Economists debate how much AI is boosting US GDP

In short: US investment in computer equipment and software has risen sharply since early 2025, but economists say GDP figures may still undercount AI’s real effect.

What's going on

US economic data shows a noticeable rise in spending on computer equipment and software as a share of real GDP, which is a measure of how much the economy produces after adjusting for inflation. Financial Times writer Robert Armstrong notes the increase starts around early 2025 and is likely connected to the AI boom.

Even with that jump, Armstrong argues it is easy to overstate what it means for overall growth. Some parts used to build AI data centres are imported, so GDP also records a negative offset under trade (like buying more from abroad at the same time you build more at home). He also notes that consumer spending is still a much larger part of the economy overall, even if tech investment looks large in one quarter.

Another issue is measurement. AI tools are improving quickly, and the cost of getting a useful result can drop fast. That is similar to how computers got more powerful over time for the same price, but potentially faster. If official statistics do not fully capture these fast quality gains, real GDP could be understated.

A recent paper from economists Anton Korinek and Patrick McKelvey at the Peterson Institute estimates US AI compute spending rose from $37bn in 2023 to $219bn in 2025. They also estimate the price of AI “inference” (the cost to get an AI system to produce an answer) fell 94% over that period.

What to watch

The big open question is how much of this cheaper AI output is actually used for productive work, rather than low value content. Future GDP and productivity updates may hinge on how statisticians measure rapidly improving AI services.

Source: Financial Times