The title of this week’s first daily mailer was “How much is too much?”
The reference was, of course, to runaway AI capex spending. Not so long ago — where that means as recently as end-Q3 2025 — investors were rewarding spending plans with no questions asked.
I’d argue Oracle’s $18 billion debt sale in September marked the beginning of a shift in that psychology. Starting with mega-cap tech earnings the next month, markets began to cast a more wary (or, if that’s too strong, we can say “discerning”) eye towards unbridled spending.
Fast forward four months and the hyper-scalers have sold another $100 billion in debt and tipped capex north of $650 billion for 2026. Analysts expect the two-year total for 2026 and 2027 to approach $1.5 trillion. For some professional capital allocators, enough’s enough.
There’s the chart. You’re likely to see it a lot on Tuesday. It’s from the latest installment of BofA’s closely-watched Global Fund Manager survey, and it shows that the net share of panelists saying companies are over-investing posted a very large month-to-month increase.
There’s a lot to glean from the visual in addition to the “too much capex” point. For two straight decades, investors generally believed corporates were under-investing. The recent shift thus marked a sea change and at least according to these survey panelists, it’s too much of a good thing already.
Just 20% of survey respondents said companies should improve capital spending when asked about the best use of cash flow. As the figure below shows, that share’s rarely been lower.
A third of panelists said companies should return cash to shareholders, the highest share in a dozen years. 35% said management should improve their balance sheets.
“Capex is too hot right now,” BofA’s Michael Hartnett said, in the editorial accompanying the survey results.
Meanwhile, in a nod to an AI-driven “loan-ageddon,” the share of respondents who said private equity and private credit are the most likely source of a credit event rose to 43%.
As the figure shows, that’s up markedly from January. Hyper-scaler capex was number two on that list, commanding a 30% share.
So, if a meaningful share of the angst in private credit is now a function of the AI disruption threat, three-quarters of the perceived credit event risk in markets is now tied to AI.
Ironically, one side of that assumes AI lives up to its promise (e.g., disruption from agentic AI undermines software businesses, causing private loans to sour) and the other side assumes it doesn’t (i.e., incremental hyper-scaler capex turns out to be good money after bad with the benefit of hindsight).
This month’s fund manager poll included responses from 162 participants who between them manage nearly $450 billion, not even enough to fund nine months of hyper-scaler capex in 2026.
