How the World Works
Latest data
Link 5 · Real economyIn the chain

How much of America's growth is AI? Between a sliver and a third

Economists disagree widely on how much artificial intelligence adds to US growth: from 0.1 to about 0.8 of a percentage point. The gap comes from three channels that are hard to measure: investment, imports and the stock market.

US growth AI? invest − imports + wealth
0.1 ptGoldman Sachs estimate
~1/5Fed working paper (investment only)
~1/3ING estimate for 2026
2.2%Railroad build-out, % of GDP a year (1870–90)

What happened

The US economy has been growing faster this year, and the Fed's chair described growth as strong on 28 October. Much of the debate is about one question: how much of that growth is the artificial-intelligence boom? There is no "AI" line in the official GDP figures, so economists have to estimate it, and their answers differ widely.

  • Goldman Sachs: about 0.1 percentage point over the past year.
  • Nationwide: about 0.6 point, a quarter of growth, including side effects such as construction jobs and stock-market wealth.
  • ING: about 0.8 point, a third of its 2.4% forecast for 2026.
  • A Federal Reserve working paper found AI-related investment worked out to about a fifth of growth over a year.

Why the estimates are so far apart

Investment counts, but imports subtract. Spending on data centres, servers and software is business investment, which adds to GDP. But GDP measures what is produced in the country. Many AI chips and computers are made in Taiwan and South Korea, so when a US company buys them, the import is subtracted again. A huge investment can add surprisingly little to US growth. That is the main reason Goldman's number is small.

The wealth effect. AI hopes have lifted share prices, and households that own shares tend to spend more when they feel richer, even if the gains are only on paper. By some estimates, about half of AI's contribution to growth comes through this channel. It works in reverse too: if share prices fall, spending can slow.

What is not measured. Bloomberg Economics argues that official statistics miss some of AI's effect, and that growth in 2025 would have been 3.1% rather than 2.3% if it were fully captured.

Why it matters

If AI is a large share of growth, then a pause in AI spending would hit the whole economy, not just technology firms: construction, power, and the spending of better-off households. The scale is unusual. One projection puts AI-related investment at about 3.6% of US GDP a year until 2032, above the railroad boom of 1870–1890, at about 2.2%.

Where it sits in the chain

This is Link 5, the real economy. Investment is the most volatile part of GDP, and this time a single theme drives much of it.

What to watch

  • The first estimate of US third-quarter GDP, due at the end of October, and the contribution from equipment and software investment.
  • Big technology firms' spending plans for 2027 (see Alphabet's results).
  • Share prices: a sustained fall would test the wealth effect.
Where this sits in the chain
  1. 1Inflation
  2. 2Central bank
  3. 3Expected rates
  4. 4Long-term yields
  5. 5Real economy
○ External shocks → Link 1
○ Government finances & politics → Link 4
Concepts in this story
Story thread · 4 storiesThe AI investment boomSpending on artificial intelligence lifts chip exports and business investment, and draws record sums of new shares, debt and leases to pay for it.
Newsletter

The Sunday note

One email every Sunday: the week’s most important stories, the mechanics behind them, and the numbers to watch next week. Free.

We use Kit to send the newsletter. Your email is used only for it, and you can unsubscribe in one click.

Sources

  • Estimates by Goldman Sachs, Nationwide and ING, a Federal Reserve working paper (July 2026) and Bloomberg Economics, as compiled by Bloomberg (28 October 2026)
  • Projection of AI-related investment by Stijn Van Nieuwerburgh, Columbia Business School, as reported by Bloomberg
  • US GDP growth: Bureau of Economic Analysis