Prepare for Bubble Trouble

All the disquieting signs have been hiding in plain sight. Anybody who has lived through previous market manias felt the ominous presence of deja vu.

The problem is, many people in the market today have no memory of how manias end. They’re too young, too experienced, including some cast and crew at VC firm Andreessen Horowitz. Earlier this week, in an opinion piece published by MarketWatch, Brett Arends had some cheeky fun at the expense of those callow youngsters who’ve never gone through a chastening market cycle. Here’s an excerpt from that piece:

Wall Street manias and crashes, sadly, are about as variable as a Scooby-Doo episode. Some of the names and details change, but the fundamental story is always the same. Are you still astonished that the ghost or the mummy or the skeleton or the monster that’s been chasing Shaggy and Scooby through the haunted house turns out to be the amiable old guy we met in the opening scene? Or that he would have gotten away with his dastardly plot, too, if it weren’t for these pesky kids and their dog?

If so, I have some very exciting leveraged technology ETFs for you!

Some people trace the stock-market mania story back to the railroad boom in the 1800s, the South Sea Bubble in the 1700s or the Dutch mania for tulip bulbs in the 1600s. But even a few thousand years ago Ecclesiastes was warning us that cycles repeat, and “there is no new thing under the sun.”

Ecclesiastes, scourge of hubris and vanity, would have a field day with the frothy AI market.

Questions Abound

With the passing of each day, we seem to be heading toward the diminution, if not outright implosion, of what is irrefutably an AI bubble. The principal questions, as always in bubbling markets, are when the bubble will deflate (or burst), what event (or events) will trigger the implosion, how severe and broad the economic costs will be, and how quickly the market will readjust to a post-bubble equilibrium.

There will be other questions, too, such as who will get the blame for the fallout, which stocks will suffer the most, how many AI-related companies will cease to operate independently (and who will they be), and which companies will emerge as long-term winners. Place your bets on the prediction markets, if you don’t mind being occasionally fleeced by insiders with privileged information.

The truth is, we cannot know the precise answers to these questions because they traffic in future events, and future events are always governed, to greater or lesser degrees, by contingency. You can run the numbers, assess probable outcomes, establish odds. All of that, however, resides in the realm of parimutuel wagering, which brings us (and you) back to the prediction markets.

Remember, though, that a parimutuel market is a gambling den. You can win, sure, but eventually, if you wager compulsively or impulsively, you’re likely to lose . . . . a lot.

Anybody who’s bet on horse races or other sporting events will tell you authoritatively that prohibitive favorites don’t always win. Earlier this week, for example, in an NHL hockey game, the Montreal Canadiens had a 4-0 lead with just three minutes remaining against the visiting Winnipeg Jets. That sort of lead, in a hockey game, should be insurmountable. Somehow, it wasn’t. The Jets stormed back and won in overtime, 5-4. Admittedly, that’s an exception, but these things can and do happen.

The Coming Crackup

Probabilities still have their uses, though, because they can help us to prepare for potential outcomes. Let me reach for an analogy: I walk past an intersection regularly where the traffic is ludicrously heavy, the driving often distracted and erratic, the drivers frequently impatient, and the pedestrians and bicyclists numerous. Accidents will happen there. I am sure of it. It’s a foregone conclusion. But I cannot tell you when they will happen or whether any serious injuries will result from them.

That’s how I view the AI bubble. All the factors have converged to set the stage for a crackup. We have ludicrous circular financing; an unsustainable rate of unprecedented and increasingly speculative capital spending; nouveau-riche AI providers (so-called neoclouds, often run by, ahem, checkered management teams), which posses little to no competitive differentiation; distended market valuations of too many AI-related companies; increasingly stretched private credit; and the rise of open-source, open-weight competitive threats to current AI model kingpins.

We would all be well advised to see the risks — they’re well documented — and prepare accordingly for the most likely scenarios.

You can prepare in a range of ways, depending on whether you’re an AI infrastructure vendor, a provider of AI models, a venture capitalist, an investment bank, a market analyst, or a humble investor in the stock market.

That last category is where I reside these days. I have taken resolute steps to limit my exposure to an AI market shock, though doing so has not been easy.

If you’ve bought ETFs, even those not directly exposed to the AI trade, you might find that your investments are more AI-related, even at one or two removes, than you assumed. I advise looking at the portfolio of stocks contained within an ETF.

You can still own some tech, of course, but use discretion. A rising tide lifts all boats, but when the tide goes out, not everybody in the vendor community will have the wherewithal to stay afloat. As an investor, there’s no substitute for thorough research. Look critically at the balance sheets, debt levels, cash flows, and strategic and tactical vulnerabilities of these companies. They’re not all the same; the differences will matter enormously when the crunch comes.

Value of a Dark Imagination

Finally, don’t simply hope for the best, or assume that the good times will always roll. Imagine all the things that might go wrong, and consider how you might react to them.

There will be survivors, of course, and those companies will reap lucrative rewards as AI consolidates and matures as a market. But we know from previous bubbles and their aftermaths that many ostensible contenders are ultimately revealed to be performative pretenders. Some put themselves in the past tense through their own strategic miscalculations and poor execution.

That’s why you should review and evaluate the histories of executive teams, board members, and principal backers of these companies. If you detect a surplus of rogues along mahogany row, remember the adage that leopards don’t change their spots.

As a modest investor without the power to move markets, I diversify my holdings. The irony, if it qualifies as such, is that I now have limited investment exposure to tech even though I spent decades in the industry, including considerable time as a market analyst. Allow me to posit that my aversion to disproportionate tech investment is not despite my professional experience but because of it. I’ve seen how quickly the tide can turn, how hugely successful technology companies can suddenly find themselves at mortal risk.

I constantly evaluate my portfolio, looking at how I can further balance and diversify my positions. Sure, you sacrifice the massive returns that can follow big, homogeneous bets, but bear in mind that those are still bets, and that winning and losing, no matter how clever you might believe yourself to be, is almost always beyond your immediate control. There are things you know, given the available information, and things you don’t know.

Even the most knowledgeable among us lacks complete knowledge. Blind spots can be fatal. You can mitigate them by trying to see what you don’t want to see or what you don’t expect to see.

The form charts of bubble-and-burst history strongly suggest that the AI market has morphed into a market bubble. We know that some investors suffer a lot worse than others when bubbles get punctured. Sure, you might get the timing wrong, leaving the festive AI table when it’s still piled with riches. But spare a prudent thought for something other than a best-case scenario, then prepare accordingly.

I’ll be back with a post that examines an underappreciated oddity in the AI market.