The AI Jitters (and three grimmer interventions)
An AI non-cause for worry, followed by causes for worry abroad, nearby, and everywhere
The Fed raised rates for no reason, and shortly thereafter, the market hit a new all-time-high. Both Nasdaq and the S&P peaked this week, although the S&P is barely visible against Nasdaq’s extraordinary tech-driven growth. Is it a bubble? Again, no, not really—if anything, stocks are relatively cheap, compared to their forecast earnings growth.
The AI Jitters (again)
The Fed raised rates for no reason, and shortly thereafter, the market hit a new all-time-high:
Both Nasdaq and the S&P peaked this week, although the S&P is barely visible against Nasdaq’s extraordinary tech-driven growth
Is it a bubble? Again, no, not really—if anything, stocks are relatively cheap, compared to their forecast earnings growth.
What is true, however, is that earnings growth is increasingly dependent on AI capex. It’s semis and energy that are doing all the growing, and a majority of that comes from hyperscaler investment in infrastructure, the demand for which is premised substantially on frontier lab investments in model training.
So, when the FT reported that OpenAI annualized revenues are $20B less than previously signalled, this is what happened next:
Nasdaq took a tumble, particularly some of the big dawgs in the AI infra supply chain, like Nvidia and Oracle, presumably because people interpreted the OAI “undershoot” as a bearish signal for ongoing demand for compute.
Well, that’s certainly Mr. Market’s prerogative, but (a) there was no “undershoot,”—$50B reflects revenues net of the share that goes to the cloud partners, while $70B reflects gross revenues, and the (widely reported) inside-baseball explanation for this is that Anthropic had been counting the latter, while OAI had been counting the former, but since the latter is a higher number, OAI said ‘hey we can do that too, but also this is the real number.’
But. more importantly, (b) at $50B in ARR, that makes OAI ~20% bigger than Salesforce, a scale that it has achieved in a fraction of the time:
$50B and growing 229% is just not a thing that Mr. Market sees every day.
This is broadly the entire universe of tech companies, by multiple and NTM growth:
The vast majority of them do not generate $50B in ARR, and basically none of them are growing more than 50%, let alone 229%. (Those little dots off to the right are neoclouds, which are growing quite rapidly off a small base, serving AI compute.)
Will there be trials and tribulations along the way? Surely.
But the hype is more than just hype, and at these levels, there’s bound to be some nosebleeds, but people react to every little wiggle like the world has changed. Oh no, Ramp’s token spend data shows some deceleration!
Yes, it’s true, and it’s worth noticing that July’s upward inflection appears to have mean-reverted to January’s upward inflection, which is still an astronomical pace, even relative to the previous January’s pace, which was stilly pretty fast, but now looks paltry in hindsight.
Random Walk has about as much certainty of the future as anyone else (which is to say, very little at all), but I maintain my increasingly less-cautious optimism.
1. Europe Needs An Intervention
Random Walk has made passing references to Europe’s fade before.
Let’s just start with the facts.
- It’s an under-appreciated tragedy that the once great civilization is rapidly devolving into a quasi-totalitarian mausoleum, entombed by life-draining regulators, and overwhelmed by catastrophic immigration policies.
- Western European leadership busies itself with scolding, rule-making, and persecuting wrong-think in a feeble effort to keep the peace just long enough for aging Europeans to consume their pensions before they finally die-off forever.
- If there are any history books written to remember them, they will likely be written by the descendants of the scolds (under the watchful eye of the restored and embiggered Andalus), who will blame the “far right” for ruining everything.
That’s the base case, at least, but it ain’t over til it’s over.
Anyways, here’s some more tragic data on that front, this time courtesy of Stripe (led by two fearless Irishmen—perhaps the last two—for what it’s worth).
First, Stripe asked the European founders on their platform whether they would reincorporate in the same country again, and the results are a staggering signal of no-confidence:
The majority of founders in the UK and every EU country (in the table at least) would take their business elsewhere, on the second turn.
The people who have built businesses in Europe apparently know well enough to not make that mistake twice.
