Over 90% of AI spend comes from the top 10% of customers: broad usage, narrow payment is a fragile base

Over 90% of the money going into AI models and data centres comes from the top 10% of customers. That is the number I would put in front of any board pricing its AI exposure.

The figure comes from Apollo, quoted in Derek Thompson's 26-fact survey of AI and the economy, and several of his other facts sit uncomfortably beside it. Adoption has been faster than any technology on record: about three years to reach half of Americans, against decades for the telephone. Yet by Andreessen Horowitz's data, less than 3% of US households pay for an AI subscription. The cost of building AI is approaching US$1 trillion a year, and 19 of the 25 best-performing stocks in the market are tied directly to the build-out.

Then there is the wobble. Ramp's chief economist reports that in August the top 1% of businesses on its platform cut AI spending per employee by 10%. When spending is this concentrated, a handful of buyers trimming budgets is not noise. It is the signal, and it arrives while the spending commitments are being made years ahead.

Thompson also notes that, as a share of the economy, this is the most expensive infrastructure project since the 19th-century railways, a build-out that changed the country and also produced a run of bubbles and recessions. I read that as a reminder that the physical asset can be real while the revenue underneath it is still unproven.

Broad usage with narrow payment is a fragile base for a capital programme of this size. The revenue case depends less on mass adoption than on a small group of heavy users continuing to scale their spending.

For Australian investors and suppliers riding the AI trade, the practical question is simple. How many of your end customers are actually paying, and what happens to your forecast if the biggest ten slow down?

Source: Derek Thompson, "The 26 Most Important Facts About AI and the Economy", 8 Oct 2026, https://lnkd.in/gtrQjmMW

#Investment #CapitalMarkets #Strategy