AI spending becomes growth only after it changes output

themePublished 19 October 20262 min readequities, macro

Key points

• Investment spending can lift GDP before any productivity gain is visible.
• The IMF's upside includes stronger AI investment and productivity. Its risks include a reassessment of AI expectations.
• The OECD says a downside risk is that AI investment returns fall short of the earnings implied by current spending plans.
• No official productivity percentage for AI is quoted, because none was on the pages used.

What is going on

In the October 2026 World Economic Outlook the IMF says investment in artificial intelligence is one reason activity has held up, and it separates an upside of stronger investment and productivity from a risk that expectations are marked down. The statistical cutoff for that chapter is 1 October 2026. The OECD, on 23 September 2026, says AI-related investment supported production and trade, and lists returns that disappoint relative to the earnings implied by spending plans as a risk to its own projection. The OECD's global growth projection is 2.9 percent for 2026. The IMF's is 3.1 percent. They are not combined here.

Why it matters

The accounting order matters. Capital expenditure shows up as demand for the supplier and as an asset for the buyer. Productivity shows up later, in output per hour, and only if the tool is used. Complementary capital, redesigned work and skills sit between the invoice and the statistic. A firm can therefore report that it 'adopted AI' while its measured labour productivity does not move. Implementation costs, including the power and the people who check the output, are part of the investment. Leaving them out makes the return look easier than the project is.

What to watch

A fair test, of the kind the OECD's sentence implies, compares later earnings with the earnings that current spending plans appear to require. Official productivity releases, when used, need their own statistical office and a period. This note will not treat a vendor survey as that release. Diffusion across industries is the evidence that the spending left the technology sector. Concentration of the spending inside a few buyers is evidence that it has not.

Risks

Measurement can miss quality improvements, and it can also credit a boom in capital spending as if it were already efficiency. Gains can accrue to customers through lower prices, which is a social benefit and a poor margin for the seller. The IMF's inflation projection of 4.7 percent in 2026 is a reminder that the same world can have an investment boom and inflation that is not back at target. This note does not convert either institution's projection into a sector weight.

What would change the reading

The optimistic reading weakens if capital spending stays high and official productivity statistics, cited properly, do not follow. The sceptical reading weakens if productivity does follow outside the firms that sell the tools. Until one of those documents is on the page, the honest statement is the one both institutions already allow: the spending is visible, and the return is contingent.