AI spending could hit 9% of US GDP by 2032

The surge in artificial intelligence investment by major technology firms is projected to require annual spending equivalent to approximately 9% of the United States' gross domestic product. This figure mirrors the proportion of household income Americans currently allocate to food purchases.

Scale of financial burden

This estimated expenditure level represents roughly double the total national outlay on all energy sources, computers, and software combined. It also equals approximately seven times the aggregate consumer spending on telephone services, streaming platforms, and internet connectivity.

The analysis originates from Stein Van Nieuwerburgh, a finance professor at Columbia University. He concludes that the current expansion in AI infrastructure surpasses any previous investment boom in American history. The critical variable, according to his assessment, is the revenue volume the industry must generate to justify these massive capital deployments.

Revenue requirements for viability

Van Nieuwerburgh’s model begins with actual data center construction plans and average build costs. He operates under two specific assumptions: each dollar of revenue generates fifty cents in cash flow, and investors demand an unlevered return on invested capital of 10%.

Based on these parameters, AI revenues must reach $3.5 trillion by 2032. This target corresponds to 8.8% of the US GDP, assuming nominal annual growth of 4%. The scenario presumes that AI companies can maintain their current pricing structures relative to computing power.

However, Van Nieuwerburgh notes that scarcity of computing capacity currently sustains high prices and profit margins. He argues that AI firms implicitly assume they can continue charging rates reflecting this scarcity through 2032. Market signals suggest intensifying competition as computing capacity quadruples, raising the risk of price declines alongside production expansion.

Jevons paradox and efficiency gains

In the technology sector, falling prices do not necessarily indicate shrinking market value; they often reflect rising efficiency. Data from the AI sector shows model capabilities doubling approximately every four to five months.

According to Epoch AI, the effective price for a specific level of AI capability has dropped by 47% per quarter since 2023. This rate of decline is about six times faster than the reduction in computing power costs.

Optimists argue that lower prices and higher capabilities will expand demand more than the price drop reduces revenue. This phenomenon, known as the Jevons paradox, is named after British economist William Stanley Jevons.

Corporate performance benchmarks

Current demand indicators provide some support for this hypothesis. A panel of experts surveyed by a group led by Ezra Karger of the Federal Reserve Bank of Chicago forecasts that OpenAI and Anthropic will generate $300 billion in annual combined revenues by 2030.

This projection may appear conservative given current trajectories. The two companies already report a combined run-rate of approximately $180 billion, growing at double-digit rates on a quarterly basis. Proponents of AI suggest the technology will become a third factor of production, alongside capital and labor.

If this scenario materializes, spending equal to 9% of GDP on AI might not seem excessive. Labor currently accounts for about 51% of the US GDP. However, such a shift would represent a significant structural transformation in the economy.