....................... This AI-related infrastructure boom is larger than what we saw during the telecom and broadband buildout in the late 1990s, and about as large as the railroad boom back in late 1800s. With the railroad boom, the companies built too many lines too fast, creating duplicate capacity. These sat idle for a while (hence the crash after the boom), but eventually became useful once again as the economy expanded. With the telecom boom, companies laid a lot more fiber optic cables than was needed back then (and so these companies lost money), but this “dark fiber” became useful in later years, especially after all of us started streaming cat videos on Youtube and later spent evenings at home with Netflix. The pattern we’ve seen has been:
- A massive capex boom
- Overbuilding
- A shake-out, leading to a crash
- Long-term benefit
The question right now is whether the current generation of data centers will actually lead to a long-term benefit. One big difference is that the lifespan of this AI infrastructure is much shorter than say railroads and broadband fiber optic cables—datacenters that are deployed with the latest, most sophisticated chips (which are used to train AI models and provide responses to our AI prompts) are likely going to be near obsolete in 3-4 years as chip technology advances. The depreciation of these assets happens much faster, in contrast to what we saw with railroads or telecom. (“Moore’s Law,” first stated in 1965, noted that the number of transistors in an integrated circuit had been doubling about every two years. That observation has continued to hold roughly true the last 60 years.)
Another difference is that there are only a handful of players are involved in the current AI boom, since only these large companies have the funds and scale to operate at these levels ...........
"At any moment of time, there are myriads of feedback loops at work, some of which are positive, others negative. They interact with each other, producing the irregular price patterns that prevail most of the time; but on the rare occasions that bubbles develop to their full potential, they tend to overshadow all other influences..” ~ George Soros
Summary: This is still a market you want to be long and buying. Sure, things are a bit overextended over the short term, and we could see a pullback soon. But we expect any pullback to be mild and an opportunity to add risk. Positioning and sentiment largely remain offsides (w/ CTAs & Vol control the sole exception), and there’s still lots of room for both to drive risk assets higher. Bonds and commodities remain in major compression regimes, suggesting BIG trends are coming. Our growth leads tell us reflation, not stagflation, is on the way .......
Below are the forward returns following a liquidity reading of 90%+. Gold was higher 80% of the time over the following 12 months, with average gains of 16%. And the SPX was higher 90% of the time, with average gains of 14%.
.......... Our Market Implied Macro Regime indicator isn’t currently pricing in significant odds for a specific regime. But I suspect we’ll see the probability of an “Overheating” regime rise over the coming months.
There’s A LOT of market data out there.
So much, that at any given time you can find a chart, stat, or study to back up whatever story you want to tell.
That’s the real enemy in this game…the bias we bring to the data.
The antidote is simple but not easy: weight the evidence.
Don’t stop at finding a data point that matches your opinion.
Stack the signals, count them up, and let the majority call the play. .........
I’ll always rip through hundreds of charts each week as it gives me a feel for the market that no model can replicate.
But I continue to added more composite frameworks to weigh the data I see with my eyes.
These frameworks don’t just spit out noise; they filter, weigh, and separate the rules from the exceptions.
............. Offense is on the field….ACT LIKE IT!
We can simplify the value we receive from our stocks into three simple categories:
- the cash we receive today (dividends and/or buybacks)
- the cash we receive tomorrow (growth in that available free cash flow)
- the change in the market’s valuation of that cash flow (e.g. the P/E ratio rising, or “multiple expansion”)
I call these the three engines of value, and any stock’s appreciation is a simple mathematical result of these factors.
Since the value we receive from these categories (whether measured in dollars or percentage returns) are fungible, I find it odd that investors tend to prefer prioritizing one of these engines over the other. All three can drive results.
“Quality” investors tend to prioritize returns on capital and the long term runway for growth, but ignore the impact that a shrinking multiple has to a stock’s performance — even the long-term performance. Costco is a world-class business, but trades at 55 P/E. Let’s say in a decade, the company replicates the exact same growth rate it had last decade (no easy feat given the larger starting base). If it trades at a reasonable 25 P/E (near its historical average), shareholders will earn just a 4% total annual return including dividends over the next 10 years. And if this multiple happens to come revert to its average in 5 years, the returns become negative over that period. A high enough price can lead to a risky investment even for the highest quality companies.
I think saying things like “we only invest in quality companies” is often a statement that does more harm than good to a portfolio over time, because it too often leads to the investor paying too much for these businesses collectively. The result here won’t usually be disastrous, but is often subpar.
