The Problem With AI Economy


As you know, my working theory over the last few months has been Run It Cold Now, Run It Hot Later.

Given how much the Run It Hot Later seems to be the consensus, in this piece I’ll focus again on the Run It Cold Now portion – specifically: what are the risks of a proper growth scare? And how do we trade it?

The Run It Cold Now theory is predicated on the tariff-as-a-tax negatively impacting the US economy without any real fiscal offset until the OBBB kicks in early 2026. Yet people are watching hold fine and the stock market soar, and they wonder what slowdown are we even talking about here?

The chart above shows one of the key issues with today’s economy: AI-related Capex spending accounted for more GDP growth than consumer spending in the first 2 quarters of 2025. Absolutely staggering.

Look, there is no big issue with capex-led growth as long as:

1) Capex turns to be profitable with decent ROI (otherwise earnings will have a problem);

2) Capex doesn’t overly rely on debt (otherwise we risk the typical private debt bubble burst later);

3) Capex creates jobs, so that growth is organic and well-distributed (does AI create jobs? Mmmhhh)

We can’t say much about the ROI yet, but we can definitely say that AI-related capex could soon enter the (more dangerous) debt-funded stage in 2026.

The chart below shows the big-tech announced capex spending as a % of their EBITDA – it’s already over 65% on average, exceeding the AT&T (NYSE:T) spending of 1998.

To keep up this pace next year, companies will have to resort to debt-funded AI capex:

AI Capex mechanically adds to US GDP even before we get to talk about the ROI.

But the biggest issue with AI Capex is that it doesn’t really add jobs for the median American for now, and hence, we are left with two economies: a hot AI-related economy, and a broader labor market struggling under the fiscal tightening induced by tariffs.

A simple analysis of US jobs created since 2015 shows how the policy-affected industries account for over 40% of the overall jobs created, and how the hiring slowdown in these industries is turning into net layoffs.

The first chart shows how the sectors affected by tariffs (e.g. manufacturing, trade etc) and by higher rates and policies (e.g. housing and government hiring) account for 40%+ of total job creation since 2015, while the tech-related job creation is very small.

The table below shows the 3-month average hiring pace in such tariff and policy-affected industries: a modest net hiring has turned into net layoffs already.

AI-related capex will sustain GDP and a soaring stock market at least until the ROI question gets some answers in 2026, but the broader job market is in trouble:

In a world where the AI economy is running hot but the blue-collar job creation struggles, how should investors approach markets? Is AI a bubble that will inevitably burst?

The anecdotal evidence I hear is that some investors are still like the old man yelling at clouds that the bubble will burst anytime soon.

Their key point is that this Capex ROI is nowhere to be seen, and that AI-related companies are just pushing each other’s stock prices by announcing deals. They don’t say it out loud, but they think it’s a Ponzi scheme.

But here is what really makes bubbles burst: private sector debt-funded manias:

As my friend Dario Perkins shows, the most recent example is the Chinese housing market, but there are plenty more: the Asian tigers crisis and the US GFC were triggered by an excessive use of leverage from the private sector.

Yet out of the ~400bn in US AI Capex in 2025, about ~140bn has only been debt-funded, and these are the debt/revenue ratios for the largest US AI-related companies at the moment:

Next year, AI Capex will be even larger and more debt-funded. Then we’ll eventually get to the point where ROI matters because companies must service debt – and companies can’t print money like the government.

But it can still take several quarters, and do you want to be the old man yelling at clouds for so long?

Thanks for reading!

***

This article was originally published on The Macro Compass. Come join this vibrant community of macro investors, asset allocators and hedge funds – check out which subscription tier suits you the most using this link.





Source link