Now we’re supposed to believe investors are newly skeptical of AI spending.

Just not at Alphabet. And not at Amazon either. And never mind that Nvidia added nearly half a trillion dollars in market cap over the space of just three sessions this week, briefly becoming the world’s first $5 trillion company in the process.

Mainstream financial media outlets are notorious for letting the price action dictate the narrative, and God knows I’m sympathetic. There’s nothing more uncomfortable editorially than a prominently-positioned article about equities which appears to contradict that day’s cash session on Wall Street.

Simply put: You have to tell the same story that stocks appear to be telling — or risk readers asking whether you’ve checked out for the day, or have otherwise stopped paying attention.

In addition, there’s a sense in which the price is the story. If Meta falls 14% in two sessions, shedding $250 billion in market cap in the process, and the only plausible trigger was management’s “TMI” on expenses, what can you say? Other than that investors were spooked about the temporal disconnect between massive outlays (now) and returns on those investments (later)?

There’s the simple chart: It was a helluva week for Alphabet and Amazon, not so much for Microsoft and Meta. Note the subheading and the annotations. This week’s price action was just a flip-flop of the prior quarter’s mega-cap earnings reaction.

Company analysts might quibble with this — or maybe not considering they’re predisposed and in some sense obligated to staying bullish on the names in their coverage universe — but if we’re honest, the June quarter for Microsoft and Meta looked a lot like the September quarter. The only thing that really stuck out as a differentiator was Meta’s lengthy expense commentary (that, and a one-off tax charge).

A quick review of the June quarter for Alphabet and Amazon likewise admits of very little in the way of differentiators versus the September quarter — certainly not enough to explain the enormous disparity between the earnings-week reaction in the shares.

For three straight quarters now, I’ve dutifully parroted the boilerplate color which says it’s “show me” time on AI capex — that investors are becoming impatient and need to see proof that tens of billions in spending is paying off. And I’ve parsed mega-cap earnings through that lens while explaining the price action by way of supposed “nuance” in management’s capex and expense commentary.

Having done my “job” in that regard, I’m free to say I don’t think we’re anywhere near the moment when investors actually lose patience with these companies’ AI outlays. The reaction to Meta’s expense commentary suggests otherwise, but as I put it while editorializing around the company’s quarter, I think they just used too many words. I think they pounded the table too hard. If you shout loud enough and long enough about how much money you’re going to blow through next year, you’ll eventually succeed in talking the stock down.

The figure below’s from Goldman’s David Kostin. You’ve probably seen it by now even though it’s only a day old. (I feel sorry for people who can’t write and depend on the exclusivity and novelty of sell-side notes to pitch their content. Because the sell-side notes ain’t so exclusive and novel anymore, in case you haven’t noticed.)

Kostin’s repeatedly suggested there’s meaningful upside to these estimates, and indeed they’ve been ratcheted higher every quarter. His point: Even company analysts are having a difficult time keeping up with hyper-scaler capex plans despite having a direct line to management.

If you’re curious as to how concerned the market actually is about Meta’s spending, I’d encourage you to ignore the stock and focus on the record $125 billion of orders the company received for a $30 billion bond sale this week. That’s, um, heavily oversubscribed, and doesn’t exactly suggest investors are reluctant to fund Mark Zuckerberg’s AI push. Rather, it suggests people are beating down the door to fund it. The longest-tenor tranche of the offering — a $4.5 billion, 40-year slice — priced 30bps inside the initial chatter.

Needless to say, that sort of demand will green-light more offerings. With IG spreads as tight as they are, management teams can easily convince themselves that returns on AI investments will be multiples of what a high-grade issuer will pay to fund those investments in the blue-chip corporate bond market. Whether that’s an exercise in self-delusion is anyone’s guess right now, but it could just as easily be true as false.

AI’s becoming subtly ubiquitous. You’re using it even when you don’t realize it, and I’m skeptical of the line that says most of that’s free — that even as companies are monetizing enterprise-level AI services, they aren’t generally being paid for mass-market AI. Nothing’s free, folks. Or nothing that’s worth having. You might not see a bill for the AI services you’re using, but I can assure you you’re paying for them somehow.

It’s not realistic to expect the hyper-scalers to fund all, or even most, of these investments out of free cash flow. That’s not how corporate finance works, and frankly, there’s a sense in which suggesting otherwise is to misunderstand what “FCF” actually is from a definitional perspective. We also seem to be conflating capex, expenses and the non-technical catch-all “spending” and doing so not merely as a matter of convenience for conversational purposes, but as a matter of course — as though we actually don’t know there’s a distinction.

This is another one of those awkward moments when I find myself having to reluctantly lecture people who should know more about a given subject than I do, but apparently don’t. (I dread these moments because it scares me: How is it that someone like me knows so much more about so many things than people who are vastly smarter and wildly more successful? It’s like finding out your parents don’t have all the answers, and it happens to me at least twice a month in all sorts of contexts.)

None of the above is to suggest that all of this spending’s a good idea. I’ve spilled gallons upon gallons of digital ink making the case that it isn’t, or might not be. All I’m saying is that it’s a mistake to extrapolate a newfound skepticism for AI spending from the post-earnings price action in the mercurial equity of two hyper-scalers. It’s even sillier to suggest there’s some happy medium between enormous and ginormous AI spending, where the former’s just table stakes and the latter’s too much of a good thing.

We’re nowhere close to peak AI spending. I think investors understand that, and I don’t think they’re anywhere near out of patience.