roughly one-third of 2027’s total AI capital expenditure.
Moody’s and Goldman Sachs both project hyperscaler capex approaching or exceeding $1 trillion in 2027, with the financial pressure already showing up in deteriorating credit metrics.
Binary outcomes and what comes next
The Wachters frame the situation in starkly binary terms. Either AI delivers a productivity boom substantial enough to translate into significant GDP growth, or the world is staring at one of the largest capital misallocations in history.
Analysts suggest that if capex growth stagnates after 2027, the industry’s focus may shift toward software development and monetization solutions rather than continued infrastructure expansion.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our Editorial Policy.
English (US) ·