How crucial are AI investment and AI-enabled cloud services to overall corporate profit growth in America?
I suppose you know the answer, but in a word: “Very.”
This is another riff on the “extreme concentration” narrative, which doubles as a crash canary for the doomer set. It’s a bit ad nauseam on the score, but these are the sort of statistics which, if you don’t highlight them at regular intervals, you chance being remiss in the face of what hindsight will call an “obvious” market top.
With that in mind, the figure below from Goldman’s Ben Snider shows that between them, AI investments and cloud services were responsible for a quarter of aggregate index EPS growth last year. That share’s expected to rise meaningfully in 2026 to almost 40%, on par with 2024.
At least as it relates to the composition of profit growth — and if profits aren’t what counts, then what is? — the “broadening out” thesis has yet to materialize.
“The bulk of the AI-related earnings are attributable to the revenue beneficiaries of the AI capex investment boom, the majority of which has accrued to semiconductor companies,” Snider remarked, adding that although cloud services revenues are small by comparison, those sales “are doubling annually.”
If you’re wondering what the earnings growth contribution breakdown looks like by company, all you need to know is that Nvidia by itself is seen accounting for 24% of S&P 500 EPS growth this year.
Now for the real eye candy. The figure below, from the same Goldman piece, shows you what it looks like when erstwhile capital-light cash machines wittingly transform their business models overnight.
Earnings and FCF used to track each other very closely among the hyper-scalers. Now, the two have diverged entirely.
On a rolling 12-month basis, hyper-scaler earnings growth’s about 23%, whereas FCF “growth” (with scare quotes) is negative to the tune of 32%.
As Snider went on to observe, the result is a cohort which trades very cheap to its own historical earnings multiple at just 24x (14%ile), but hugely expensive on a free cash flow multiple, where “hugely expensive” means an eye-watering 115x.
If your final question’s whether those companies have ever traded anywhere near that on a P/FCF basis, the answer’s obviously “no.”
I expect this year we will see a lot of followers in the Block framework. Massive layoffs based on an unproven assumption that AI will be able to fill the gap. Thus far AI has proven very capable at taking tests but far less so at executing on real world tasks without human input or supervision. I say this as someone who has done quite a bit with AI. But I can confidently say nothing that I did with AI it would have done on its own. This is the crux of the problem, you still need people to direct, implement, measure, sell, and adapt GenAI.
I think the Block layoffs are ai washing. They over hired in 2021 and it’s a nice excuse to make some layoffs. Of course, we can build and manage more with ai, but the same humans need to make sure the wheels aren’t coming off. Further, think about the human element. If I was an engineer making 300k at a top tier start up, I probably went to a good school, look good on paper, and I work just as hard as I have to so that I can enjoy my life. Just like any organization there are ppl that do the work and the ppl that are good at communicating.