Good morning. Apple appears to be the latest non-finance company to discover that finance is hard. The device maker announced yesterday that its buy-now-pay-later service, Apple Pay Later, was shutting down after less than a year and would be replaced by instalment loan partnerships with credit card companies. Good for them. I can’t think of a major non-finance company that has diversified into finance that didn’t live to regret it. If you can, email me: [email protected].
US markets are closed tomorrow to mark Juneteenth. Unhedged will be taking the day off, and we will be back Thursday.
Federal liquidity
About a year ago Unhedged wrote several pieces about what might happen to markets when the federal government was no longer forcing cash into the financial system. We focused on a widely used measure of federal liquidity: the sum of the Fed’s balance sheet and its pandemic-era Bank Term Funding Program (BTFP), less the Treasury general account and the Fed’s reverse repo programme.
The logic for this proxy goes like this. When the Fed buys a bond and places it on its balance sheet, that pushes cash into the financial system; the same is true when it provides funds to a bank in exchange for collateral through the BTFP. When the Treasury general account (the US government’s checking account) rises, that represents government revenue collected but not disbursed, an absorption of liquidity. The Fed repo programme, in which the central bank swaps its securities for cash overnight and pays interest on that cash, is specifically designed to absorb excess liquidity from the system in order to keep short-term interest rates within the target range. Here is a chart of the liquidity proxy and its constituents over the past five years (I show the liquidity-absorbing elements as negative numbers):
Focus first on the federal liquidity proxy, the mid-blue line, which is the sum of all the other components. It peaked at about $7tn at the end of 2021, driven up by the Fed’s bond buying (the dark blue line), and fell to $5.8tn by early 2023, as the Fed let bonds roll off its balance sheet and absorbed liquidity with reverse repos (the green line). Since then, federal liquidity has ambled sideways, as rapidly declining repos have released cash, offsetting a smaller Fed balance sheet and a rise in the Treasury general account.
It is natural to ask if the Fed is really doing quantitative tightening if the bond roll-off (a cash absorber) is offset by falling reverse repos (a cash releaser). In theory, the two programmes target different ends of the rates curve: QT the long end (letting long yields rise now that the pandemic emergency is over) and reverse repo the short end (keeping it high, in line with inflation-fighting Fed rate policy).
That’s a debate for another day. The reason I go through all this tiresome arithmetic today is that through much of 2022 and 2023, people were pointing at the liquidity proxy and noticing that it had a close correlation with the stock market. Here, for example, is a chart Unhedged published in the middle of last year:
Correlations like this are seductive, because they suggest that the market is determinate and its movements can be explained with just a few variables. But it isn’t, and they can’t. The market was probably influenced by, but was never determined by, government money printing. The chart below, which tracks both the S&P 500 and the liquidity proxy, indexing them to 100 as of mid-2019, illustrates this:
Since October of last year, the S&P 500 has been rising sharply and the liquidity proxy has been heading sideways. The pure “money printer go brrr” theory of the US stock market is therefore disproved. One might argue that the baton has been passed, and liquidity has been replaced by deficit spending, AI hype, excitement about rate cuts, or something else. The point, however, is to avoid simple theories of the stock market. They always disappoint you in the end.
The increasing capital intensity of Big Tech
In his Base Hit Investing Substack newsletter, John Huber of Saber Capital has made an important point about the Big Tech companies that account for so much of the US stock market’s recent increases. Stated with maximum generality, Huber’s suggestion is that several of the Big Techs are massively increasing capital expenditures, and this might have negative implications for their future growth and profitability. Here is his killer chart, which simply shows capital spending at Meta, Google and Microsoft:
The monster 2024 year is just an estimate, but it aligns closely with what the companies have said about their plans.
In slightly more detail, Huber’s argument goes like this:
-
Capex at these companies is rising fast, mostly due to the AI arms race
-
This increase will hit earnings eventually, as depreciation expense rises, and it will hit free cash flow immediately; so be conservative and use price/free cash flow not price/earnings multiples to value these companies
-
Higher investment is not a bad thing if the returns on the investment are good, but it’s not clear what the returns on these large AI investments will be
-
Growth in invested capital at these businesses is accelerating quickly, and all else equal that means lower returns on invested capital and lower earnings growth in the future
-
These are great companies, but may only be average investments at the current price
I think Huber has put his finger on something very important, especially with points 3 and 4. I would frame point 3 slightly differently, by asking what the competitive advantage these companies’ investments in AI will create. All three of them have uncrossable competitive moats in their core businesses. Will they throw vast sums at AI only to discover that AI services (whatever they turn out to be) are a vicious competitive battlefield?
On point 4, I would note that the three companies he focuses on are all investing heavily, but there are important differences among them. If you look at capex as a percentage of sales, while each dollar of sales is becoming more capital intensive at Meta and Microsoft, this is not true of Alphabet (the difference is that Alphabet has grown sales faster than Microsoft, and ramped up Capex more slowly than Meta)
It is also important to remember that despite huge increases in capital invested over the past decade (Meta, for example, has seen invested capital increase fivefold) returns on that capital these companies have remained high and, in the past few years, even improved:
The Big Techs have scaled profitably so far, in other words. The question Huber rightly raises is whether the AI wars — a new business for all of these companies — might change that. Huber, in other words, has turned AI hype on its head.
One good read
America gets the flows.
Read the full article here