Is AI Capex Holding Up the Economy? The Data-Centre Boom and Its Macro Risk

Macro Guide, 2026

By Ken Chigbo, Founder, KenMacro, UK macro desk.

Find Ken on X @realkenchigbo · YouTube @Realkenchigbo · MACRO MASTERY Discord

Updated 2026-06-08

What hyperscaler capex actually is

The hyperscalers are the giant technology firms building the plumbing of the AI era. Their capital expenditure covers data centres, the chips inside them, and the electricity and cooling that keep everything running. The numbers are eye-watering. Banks and research houses such as Goldman Sachs put combined hyperscaler spend in the hundreds of billions of dollars a year, with widely cited estimates running anywhere from around 500 billion upwards, and the figures keep getting revised higher. Treat all of these as estimates rather than hard counts, because firms report differently and analysts make assumptions. What matters for the desk is the scale. When a single category of corporate investment grows this fast, it stops being a sector story and becomes a macro one. It feeds straight into business investment, one of the building blocks of GDP, and that is where the growth question begins.

How capex flatters the growth read

Strip a quarter of US output into its parts and business investment has leaned hard on AI-related spending. Some analysts estimate that AI capex has accounted for a notable portion of recent GDP growth, in certain quarters punching well above what a sector this size would normally contribute. Again, these are estimates and they move about, but the direction is not in doubt. The desk reads this two ways. First, it tells you the headline growth number may be carrying a heavy AI passenger. Second, it means the growth read is more concentrated and therefore more fragile than a broad, balanced expansion would be. A figure that looks healthy on the surface can rest on one spending cycle. That is not a reason to panic. It is a reason to know what is actually driving the print before you trade off it.

The bull case: real investment, real output

There is a serious argument that this capex is exactly what a productive economy should be doing. Companies are building genuine capacity: compute, networks and power that other firms can rent and use. If AI tools lift productivity across the wider economy, today’s spending becomes tomorrow’s efficiency, and the investment pays for itself in higher output per worker. On this view the capex is not a sugar high but the early, lumpy stage of a real technology build-out, much like railways or fibre before it. Demand for compute has so far kept outrunning supply, which is why the spending plans keep rising rather than stalling. For the desk, the bull case supports a constructive growth and equity regime: investment-led expansion, firm corporate earnings, and a reason for risk assets to hold a premium. The key word, though, is returns. The build only justifies itself if the revenue eventually shows up.

The risk case: an air-pocket if returns disappoint

Flip it round. If the AI investment cycle slows, or if the returns on all this spending disappoint, the same force that flattered growth goes into reverse. You get a growth air-pocket, where a big chunk of investment simply stops contributing, and an equity de-rating as the market repriced earnings and multiples. The knock-on effects spread from there: hiring tied to the build cools, credit conditions tighten, and sentiment turns. Two specific worries sit on the desk’s radar. One is circular financing, where chipmakers and infrastructure players help fund their own customers, which can flatter demand and obscure how much is genuine. The other is a power bottleneck, because data centres need electricity at a scale the grid cannot add overnight, which could cap the build or push costs up. None of this is a forecast. It is the shape of the downside the desk wants to be positioned for, not surprised by.

Why the desk treats this as a regime swing factor

On the desk, AI capex is not a stock-picking theme. It is a swing factor for the entire growth and equity regime. Because so much of the recent growth and so much of the equity market sit on a narrow base of AI-exposed names, the macro read and market concentration have become two sides of the same coin. When spending guidance holds firm and power keeps coming online, the constructive regime stays intact and risk can stay leaned in. When capex guidance softens, power constraints bite, or credit spreads start widening, that is the tell that the cycle may be turning, and it usually shows up in markets before it shows up in the official data. The job is not to call the top. It is to watch the right signals, size positions for a regime that can change quickly, and avoid being all-in on a single growth engine.

The desk’s checklist

1. Anchor the scale before you trade the theme. Start with the size of the spend. Note that research houses put combined hyperscaler capex in the hundreds of billions a year and treat the figures as estimates. Knowing the scale tells you why this is a macro factor, not just a tech story.
2. Decompose the growth print. When a GDP number lands, look past the headline to business investment and how much AI-related capex contributed. If one spending cycle is carrying the print, the growth read is more concentrated and more fragile than it looks.
3. Track capex guidance, not just history. Earnings-season capex guidance from the hyperscalers is the leading edge. Rising or steady guidance supports the constructive regime. A cut or a hint of discipline is the first crack the desk wants to catch early.
4. Watch the power bottleneck. Data centres run on electricity the grid cannot add overnight. Follow power availability, grid constraints and energy costs, because a hard power ceiling can cap the build regardless of how strong demand looks on paper.
5. Check for circular financing. Be alert to chipmakers and infrastructure players funding their own customers. Circular flows can flatter demand and hide how much is genuine end-user appetite, so read who is paying whom before you trust a demand signal.
6. Use credit spreads as the macro tell. Credit spreads widening alongside softer capex guidance is the combination the desk treats as a regime turn. Equities can stay euphoric late, but credit tends to flag stress earlier, so let it cross-check the equity story.

Frequently asked

Is AI capex really holding up the US economy?

It is propping up a meaningful part of it. Research houses estimate hyperscaler capex runs into the hundreds of billions a year and that AI-related investment has accounted for a notable share of recent GDP growth. Those are estimates, not certainties, but the scale is real. The honest answer is that the economy is leaning on this spending more than a balanced expansion would, which makes the growth read more concentrated.

How much are the hyperscalers actually spending?

Combined hyperscaler capital expenditure is widely estimated in the hundreds of billions of dollars a year, with figures from banks and research houses often cited from around 500 billion upwards and rising. Firms report differently and analysts make assumptions, so treat any precise number with caution. The point is not the exact figure but the order of magnitude, which is large enough to move macro aggregates.

What happens if AI spending slows down?

If the cycle cools or returns disappoint, the force that flattered growth runs in reverse. The desk’s risk case is a growth air-pocket as investment stops contributing, an equity de-rating as earnings and multiples reprice, and knock-on effects on hiring, credit and sentiment. This is a scenario to be positioned for, not a prediction that it will happen on any particular timeline.

What is circular financing and why does it matter?

Circular financing is when companies higher up the AI supply chain, such as chipmakers, help fund the very customers buying their products. It can flatter reported demand and make the build look more self-sustaining than it is. The desk watches for it because demand that is partly financed by the seller is a weaker signal than demand from independent end users paying their own way.

Why does power matter to the AI capex story?

Data centres consume electricity at a scale the grid cannot expand overnight. If power availability, grid capacity or energy costs become a hard constraint, they can cap how fast the build proceeds no matter how strong demand looks. The desk treats power as a real-world bottleneck that could throttle the investment cycle, and tracks it as one of the early signals of a slowdown.

How does the desk trade around AI capex?

The desk treats AI capex as a swing factor for the growth and equity regime rather than a single trade. It watches capex guidance, power constraints and credit spreads for the turn, and keeps in mind that this growth engine and market concentration are linked. The aim is to size for a regime that can change quickly, not to call the exact top or bottom of the cycle.

Educational analysis only, not financial advice. KenMacro has commercial partnerships with some firms referenced and may earn a commission if you open an account, at no cost to you. Manage risk against your own circumstances.