What must happen for AI’s trillion-dollar gamble to pay off
| Source: MIT Technology Review AI
Tags: hyperscalers, AI investment, data centers, Jessica Wachter, capital expenditure, Microsoft, Alphabet, Amazon, AI bubble
Hyperscalers need to 2.7x their productivity by 2030 to break even on $1.1 trillion in projected AI data center spend, or risk bankruptcy — a Wharton finance professor's accounting analysis frames the AI buildout as potentially the largest capital misallocation in history.
Details
Wharton finance professor Jessica Wachter cut through AI hype with a simple accounting question: given hyperscalers (Alphabet, Microsoft, Amazon, Meta, Oracle) projected to spend $1.1 trillion through 2027 — rising to $5 trillion over four years — how fast must earnings grow to justify it? Her answer: 2.7x productivity growth by 2030, assuming 15% return on capital and asset depreciation. Total AI revenues in 2026 are estimated at $150-200 billion against ~$750 billion in annual spending. Former SEC chair Gary Gensler, now at MIT Sloan, frames the gap plainly: the spending does not have commensurate revenues yet. Whether those revenues materialize depends on AI productivity gains paralleling the US IT boom of the mid-1990s — compressed into a fraction of the time. Wachter warns that failure would mean the largest misallocation of capital in history. The analysis also examines demand-side uncertainty: AI adoption requires a critical mass of workers and companies using AI in ways that generate measurable returns. It does not conclude a bubble is inevitable, but establishes concrete, verifiable benchmarks for when one becomes unavoidable. Wachter was previously the SEC chief economist and director of its economic and risk analysis division. The spending-to-revenue gap is not in dispute — only whether it will close fast enough.