“Value” investors often prioritize multiple expansion, and thus look for catalysts that might change the market’s perception of the business, causing its P/E ratio to rise. One thing growth investors might underestimate is that a rising P/E ratio can lead to a big tailwind even over a 10 year period. A stock that rises from 10 P/E to 25 P/E over a decade adds 10% per year to the annual return from the P/E ratio alone. If this company grows at just 4% and can pay out half of its earnings as dividends, the total return for these three engines will equal approximately a 20% annual return over a 10 year period. Not bad for a mature, 4% grower.
I see this undervalued-mature business theme play out frequently in large cap stocks when a business hits a rough patch. United Health Group (UNH) is a prime example right now; the health care industry is a large and important industry that I believe will be larger and just as important a decade from now; UNH has a wide moat in that industry and has issues which appear fixable over time, even if they take a few years. Ben Graham once said that the great thing about unpopular large caps is the market corrects itself very quickly if the company’s issues get resolved. I believe this could happen with UNH.
I also see this occur when a business has no specific issues but the industry is facing a headwind or perhaps there are fears of a recession. e.g. JP Morgan (JPM) is up nearly 3x in the last 3 years, a 44% CAGR for perhaps the world’s most well known bank. It was mispriced 3 years ago not because of company problems, but rather the fear that a recession would dent earnings in the near term. Stocks overreact to near term pressures even when those pressures have little impact on the total amount of future cash flows. JPM’s mispricing corrected itself rather quickly (as often occurs, as Ben Graham observed), but the results still likely would have been excellent even if the result took 5-10 years instead of only 3.
So, paying low prices obviously creates better long term results, but just as growth investors sometimes extrapolate current growth rates too far and almost ignore the P/E multiple; value investors sometimes spend too much time thinking about the multiple and not enough time on the value that can come from the long-term business results. One of the best features of a cheap stock is a high FCF yield, and as long as the stock stays cheap, the yield remains high and with proper capital allocation, this low growth cheap stock can turn into a compounder.
......................... Division by zero is known as a “singularity.” It’s the point where equations break down, values become “indeterminate,” things stop working normally, and variables shoot toward infinity and suddenly collapse on the other side.
Black holes are called singularities, because the normal equations of physics become meaningless – things like density and curvature go to infinity.
I’ve never liked division by zero. In college, it was a bane of my existence, particularly in computer programming class. Aspiring geeks like me would be huddled in Northwestern’s Vogelback Computing Center, which could easily be mistaken for a bomb shelter, and you’d wait half an hour for some bored work-study student to push your stack of punch cards, wrapped in a printout, into your slot amid a beehive of mail cubbies, only to find a crash log of cascading “Fatal error!!” messages because you forgot to assign a variable and the computer tried to divide by zero.
I started my own investment company in 1985, and over the next 25 years, the relentless pursuit of geekiness was wonderfully rewarding across market cycles that included both bubbles and crashes. Those admirable outcomes hit an excruciating roadblock (which we’ve since resolved) when another singularity emerged in 2010.
That’s when Ben Bernanke introduced a full-throttle, hair-on-fire, “divide-by-zero” singularity into the U.S. financial markets.
Under Bernanke, the Fed recklessly drove the quantity of zero-interest Federal Reserve liabilities from just 6% of GDP in 2008 (a level historically associated with T-bill rates of roughly 4-6%), past 10% of GDP (a level associated with T-bill rates between 0-2%), past 14% of GDP (at which point T-bill yields hit zero), all the way to 25% of GDP by the time his term as Fed Chair ended. The expansion continued – abetted by Janet Yellen and initially by Jerome Powell – until the quantity of zero-interest base money created by the Fed peaked at a deranged and unprecedented 36% of GDP in early 2022.
The Fed created these zero-interest liabilities through “open market purchases” that took interest-bearing Treasury securities out of the hands of investors – and replaced them with zero-interest cash (mostly held by investors in the form of bank deposits).
Somebody had to hold these zero-interest Fed liabilities, in the form of cash, at every moment in time. Cash doesn’t “turn into” something else when you use it. It just goes into someone else’s hands, someone else’s bank account.
Nobody wanted it. As the Wiggles song goes, “Hot potato, hot potato.”
Several years of “quantitative tightening” have gradually brought Fed liabilities down to about 22% of GDP. These Fed liabilities primarily take the form of bank-reserves, currency in circulation, and “reverse-repurchase” agreements with money-market funds. The Fed would have to cut its balance sheet by another two-thirds, to roughly 7% of GDP, in order to achieve T-bill rates of 2-4% by market forces alone.
