Is This Cycle Different? The People Wanna Know
The trillion dollar setup; some charts for AI cheer; of cultural malcontents; Against Dual Mandatism
As the market continues to vent its AI Jitters, I stumbled across this chart that is a near perfect encapsulation of where things stand.
1. Is This Cycle Different? That’s The Trillion Dollar Question
Historically (1950-2019), forward earnings multiples are always lowest at the peak of the cycle. That is why when EPS is already +30 over the past 12 months (as it is)—reflecting a year’s worth of strong cyclical upswing—forward multiples have typically been on the lower end, closer to 12x (as per the blue rectangle and orange circle). No one wants to buy last year’s earnings growth.
Right now, however, forward multiples are relatively high (~18x), which is very unusual following a period of already strong earnings growth. But what’s also unusual is that expectations for forward earnings are also still very high, which typically does not happen when earnings have already grown so much.
In other words, on a “cyclically adjusted basis,” prices are quite high . . . but only if you assume that this cycle is like the cycles that came before it. If this cycle keeps going in-line with guidance, however, then valuations will look pretty cheap, in hindsight.
And therein lies the crux of the thing. Mr. Market simply does not believe that this cycle will be different than the ones that came before it. Investors do not believe the forward capex guides. They do not believe that demand for compute will continue to support trillions of dollars of annual investment into compute, such that the unprecedented thing will be even more unprecedented than it already is.
2. Charts of AI Cheer (And Some Perplexing FUD)
Now, as Random Walk has laid out many of times before, there is plenty of good reason for skepticism (but also optimism).
That said, it does seem like the FUD (while not entirely unfounded) has gotten a bit head over skis.
Oh no, Anthropic is dead in the water, but so is Gitlab because Anthropic is going to eat its lunch:
Pick a lane, man, either Anthropic is toast, or Gitlab is, but both is pretty rich.
It cannot be simultaneously true that
- the prize of “winner takes most” for this world-changing tech is so large that the labs have no choice but to commit unsustainable amounts of capital on training new models;
- the natural price equilibrium of these uber-expensive models is a commodity-like number converging on $0; and
- in a universe where this world-changing tech is priced close to free, that no one will want to use it, such that compute will lay idle, the world over.
If the commodity version is good enough, then the labs don’t have to keep the arms race going. Good for them, good for inference, and probably less good for the chipmakers.
If the prize of being better is worth it, then why shouldn’t it attract sufficient capital to win? And upon winning, why would the natural price be commodity-like?
In other words, it’s hard to stitch together an argument where everyone loses. If it’s valuable and expensive to differentiate, then token prices cannot be commoditized. If tokens can be commoditized, then there’s no reason to keep up the arms race, but there’s also no reason that compute would stay idle. If the arms race is inevitable, then it can’t be true that tokens are a commodity (or that compute will stay idle, especially if prices are low).
Anyways, these are not fully developed thoughts, but just an opening for a brief dose of chart-optimism.
First, a fun little observation about expectations for AI lab revenues. It turns out that forecasters (of the Forecasting Research Institute) have chronically undershot actuals:
The high-end of estimates (from Sept. ‘25) for July ‘26 revenues was $25B. Actuals were ~2-3x higher than the high-end of the forecast.
Second, despite all the handwringing around token spend (derived from useful, but unquestionably biased samples), the story actually looks pretty good:
Total token spend (per Ramp) mean-reverted to its astronomical pace, and appears to be reaccelerating. That’s not so scary.
Token spend by cohort also shows no signs of slowing:
Again, via Ramp, for all of the top 1%, top 10%, and the median business AI spender, demand is going up (pretty steeply).
And the median business user spends only half of what the median paying household does, so there is plenty of room to run!
But about the Capex ROI? Will it ever be enough to recoup those trillions?
Well, that deserves a longer post, but using Goldman’s napkin-math for hyperscaler payback hurdles, and some napkin math projections of cloud-revenue growth . . . payback does not seem that far-fetched:
Using either current growth, or the growth of the backlog, with some assumed decay, hyperscalers could hit their 20% ROIC hurdles sometime between ‘27 and 28’ (and onwards and upwards from there)!
Will that definitely happen? Of course not. Does it account for the additional revenue that’s gotta be earned to justify all that cloud spend? Nope.
But, I’m outta time, and this was just a simple chartapalooza of cheer.
