Have The Hyper-Scalers Lost Their Minds?

For the first time, I’m starting to wonder if hyper-scaler management might’ve lost the plot in earnest.

By “in earnest” I mean I’m beginning to suspect a case of mild delirium might’ve taken hold in the C-suite among the big AI spenders. (Some of you will immediately quip that “mild” is an understatement of trillion-dollar proportions.)

It’s not that I expected capex plans to be cut this quarter. I obviously didn’t. No one did. Nor did I anticipate management would evidence a rekindled commitment to capital discipline despite clear signs of market angst around the trajectory of AI-related borrowing and spending.

There’s been no change to analyst estimates for 2027 hyper-scaler capex, which is seen at $1.2 trillion, and everyone knows the free cash flow situation’s going to get worse before it gets better.

The figure above shows you historical and projected FCF for the big-spending tech heavyweights.

So, again, it’s not that I’m surprised by anything we learned over two weeks of hyper-scaler reports. Indeed, we didn’t learn anything. That’s kinda the point. They’re still spending, planning to spend more and with FCF turning negative, funding that spending will entail debt or equity raises. (Both.)

But I wonder if management learned anything from the decidedly adverse price action seen in shares of Meta, Amazon and particularly Alphabet following earnings. Microsoft held up better, but note from the chart that FCF’s still positive there, and not expected to turn negative.

With that in mind, we’ve just seen two more hyper-scaler debt deals, and no sooner had they launched than Bloomberg was out previewing another one.

As the figure shows, Alphabet launched a seven-tranche EUR deal today, with maturities from two to 19 years (it’s the second time Alphabet tapped the EUR market this year) and Oracle a $10 billion, six-tranche deal with maturities from three to 20 years alongside a $10 billion stock offering.

That’s on top of Microsoft’s 10-tranche $25 billion deal which priced last week, and ahead of what Bloomberg says could be a $20 billion, cross-border placement from Meta. And don’t forget about the pending $40 billion SpaceX deal.

This is too much supply, and that’s showing up in the pricing. Alphabet’s paying more this time in the EUR market than it did in May and as Bloomberg noted in a separate piece, the company’s existing euro bonds “have widened as much as 60bps since they started trading.” That as investors in Alphabet’s massive equity offering are deeply underwater following last week’s rout in the stock.

Meanwhile, Anthropic amended its S-1 to include Q3 results. The updated filing also tallies the company’s commitments and lease agreements, which come to “approximately” $895.4 billion. I don’t know why “approximately” sounds funny there, but it does.

Maybe Claude can tell you why that’s not a scary sum, but… well, the only way $895.4 billion (“approximately” or exactly) isn’t a large number is if you’re Elon Musk on a day when SpaceX is up big.


 

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.

3 thoughts on “Have The Hyper-Scalers Lost Their Minds?”

  1. Recently, I read that Gemini, is hurting advertising earnings, and that Gemini is a pretty good AI that will be enough for a lot of users. The money those other big companies are pouring into building even more powerful AI might be wasted because for many (most?) people, Gemini answers their needs.
    Then today, in Tom Friedman’s column in the New York Times, he says China is moving forward very fast with “good enough” AI , and doesn’t need to spend billions to make AI just a little bit better. I am paraphrasing, but to me these things I am reading seem reasonable. Will there ever be free cash flow to pay back these big investments?

  2. If they’re really convinced of the value of being the ones to find the holy grail, I can see logic and justification for going for all the cash they can gather while the gathering is still good (even if not so good as it might have been). A couple years of negative FCF isn’t likely to sink the big hyperscalers – if it only goes on for a couple or three….though I wouldn’t touch Oracle or some others. I’ve always thought companies with capital needs for a perceived home-run reward should grab all the cash they can when it’s available, never mind some dilution or loss of market cap – as long as they’re not going have a continuing burn rate that outstrips the ability to attain or restore positive cashflow. But what do I know….

  3. The bigger concern to me is if/how everything around the AI rapid evolution may affect core (making them legacy?) businesses – like the huge hit Intel took from weakening sales into the PC market. Some version of that is likely to come for AMD, perhaps MSFT as well; and what about Google’s search/advertising biz, and AMZN’s shopping domination. Then there’s AAPL; not sure what to make of them…

10th Anniversary Boutique

Coming Soon