Nobody likes to hear “I told you so,” even when it comes from someone, in this case me, they enjoy reading or listening to.

But here it is anyway: I told you so. Honestly I did.

The last Weekly, entitled “Something Wicked This Way Comes?” (remember Dan’s dictum: “If you think you might be wrong, or that we might get in trouble, just pose it as a question”), reads prescient seven days on.

“Although the jury’s still out on whether we did in fact witness ‘the start of something’ mid-month, this week had the same ominous feel as its foreboding predecessor,” I wrote, conducting a vibe check on October 23 amid proliferating AI angst.

I went on to warn that this month’s neocloud mini-panic, the adverse reaction to Meta’s spending plans and widening in all things “AI spreads,” together raised the already high stakes for earnings reports from Alphabet, Amazon and Microsoft. To wit:

We’re not likely to get any allusions to capex restraint from [the] hyper-scaler reports. In fact, we’ll probably get the opposite. Here’s hoping the Google Cloud, AWS and Azure growth prints are robust enough to justify the profligacy.

Fast forward seven days and I think it’s fair to call that a pretty accurate read.

Google Cloud revenue more than doubled in Q3, Alphabet said Wednesday. 24 hours later, Andy Jassy said AWS sales rose 39% last quarter. Both of those prints exceeded consensus, but it didn’t matter. The beats were overshadowed by escalatory capex, plunging free cash flow and, insult to injury, tentative evidence of weakness in the companies’ core businesses.

The figure below gives you a sense of how AI’s impacting cloud sales at Alphabet, Amazon and Microsoft.

As a technical aside, a double y-axis is necessary here. If you plot these growth rates on the same axis, it makes it appear, to the casual observer, that Alphabet’s outperforming Amazon and Microsoft, when in fact the growth rate disparity’s a function of AWS and Azure being more mature businesses than Google Cloud.

Plainly, AI’s driving an inflection in big-tech’s cloud businesses. There’s little question about that. The concern is that it’s not coming fast enough to justify the rate at which they’re spending. And do note: Meta still doesn’t even have a public cloud business, although they’re apparently building one.

This week’s earnings releases also raised uncomfortable questions about the extent to which AI may be cannibalizing the core businesses of the big-tech firms at the bleeding edge of the revolution. There’s not a ton of evidence to support that just yet, but… well, suffice to say I do more searching on ChatGPT than I do on Google these days. And as I’m sure most you’ve noticed, Google now steers you aggressively towards a Gemini-powered search experience which, in addition to being inferior to “regular” Google in my opinion, could be costing Alphabet on the paid clicks front.

In the meantime, investors are left to ponder the highly uncomfortable prospect that the crown jewels of American capitalism have become debt-laden, capital-intensive, equity-issuing, cash burners. Apropos, S&P cut Meta’s outlook to negative on Friday evening, citing, “uncertainty around how long capex will remain elevated and FCF will be negative.”

The figure shows you the TTM FCF metric for Meta (which reported negative free cash flow for the first time in Q3), Alphabet and Amazon.

In the rationale for cutting Meta’s outlook, S&P noted that its projection for a “dramatic increase in the company’s leverage” is mostly a function of Meta’s data center leases, not debt sales. S&P went on to say that although Meta’s massive equity raise may cover the company’s FCF deficit next year, the ratings agency “expect[s] the company will solely use debt” to cover 2028’s projected shortfall.

Of course, the worse the outlook for your credit rating, the more it’ll generally cost you to borrow, which means you might be compelled to issue more stock instead, raising the specter of dilutive events for existing shareholders. And around we go in the sort of unpalatable loop no one ever thought they’d have to consider vis-à-vis mega-cap US tech.

As Bloomberg not-so-gently noted on Friday, investors who participated in Alphabet’s June equity offering are underwater badly following the worst week for the stock since March of 2020. Investors in Meta’s offering are at risk of finding themselves in a similar predicament if sentiment doesn’t improve.

None of the above’s to suggest this whole project — the “leveraged buildout,” as I’m fond of calling it — isn’t worth it. I think it is. I think AI, even in its current form, is absolutely revolutionary. Just in my own little sandbox (no programming pun intended), I use ChatGPT for the majority of my data-gathering and processing needs now instead of the Bloomie. That says a lot. Consider how expensive a Bloomie is. I can more or less replace it with a basic ChatGPT subscription that costs me $21.85/month. I’m going to keep the Bloomie for now, but maybe not for much longer.

So, when I write, as I did in the last Weekly, that “we’ve reached some manner of inflection point” beyond which the hyper-scalers — and really everyone involved in the AI race, from the HBM oligopoly to the neoclouds to trillion-dollar data center REITs with interstellar side hustles — will be forced to contend with “pushback,” I’m merely suggesting that shareholders and capital markets are in the process of demanding these companies show some near-term discipline.

Late Friday, the FT reported OpenAI cut its target for a new funding round by $10 billion in light of “a brutal selloff for tech stocks and growing anxiety about the pace and extent of AI spending.” The company’s now looking to raise $20 billion at a $1.25 trillion valuation from an investor group that includes the usual heavyweight suspects including Andreessen Horowitz, Sequoia and SoftBank. The biggest investor in the new round, the FT said, is the United Arab Emirates.

Addendum

On Saturday morning, I casually asked ChatGPT to solve Professor Brand’s blackboard equation from Interstellar.

It offered a set of preemptive caveats, the gist of which seemed to be that I was asking for the impossible. Not being a PhD in Aerospace Engineering — and unable to raise Roberta Sparrow for a Zoom consultation — I couldn’t make heads or tails of the model’s highly technical, quasi-apologetic preamble, let alone the document it produced after 15 or so minutes spent “Thinking.”

For whatever it’s worth, I’ve embedded ChatGPT’s “Speculative Gravity Control: A field-theory candidate” below. You know, in case anyone’s looking for some “light” Halloween reading.