3. Of Bad Ideas And Cultural Malcontents
Continuing to riff on the theme on the rising prevalence of high salient youthful folly—because again, youthful folly is ubiquitous throughout history, but what’s different now, is that everyone apparently has to care —it seems hard to overstate the role that Higher Ed, Inc. has played in getting us to this point.
Accept the premise that at least some substantial part (if not all of it) of this youthful discontent is status-driven.
People are upset, not because their material conditions have faltered (because they haven’t), but because they do not perceive their relative status gains, either now or in the future, to be sufficient to meet their expectations.
One reason for that disappointment (in addition to other reasons like white collar stagnation, plus social media’s erosion of local maxima), is that we’ve persuaded a generation (or more) of people that if they go to college (or beyond), then they will unlock a sure-fire path to some high(ish) status vocation—either as a scholar, a corner office, or some variant therein—that will not only align with their “passions,” but also bring sufficient pecuniary rewards to partake in the finer things that our $30T economy has to offer.
That claim has been quite beneficial to the provisioners of this golden unlock (i.e. Higher Ed, Inc.), who have happily gorged on the trillions of borrowed dollars that K-Street ‘bankers’ have facilitated to underwrite the cost of this supposedly pedagogic endeavor.
That claim has also been demonstrably false, as it turns out, the market is rife with credentials that are worth less than the cost to have obtained them. And, of course, it seems hard to imagine a scenario where the Lake Wobegone offering could ever have been true, but in all events, it’s not true, and therein lies the present predicament.
All of which is just a windup to introduce the following two charts.
That downwardly mobile college grads are at the heart of this thing is visualized quite nicely by this Nate Silver chart on the “very liberal” electorate:
If you want to know who is likely voting for the socialist ethno-religious supremacists, with their free healthcare and exotic stick-it-to-the-man joie de vivre, it’s the low-middle-income college-and-beyond types, more so than anyone else.
It’s of course possible that the downwardly mobile higher educated have joined the revolution for reasons other than disappointment with the “Academy’s” false promise of higher status, but Silver’s heatmap is just so on the nose.
Here is chart the second.
That Higher Ed has known this whole time that it was unsustainably peddling lemons is reflected in its own hiring practices:
The only growing category of help that 4-year colleges have deigned to hire is “graduate assistants,” (and it’s not even close).
Graduate assistants are poorly-paid, untenured ex-grad students, who as grad students, provided graduate programs a reason to exist, but upon completion, have nowhere to go from there, so they become poorly-paid, untenured teachers to the next generation of rubes, in the dim but fading hope that a tenured faculty role (or the private sector) will rescue them. OK, that’s a bit unkind, but it’s still pretty much correct.
It’s of course mechanically true that not all grad students can become tenured faculty (unless the overall universe of the educated grows in-kind, which it cannot, in a population that does not grow). But it’s hard to think of a reason that graduate assistants would be the only net-new hire, if there wasn’t a glut of ex-grad students with nowhere else to go.
But rather than respond to the obvious glut of grad students by bringing these programs to heel, Higher Ed has used them as low-cost overhead to keep the rest of machine coughing and sputtering along.
And yes, it would seem that these are some subset of the people disproportionately fond of that new breed of “elite.”
4. Against Dual Mandatism (a rant)
There are many people, it seems, who strongly believe that if inflation is high, the Fed must raise rates. This claim sometimes gets paired with the notion that “AI is crowding out” other investment, and while no one seems to think that raising rates will change that much, there’s a guilt-by-association implication that “AI is overheating,” and as we all know, “overheating” is a bat signal for the all-powerful Fed overlords to raise the rates, and save us from ourselves.
QED the Fed must raise rates.
Random Walk finds the first claim deeply puzzling (as laid out many times before), and notes that the second claim is almost certainly false, which—when combined with the guilt-by-association bat-signal maneuver—makes me doubt whether these people are thinking clearly, at all.
No, the non-growth-to-contraction of private construction that preceded the AI boom (and definitely preceded the AI borrowing boom) is not an “AI squeeze out.” AI Capex is the only thing that’s bumping because it’s the only thing worth bumping when rates are high and demographics are pointed the wrong way.
Back to the story, it’s only fair to point out that the notion that Team Hike are thinking dogmatically rather than clearly is my prior, so this is not a hard conclusion for me to reach.
As Random Walk readers know, what is to be done about inflation depends on the why of inflation. While monetary causes ought to beget monetary solutions, if, otoh, inflation has nothing to do with monetary policy, then it makes little sense to throw sand in the gears, just because.
Good luck stopping this inflation with higher rates.