Consider also the relative performance of European AI businesses (on Stripe):
The y-axis is unlabeled, and it’s not obvious what makes an AI business, but you get the general idea: if Stripe is a proxy for innovative business formation, and AI businesses are a proxy for the next wave of innovation . . . then the EU isn’t just falling behind, it has barely left the starting gate.
That’s bad. Very bad. Compounding exponentials take no prisoners.
As Peter Thiel recently quipped (to paraphrase): Of the 50 richest Germans, 20 are Gen X or younger, and all 20 inherited their wealth. In the US, 12 of 50 are GenX or younger, but 9 of 12 made their wealth, while only 3 inherited it. What’s prosperous in Europe now is a function of what came before, and what’s happening now, is a signal for what’s next.
All the evidence suggests that it’s only going to get worse for Europe.
The new French malcontents are rioting in the streets (not that any mainstream media outfit would tell you), and a debt crisis appears to be underway.
There will be pockets of good cheer, of course—ASML still exists, sophisticated manufacturing and biotech still exists, there are still millions of talented hard-working people, Klarna happened, Adyen happened, Spotify happened, GLP-1s happened.
But what’s also happened is that China decided it didn’t need anything from Europe anymore:
In the bigger scheme of things, Chinese manufacturing has eaten their lunch, the next generation of Europeans will bear little resemblance to the prior ones, and the iron-gripped technocracy appears steadfast on its commitment to snuff out whatever remains of European verve and ambition.
If there’s evidence to the contrary, I’m all ears, but I don’t see it.
At the very least, there’s a war going on:
German manufacturing has been saved by a thing it has not yet ceded to China: defense goods.
It’s not an outcome that Random Walk is rooting for, but it’s increasingly hard to see how it may get much better than that. Or, at the very least, it will require some big changes, and turning the Titanic ain’t easy.
2. Beantown Needs An Intervention
It’s not only Europe that needs an intervention.
Random Walk has also made passing reference to the Curley effect, by which leadership deliberately exploits the out-group in favor of the base, with the effect of driving the out-group elsewhere, while deepening the loyalty and dependency of the in-group, all of which leads to a consolidation of electoral power that is ruinous in the longer-term, but not quite so ruinous that it can’t last at least a few decades.
The Curley effect is named after infamous Mayor Curley of Boston, a great city he once plundered, which once again appears to be pursuing a similar game plan.
Recall for a moment that net-migration tells only a partial picture. What matters isn’t just the number of people coming and going, but also their ages and incomes.
When it comes to Boston (and DC and Seattle), the picture is grim:
Boston has been a big net-loser of people and incomes over the past two-years.
Fortunately, NY appears to have staunched the outflow of incomes, but is still losing people.¹ Miami, by contrast, is a net-loser of people, but a big gainer of incomes, which is an entirely different set-up: Miami is becoming more exclusive, for better or for worse.
But, back to Boston.
Boston, of course, is home to some of the finest research institutions in the world (even accounting for recent decay), and some of the finest collections of talented young people in the world, to boot. The sports fans are bad, the drivers are terrible, and the weather is an acquired taste, but Boston has every reason to be on the perpetual up-and-up.
And yet, it’s not that. It’s the opposite of that.
Massachusetts has experienced net-negative employment growth since before the pandemic:
It didn’t start out that way. MA employment recovered in-line with the national trend until ~2023, when it diverged substantially downwards.
Whitecollar work, particularly, in the sciences, has totally stagnated:
MA had been making a lot of progress in R&D services, but again, since 2023, that trend has flatlined.
The Boston Fed offers a few other examples of regional malaise.
Elsewhere, MA’s share of VC capital and new research institutions are both in sharp decline:
VC activity has tapered off since ‘23 (when VC activity tapered off more generally, but there has been no AI recovery), and MA’s share of new research facilities has dramatically reverted to trend.
Indeed, over the longer-haul, MA has been a business graveyard:
Net business formation in MA has been sharply negative to the tune of ~17,500 businesses lost over a 9 quarter period (which was the worst in the country).²
So the picture for MA is not the best. Despite some enormous embedded advantages, MA is losing jobs, people, incomes, and businesses, in fairly dramatic fashion.
