If you own an equity index fund that tracks a cap-weighted, large-cap benchmark which doesn’t explicitly exclude US stocks, somewhere between 20 and 40 cents (depending on what index you’re tracking) of every dollar you invest goes into shares of America’s hyper-scalers.
Some argue that’s suboptimal, to employ a polite euphemism. The issue isn’t so much that the market leadership comprises such a large share of overall market-cap (that’s the definition of “market leadership,” after all), but rather that passive investing is creating a self-fulfilling, winner-take-all dynamic wherein a handful of names are becoming almost synonymous with the stock market itself.
Consider, for example, that the largest five US tech companies were worth 16% of the entire global public equity market as of late last year, according to Goldman. If you expand that to include the biggest 10 US stocks, that group accounts for a quarter of the global equity market.
There’s the chart which illustrates the stats. It’s pretty remarkable. “This degree of concentration is, in our view, unsustainable, but this is not the same as saying that we are experiencing a bubble, that technology cannot remain the dominant sector or, indeed, that the leading companies in technology cannot remain good investments,” Goldman remarked.
I bring this up not to pound the table on equity market concentration, but rather to introduce a new issue for market participants: The notion that escalatory borrowing to fund AI-related capex will mean mega-cap US “tech” begins to comprise a larger and larger share of the investment grade credit universe.
I put “tech” in scare quotes because Amazon’s classified as a consumer cyclical and Meta as a communications company. Both are a big part of the AI borrowing binge. So, not Tech as officially classified, but “tech” as we all understand it.
The figure above gives you a sense of how these deals are piling up.
Earlier this week, Alphabet raised $32 billion in less than a day across three markets, including the century bond I mentioned here, shattering UK and Swiss corporate credit records in the process. The 100-year tranche of Alphabet’s five-part sterling deal ended up 10 times oversubscribed. That strong demand naturally meant the deal priced well inside the expected spread to gilts.
Very long story short (there’s a joke in the italics), investors lined up around the block to loan Alphabet money for 100 years at just 120bps over UK government bonds. That seems pretty tight to me, even considering gilts have traded more like EM bonds than DM sovereign debt at various intervals post-pandemic.
In recent days, mainstream financial media outlets (all of them) spilled gallons of digital ink expounding on the read-across of these debt sales for credit markets. A lot that was wasted editorial energy, not because this isn’t important (it’s hugely important), but rather because the key question’s very straightforward, even if the answer’s hard to get at.
Here’s that big-picture question: To what extent will the hyper-scaler debt binge end up altering the composition of the IG credit market such that more of every passively-allocated dollar is tied to US “tech”?
And here’s the partial answer: Mechanically, these enormous bond offerings will increase tech’s share of widely-tracked indexes, which is to say the reference point(s) for retail corporate credit products.
That’s not inherently problematic, but it could ultimately leave the high grade credit market with the same latent (i.e., potential) issue as stocks: The entire asset class just becomes one giant bet on AI.
I read an article a while ago titled something like “AI eats everything “. The story went on to explain data center construction and operations effect on water, electricity, land use, construction trades, and on and on. Another facet of the modern world gets added to list.
Much of tech was celebrated for being low leverage and asset light. You can throw that playbook in the trash….
What does that mean?
It suggests to me that valuations need to adjust down in the next 3-5 years during the build out phase at least for the hyperscalers.
If the investments generally pay off, the winner’s valuations can bounce back after the build out.
Valuations in reference to what? Costco trades at a PE of 53. Walmart at 47. Alphabet is 28.6. Oracle is 29.6. The tech companies are profitable. They are investing in their companies under unusual tax-advantaged conditions, and they are dealing directly with the President to secure the best international trade advantages available in the U.S.
One explanation may be that Walmart and Costco have come to be regarded as “consumer staples” as evidenced by their inclusion in some consumer staples ETFs. That partly explains their high P/E ratios.
There was a time long, long ago when META, GOOG, AMZN and AAPL were widely considered to be in the same club. As RIA noted, as asset light profit heavy companies. In a matter of weeks, with the exception of Apple, they have squandered some of that staple designation. Now their relatively low P/Es may just be starting to reflect that shift in perception?
My initial take on the 100-year Google bond was that the maturity was an estimation of when their capex tsunami might start to break even. and generate earnings. Just joking but I wonder if longer maturity hyperscaler bonds might be a SAFER bet than the shorter-term issues. Remember, bond prices are based on the premise that you will eventually get your principal returned to you at maturity. I dunno, but I question that when it comes to META and ORCL. But as our Dear Leader suggested, any buyer of a bond fund outside of the government-only variety will unwittingly be holding an outsized weighting of their paper.
But who care if they juice up my quarterly returns for a year or two??? After that I’ll be dead and it’ll be your problem.
.
My guess is that these hyperscaler bonds will soon be changed by the market and forced to re-finance with principal down and rates upward. I have invested mostly in bonds for decades but I won’t touch any of this junk with a stick. As the AI build out slows, and it will and when profit grow slacks, as it will, all that bond capital will be back in the buy-back market which means it’s not invested.
I have been largely avoiding the AI stocks for a while now because I think at the end of the day, most of these companies will be unsuccessful at monetizing their technology. I have the luxury of not investing other people’s money so if I underperform, it is no big deal. (I remember the 2007 quote attributed to Chuck Prince of Citigroup “as long as the music is playing, you’ve got to get up and dance.” I know from experience that when the music stops, I will not be able to find a chair.) Avoiding exposure to this debt fallout might be a little more difficult.
Same for me. I measure my future by finding and and analyzing inevitabilities. There are two kinds of inevitability. One involves a future that cannot happen and one involves a future that must happen. The trick with AI is to determine which the latter is. Here’s a question? What’s the likelihood your house will ever be served by a human sized robot, tracking dirt everywhere, making your favorite meals just like your momma did, completing routine oiling and greasing, watching one of your favorite movies and knowing just why you love it? Come on, half the country cannot even find a place to rent. Fifteen percent haven’t got enough food or water. With a maximum traditional 30% income cap on housing costs only 10% can buy an average house. A stupid chat room I frequent just rewrote my meager contribution to eliminate the word idiot when the post actually spoke of clowns and idiots. This is useful AI. No revenue was created. Do you want to use robot doctors and dentists? Not me. Profitable AI applications with high growth income potential are not inevitable. Needing food is. Making food and drugs using machines is likely. Most clothes, however are still made by hand because its cheaper than it is to use machines. Normal people can’t even afford housing or transportation. AI is not a nail and I will never be a hammer.
Too big to fail.