Ah! But the inflation! The dual mandate! The overheating!
Let’s assume for a moment that raising rates will lower inflation and it will do so without causing a recession. The rejoinder still makes little sense.
- First of all, the Fed actually has a triple-mandate: stable inflation, full employment, and stable long-term rates (more on that below).
- Second of all, private sector can’t really “overheat,” on its own. Those markets are self-correcting, and the cure to high prices is high prices. Whether and how soon that cure is reflected in measured CPI may take some time, but if you’re worried that AI investment is a bubble, then presumably you think a correction is in the offing, and if not, then AI shouldn’t be inflationary in the longer run, but in either case, no one needs the Fed’s help (let alone the Fed’s hurt).
- Third of all, the only clear and obvious overheated part of the economy is the fisc. Raising rates will not moderate the fisc’s overheated consumption. At best, a rate hike will moderate consumption in the properly heated (overcooled, even) parts of the economy, which will only make the fisc’s overheat worse (to much worse).
In other words, the argument for hikes, properly stated, is that “the Fed is doing it’s job successfully, because by deflating everything else to make room for the expanding inflationary balloon that is Uncle Sam, over which the dual-mandated Fed, has no control, but since it must control inflation, it must burn as much healthy growth that is necessary to accommodate cancerous fiscal expansion (which, in the end, we will be unable to control, but this is nonetheless not futile).”
It’s not a very compelling argument, and of course it’s not the argument that its proponents would actually make, because I do not think they have thought it all the way through. It’s “prices up, therefore rates must go up” dogmatism, and the rest is just details.
Who is crowding out whom?
Perhaps more importantly, there is no reason to assume that raising rates will lower inflation, as opposed to simply make matters worse.
Right now, the best estimates of the added interest expense for every 10bps of hikes is ~$10B/year (but rising over time, as public debt reprices, and the interest compounds the deficit). The best estimates of the decreased consumption for 10bps of hikes is ~$5-7B/year, assuming that already depressed rate-sensitive consumption can be depressed even further.
$10B more in exchange for $5-7B less. In other words, the net impact of raising rates is to increase overall spending.
OK, sure, but the interest to bondholders doesn’t flow through to consumption 1:1 . . . a lot of it just gets reinvested into more debt.
And you say that like it’s a good thing? “More debt,” is why we’re in this inflationary mess in the first place. So, not only is a rate hike net-stimulative, it’s compounding the deficit, i.e. the very thing that’s driving “excess demand.” You cannot paper-over the inflationary impulse of higher rates by saying “oh don’t worry, it’s just more paper.”
And even if you could, everyone would agree that, at a certain point, if net-new interest payments were that much larger than decreased consumption, they would be inflationary (because even if the pass-through is something less than 1:1, the pass-through is not zero). That’s roughly what “fiscal dominance” means, i.e. every monetary attempt to reduce inflation creates an even larger fiscal inflationary impulse.
And since it’s hard to imagine how compounding the deficit while slowing private consumption does anything but push net-new interest payments even more quickly up-and-to-the-right, then what you’re really saying is “oh don’t worry, we’re not in a fiscal dominance regime yet, so that means it’s totally fine and sensible to accelerate the path towards fiscal dominance, because that is what the dual-mandate requires (however futile this endeavor will eventually be, on a timeline that we have now moved-up).”
Again, not an argument that anyone would actually make, and yet, it’s not not an argument that they are making, either.
These are smart people saying these things, so it’s hard to explain these claims as anything other than rooted in dogmatism, anchored as it is to a world before deficits, in which even then “prices go up, rates go up” was not mechanically true, but was perhaps more likely to be true than it is now, when it’s definitely not true.
In the end, the most likely outcome for all of this is that the Fed doesn’t hike much anyway, the transient price impact of Russia’s war (which has likely replaced the transient impact of Iran’s war) will fade, as will the transient impact of AI Capex on PC and device components, and we’ll return to the slightly-above trend inflation we had before (unless and until the AI bull case kicks in), and everyone will say “see, we were right, rate hikes were good,” and we’ll all have learned nothing.
Disclosures
Random Walk is an idea company dedicated to the discovery of idea alpha. It’s not investment advice, conflicts of interest abound, it’s totally unreliable, and it reflects the thoughts and ideas of Random Walk, only. Find differentiated data, perspectives and people, and keep your information mix lively. A foolish consistency is the hobgoblin of small minds. Fight the Great Idea Stagnation. Join Random Walk.