Given that Fed liabilities remain vastly above that level, the only way the Fed can hold short-term interest rates above zero at present is by explicitly paying interest to banks on the trillions of dollars of reserves it has created. Currently, the Fed pays banks interest on reserve balances (IORB) at a rate of 4.15% annually.
The chart below shows how all this works. It’s our version of what economists know as the “liquidity preference curve.” ................................
What will burst this bubble? (and based on the gap between surveys of investor expectations and cash flow growth rates and yields, I emphatically believe we’re in one). My view is simple, and it’s reflected in last month’s comment:
The bubble hasn’t burst yet because investors haven’t quite yet recognized that the highest valuations in history imply the lowest expected future returns in history. A market crash is nothing but risk-aversion meeting a market that’s not priced to tolerate risk. Every fresh record high in valuations amplifies the likely downside when that occurs, but examining the collapse of past bubbles, the “catalyst” typically becomes evident only after the market is in free fall. ....................
The chart below shows our most reliable gauge of market valuations in data since 1928: the ratio of nonfinancial market capitalization to gross value-added (MarketCap/GVA). Gross value-added is the sum of corporate revenues generated incrementally at each stage of production, so MarketCap/GVA might be reasonably be viewed as an economy-wide, apples-to-apples price/revenue multiple for U.S. nonfinancial corporations.
The current level of valuations is the highest in U.S. history, easily exceeding both the 1929 and 2000 extremes
The chart below shows the same valuation measure on log scale, versus actual subsequent S&P 500 nominal total returns over the following 12-year period, in data since 1928.
........... Look. I don’t want to criticize the well-intentioned work of others, so I’ll leave it there, but it’s important to remember that there’s a mathematical relationship between prices, cash flows and subsequent returns. It’s not surprising that theories promising “the price impact is perfectly long lasting” will become popular at exactly the same moment valuations have pushed to record highs. As Galbraith wrote about crowd psychology preceding the 1929-1932 collapse, “It was still necessary to reassure those who required some tie, however tenuous, to reality.”
So here we are, at the most extreme point to-date in what I continue to view as the third great speculative bubble in U.S. history. In my view, the appeal to an “inelastic market hypothesis” is an attempt to inject some sort of theoretical formalism around the idea that no price, no valuation, is too high. .......................
There are certain instances across history when the central feature of investor behavior is an “increasingly urgent impulse to buy the dip.” This semi-frantic investor behavior is typically associated with fluctuations that have progressively smaller ranges and progressively steeper slope.
In the book, “Why Markets Crash,” Didier Sornette described these dynamics in terms of what he called a “log-periodic bubble.” I’ve referenced Sornette’s work a few times in recent years: within days of the February 2018 peak (which was followed by a quick market loss of 10% and a secondary decline later in the year that brought the overall loss to nearly 20%), late-2021 a few weeks before the January 2022 market peak, and again in February of this year, within a few days prior to the interim peak that preceded current extremes.
Sornette offered this description of the market dynamics of bubbles in their final phase. What I’ve described as an “increasingly urgent impulse to buy the dip” is what Sornette describes as a “finite time singularity.” ....................
I’ll say this again (and again): nothing in our investment discipline relies on a market collapse, and no forecasts or scenarios are required. But my preference is to share my work, and interesting things I’m looking at, for free, despite having bottles thrown at me by drunk speculators like it’s a scene out of the Blues Brothers.
On a related observation, we’ve now gone 100 days with the S&P 500 above its 50-day moving average. That’s not our ideal situation, since even the hedging implementation we introduced a year ago prefers some amount of fluctuation to a diagonal, hypervalued advance with an increasingly narrow daily range.
For better or worse, these instances tend to be impermanent, and once you get beyond about 90 days without breaching that 50-day average (assuming investors are at least somewhat bullish), the market typically hits an air pocket of at least 4-6% within the next few weeks. It can be deeper of course, and it’s also possible we get nothing. My guess is 4-6% straight down, which is the standard outcome. That might be a starter, or it might be all we get for a while. Fortunately, no forecasts or scenarios are required.
................. Moreover, it’s taken a multi-year bubble, and an advance to the highest valuations in U.S. history, simply to bring the average annual nominal total return of the S&P 500 since the 2000 market peak to 7.9%.
QOTW:
No comments:
Post a Comment