Tech Produced 76pc of S&P 500 Profit Growth. Only 2pc of Households Pay for AI.

Technology generated about 76 percent of S&P 500 profit growth in 2026, according to a16z's State of Markets II report. In spring, about 2 percent of US households paid for an AI service.

Those two figures describe the same boom from opposite ends, and the distance between them is the story.

Capital is moving from software to steel

The report finds capital shifting out of software and toward physical infrastructure. Big Tech profits are funding AI computing capacity, increasingly with debt rather than cash flow — a change in the financing mix that has had far less attention than the capex headlines.

Demand for compute is strong enough that older Nvidia A100 GPUs still command high rental prices. New chips are not pushing the old ones out of service, which is what you would expect in a market where supply, not obsolescence, is the constraint.

Adoption is early and badly measured

Almost 30 percent of S&P 500 companies report a measurable AI effect. Only about 2 percent track a specific metric for it.

That is not a small discrepancy. It means most of the corporate claims about AI benefit are assertions rather than measurements, and that the aggregate picture of enterprise adoption rests on self-report.

The software sector underneath shows the same split. About 75 percent of public software companies are profitable, but only 30 percent grow faster than 20 percent a year — profitable maturity, not the growth the valuations imply.

Why it reads differently from Addis Abeba

Ethiopian firms are not buying GPU capacity, but they borrow in a world priced by these numbers. If the AI build-out is being financed with debt and the revenue proving it out is thin, the correction shows up as a change in global risk appetite — which reaches Addis Abeba as the cost and availability of external finance, not as a Nasdaq headline.

Open

The report is an investor's view of its own sector. It does not say how much of the debt financing the AI build-out sits on the balance sheets of the hyperscalers versus their suppliers, and that is the number that would show how fragile the structure is.

Sources: Venture Capital, 5 October 2026