Hedging The Hyper-Spenders

They’re the most creditworthy entities in the history of capitalism.

And on many accounts, they’re a safer bet than sundry profligate sovereigns including, arguably, the United States government itself.

But they’re suddenly borrowing like there’s no tomorrow, and in the process their business models have metamorphosed. No longer are they capital-light and lean. Now they’re capital-intensive and on their way to becoming debt-laden.

I’m talking, of course, about the hyper-scalers, which are expected to spend anywhere between $600 billion and $750 billion this year on AI-related capex. A meaningful share of that spending will be funded by debt issuance, and that has the potential to transform the investment grade credit universe, as discussed at some length here last week.

So far, those debt sales are going off without a hitch. Indeed, Alphabet’s sterling century bond was nearly 10 times oversubscribed.

The figure above shows you the recent widening in hyper-scaler spreads. Although the market’s beginning to assign a premium to account for binge-borrowing, these companies are hardly paying through the nose.

I’ve heard chatter recently about one or more hyper-scalers potentially announcing a capex cut. With the exception of Oracle, I’d be inclined to ask, “Why?” And also to suggest that welcome as such an announcement might be to some investors, it could send the wrong signal and thereby risk backfiring.

Imagine one of these companies did cut capex. Investors would ask, “What happened?” And then, “Are you less confident that these bets are going to pay off?” And then, “If so, what does that mean for the money you’ve already spent?” While they were asking those questions, they’d probably be selling the stock.

My guess is that while the pace (rate) of capex growth will of course slow going forward, it’ll continue to feel like full speed ahead in terms of “tech” high grade debt supply. And that means a larger and larger market for hedging instruments, which is to say CDS.

The figure above shows you the growth of the (previously nonexistent) market for big “tech” CDS alongside the escalatory debt sales since Oracle’s $18 billion offering in September.

As I was at pains to emphasize last year, when Oracle’s ballooning CDS was in the news every three or four days, this doesn’t as much reflect actual default fears, nor gambling, as it does the necessity of taking out insurance against an emergent left-tail. No one needed these contracts until the hyper-scalers started borrowing huge sums of money.

It’s not that anyone thinks the hyper-scalers won’t be able to pay it back. If that were the case, these debt sales wouldn’t be seeing record demand and pricing (well) inside the initial chatter.

Rather, it’s that, as Bloomberg put it, “some of the richest tech companies in the world are rapidly turning into some of the most indebted,” which in turn raises the following question, as posed by a CIO on record in the linked article: “Do you really want to be nakedly exposed here?”

He meant that rhetorically.


 

Leave a Reply

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

4 thoughts on “Hedging The Hyper-Spenders

  1. “Imagine one of these companies did cut capex. Investors would ask, “What happened?” And then, “Are you less confident that these bets are going to pay off?” And then, “If so, what does that mean for the money you’ve already spent?” While they were asking those questions, they’d probably be selling the stock.”

    Mulling over this, doesn’t it sound like they are now captive to the narratives they’ve spun? They’ve spent, perhaps squandered, so much money that have to keep on spending to save face and even fend off potential class-action lawsuits.* It complemants my thesis that the same mechanism helps explains the industry’s slavish adherence to the “Just make the LLMs even larger!” strategy.

    *JL & GJ – we should nose around those funds that finance class action lawsuits. These could be big ones.

  2. Showing my ignorance here.

    Lots of big numbers thrown out about hyperscalers capex spend. It isn’t clear to me how much is direct borrowing by hyperscalers (hence a liability) vs. how much is run through SPV’s (which I understand are usually carefully constructed to maximize control by, limit actual $ investment from, and limit liability of the hyperscaler).

    I would think there would a big spread differential between borrowing by the strong balance sheet hyperscalers vs. SPV’s backed by an asset of rapidly declining value (e.g. chips) and data centers at the margins of data center bandwidth that may actually not be insatiable (less training/more inference, novel AI software strategies, etc.).

  3. How does this interact with the treasury bond sales currently going onto to fund the government. I’ve read those sales are draining treasuries liquidity, curious that effect might have on the market.

Create a free account or log in

Gain access to read this article

Yes, I would like to receive new content and updates.

10th Anniversary Boutique

Coming Soon