In terms of what’s going on here, surely some part of this has to do with the change in interest rates. Plus, MA had a surge in new biotech facilities, so it’s not surprising that that would taper somewhat. And while the administration attempted to curtail federal funding in response to the universities having lost their way, a MA judge decided to protect the home team and restore that funding, so you can’t really blame the NIH.
In the bigger scheme of things, it would just seem like whatever the Bay State is doing (or has been doing) is not particularly good for the Bay State. Random Walk is no expert on Massachusetts politics, but even a cursory search of Boston’s Mayor Wu depicts someone apparently deadset on eroding quality of life in MA’s biggest city, and further diminishing the city’s extraordinary advantages she inherited.
She should stop.
3. We all need an intervention
Just riffing on the theme of things that are somewhat grim.
AI unto digital media, as digital media unto print
It really starts with a fun observation about what’s happening to web traffic, particularly for news sites:
Monthly traffic for news has plummeted by ~20% since the beginning of 2025.
Now, webtraffic data is pretty noisy, so take this with a grain of salt, but it’s not crazy to think that site visits are declining, particularly as (a) social media continues to ascend; and (b) AI is summarizing the news for people, such that they don’t have to actually click through to the news. Indeed, there’s other evidence that “zero-click” searches have risen substantially, so it’s not just the news that’s getting hammered by AI summaries.
Now, there’s a certain irony, of course, if AI does unto digital news as digital news did unto print news. I’m not sure that that’s what’s going to happen, because AI summaries have to get their substance from somewhere, and content monetization is evolving all the time.
But, really, this is a segue into something else.
Smartphones make people feel informed
Putting AI aside for a moment, and returning to point (a) above, it’s certainly the case that social media has made substantial inroads into the business of keeping people “informed:”
73% of respondents to the Pew poll say their smartphone helps them be informed.
Smartphones aren’t the same as social media, but let’s just assume that they are close enough.
Here’s the thing. Other than in those circumstances where being misinformed has a fairly immediate and identifiable negative feedback mechanism, people generally do not want to “feel informed.” They want to feel, yes. They care about status, belonging, and their own sense of virtue, and all the feelings that go along with that . . . but informed? Not really. People like being entertained, particularly if it aligns with those just-enumerated feelings . . . but informed? What good is that?
To riff on Mencken, social media gives us what we want, good and hard.
Apparently we like to feel anti-something
And for whatever reason, whether it’s particularly true now or it’s always been true, people seem very primed for negative polarization—to be against “the thing,” (and to be aligned with others who are similarly against “the thing”).
And if that’s what we want, well that’s what we get, and it’s leading to some very costly beliefs about the world.
Take data centers for example:
In just six months, a third of respondents have graduated to a bare majority of respondents who believe manifestly silly things about data centers.
- Data centers are not uniquely bad for the environment, and they certainly do not consume an unusually high amount of water. Solar farms and wind turbines? Now those have some high environmental costs, but you see . . .
- The evidence for data centers raising energy costs is pretty dubious—there’s better evidence that data centers lower energy costs—and the hyperscalers, try as they might, can’t buy grid and transmission upgrades nearly as quickly as they’d like.
- Are data centers bad for nearby quality of life? I mean, it’s an ugly warehouse looking thing, which according to some people, has an electric hum. I can understand not wanting to live right next door to one, but we build warehouse-looking things all the time, and no one seems to object.
While data centers appear to be none of the things that people seem to object to, they are, on the other hand, the sole straw that stirs the drink in our uni-cycle economy. People may not like data centers, but they’ll really hate what happens without them.
The normally fairly astute Jason Furman posted this foolish missive the other day about the relative strength of data center construction spending:
That is decidedly not what crowding out looks like.³
That’s what it looks like when data centers are the only thing worth building in a world of higher rates, demographic headwinds to demand, and a moderate surplus of shelter (and a more than moderate surplus of office).
Take data centers away, and I can assure you, single family home building does not spring back to life.
Blame the screens? Sort of
Is social media driving this ignorance?
