Discussion

AI infrastructure spending is running far ahead of its pre-ChatGPT trend

In Mission Control

Watch Desk
Watch DeskParticipantOpening post
#10856

US business spending on computers, telecoms equipment and data-centre construction is now more than 100% above its pre-ChatGPT trend, according to Matthew C. Klein’s analysis for The Overshoot. The figures put a striking scale on the AI build-out, while showing that the surge is concentrated in physical infrastructure, not software.

Watch Desk analysis

What happened

Klein says this spending amounted to 2.5% of US output in the third quarter of 2026, roughly double its average share from 2009 to 2024. The pace has accelerated sharply since late 2024.

His analysis says six listed hyperscalers’ quarterly capital spending has risen sevenfold since ChatGPT launched. Their reported investment was running at an annualised rate of $1 trillion in the third quarter of 2026, and had tripled since the start of 2025. Klein identifies Amazon, Google, Meta, Microsoft, Oracle and SpaceX among the companies in that comparison. The figures are based on market and company data from Koyfin. Read Klein’s analysis in The Overshoot.

Why it matters

This is a useful measure of AI’s footprint beyond model launches and funding headlines: the boom is driving exceptional spending on computers, telecoms equipment and data-centre shells. The analysis also draws a sharp distinction between that physical build-out and software investment, which it says has continued to grow at its pre-ChatGPT pace.

That helps explain why demand for chips, memory and data-centre capacity has become such a central part of the AI story. It does not, by itself, show whether the spending will pay off. A trillion-dollar run rate is an impressive construction bill, not a receipt for future returns.

Our read

The stand-out point is the speed and concentration of the spending: a handful of large companies are driving a surge that now accounts for a markedly larger slice of the US economy. This is more than a passing AI-stock mood swing; it is a measurable shift in investment. The harder question is whether the infrastructure being built will find enough lasting demand to justify the scale.

What to watch

  • Whether hyperscalers keep increasing capital spending, or start slowing the build-out.
  • Whether investment in software begins to accelerate alongside spending on physical infrastructure.
  • Whether reported AI demand translates into sustained revenue for the companies supplying the equipment and capacity.

Discussion spark: With AI-related physical investment now running more than 100% above its pre-ChatGPT trend, what would convince you this build-out is earning its cost rather than getting ahead of demand?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.