I think so, if only because that’s where people “get informed,” but also these splits are striking:
It’s true that no one (but Conservatives, and only barely) would prefer to live within 10 miles of a data center, but three cohorts stand out for their anti-AI everything polarization: women, 55+, and liberals.
There’s nothing intrinsic to AI or data centers that should uniquely impact those groups . . . and yet, here they are, sorted into their own tribes of luddites. It’s almost like some very sophisticated algorithm has given those people an extremely concentrated dose of whatever happens to make them “feel informed” on this issue.
Many people want to blame anti-AI sentiment on poor industry messaging. Don’t say things like “It’s going to replace all white collar work!” Don’t be so weird, Dario. Don’t be so evil, VCs.
That might be more persuasive, but it’s not just AI, of course, that’s been a target of polarizing drivel.
Random Walk can’t say for sure, but it certainly seems like both the speed and intensity of outrage cycles has accelerated substantially, of late. Whether it was BLM, #metoo, vaccines (both for and against), Israel, immigration, burritos, etc. everyone, everywhere is finding something to be profoundly stupid and angry about, in maximally polarizing ways.
Is social media to blame?
Again, probably not entirely—and first and foremost, if social media is to blame, it’s fault is being too good at giving us what we want—but if there was any mechanism that could rapidly sort people into highly engaged “information and belief” tribes, each with their own unique feed and illusion of consensus, and all of them negatively polarized, but otherwise totally agnostic as to the cohesiveness of the whole, then surely social media would be the first (and maybe only) guess.⁴
That isn’t to say that social media is bad, per se.
It has tradeoffs, and like most things, the costs and benefits aren’t likely to be normally distributed. The ungated discovery, reach, and speed of social media can be a massive unlock for some people—certainly, Random Walk has benefited from it, enormously. But, perhaps for other people (and perhaps most other people), the temptation to mainline a custom brew of infinite jest is hard to resist.
And it’s not like the prestige media is without its (many, many) faults—many of which have been laid bare by the ungated discovery, reach and speed that only social media can offer. But one thing that prestige media does have is a consensus-making function, which by necessity, has a broader base. The “news” may be varying degrees of (mis)informative, but at the very least, there are only so many different “newses” to choose from.⁵
All of which is to say is that if was a mixed up, muddled up, shook up world to begin with, that too is likely only just get started.
Footnotes
1 As a discomforting aside, net-new private sector job growth in NYC (ex-healthcare) is once again, nada:
The financial industry has done pretty well, and lower-income workers have experienced a raise, but the middle is otherwise hollowing out.
2 Apropos of nothing but data on firm formation, this was a fascinating chart from EIG on the “entrepreneurship” boom (or not):
While it’s true that there has been a huge surge in “high propensity [to hire]” firm formation, beneath the surface, there are reasons to doubt whether these firms will actually have a high propensity to hire. There is a large and unusual gap between HBAs and the two other measures of hiring propensity, CBAs (i.e. corporations) and WBAs (i.e. filed applications with planned wages), all of which suggests that whatever is going on, more likely sounds in side-hustle or solopreneur.
3 This, on the other hand, could be described as crowding out:
Consumers and companies both have relatively healthy debt:income ratios, in comparison to pre-Global Financial Crisis. What KKR politely neglects to mention is that the same cannot be said of the federal government, which has levered itself to the gills on our behalves. Or, at least someone’s behalf, but levered to the gills, all the same. Government borrowing has ballooned, while private borrowing has broadly receded.
4 Naturally, real diversity probably doesn’t help, either. I always find it amusing/disturbing when people pooh-pooh the possibility of negative externalities associated with rapid demographic change, particularly around social cohesion, and then later lament the apparent rise in hyper-partisanship. “Why can’t we agree on anything, and why do my cherished values appear to be on the wane?!” Idk…maybe because we’re literally more different than before?
5 That’s also perhaps why so many in the establishment have discovered a fondness for censorship—beforetimes, they never had to actively censor, because the form-factor was inherently censorious: distribution was costly, and access to distribution was highly gatekept. Now, not so much.