“Dangerous bubble sure to pop”. Wall Street has AI crash in the wings

“Keeping perspective” Michael Pascoe‘s mantra. He writes that the AI ​​craze not only distorts perspective but also risks blowing up reality.

In more years of market monitoring and reporting than I can count, I have never seen a time when there was such widespread belief that we had a dangerous bubble that was sure to burst, yet money continued to flow to inflate it.

The road to the “Crash of ’87”, now a minor glitch in the rearview mirror, was relatively quiet. Even the “dot bomb” bubble at the turn of the century, when all a company had to do to get a rise in stock prices was to add an “e” or “dotcom” to its name, was not as widely perceived as overcooked as the current outlook.

“This time is different”

This time he is pursuing himself with much more money, promises of what is essentially revolution, and a cyclical investment boom built inevitably on “this time is different.” This time, trillions of dollars are being spent on a dizzying mix of momentum building, FOMO (fear of missing out) and big bets that the first mover will win it all.

There are rumors in the market about sky-high valuations for inexperienced startups that only intend to build things like data centers or artificial intelligence, but these are nothing more than insignificant sideshows that will evaporate when the showdown occurs, just as shells and scams do when the dot bomb goes off.

As a general rule, things are neither as bad as you fear nor as good as you hope.

This time hope is low.

The serious game is the impact of both reckoning and the promises of artificial intelligence on the real economy, employment and wealth. This is where the potential is for a GFC scale event, not a junk bomb of 87 or a pinpoint bomb.

And this time, governments and central banks, still bearing the cost of the entire COVID effort, will be less successful in easing the pain. We are already seeing the result of a global easing cycle where global government deficit spending fuels asset inflation and the ensuing wealth effect. Looking back, after overcompensation during COVID, our ammunition is not there to fight another shock.

Two very different articles this month highlighted the impossibility of the nirvana promised by AI investment proponents. Former colleague Alan Kohler writing for ABC It was bleakest to see a future in which a GFC-sized crackdown was the least bad option.

not there

The other is a note from independent economist Gerard Minack to clients showing that AI spending may not provide investors with returns that justify it.

I will come back to this topic. First, Kohler combines AI and crypto bubbles for scary numbers:

Between $3 trillion and $6 trillion has been invested in creating AI infrastructure and software, accounting for nearly all of the US’s economic growth last year.

The 10 largest American AI companies have driven most of the US stock market’s gains over the past two years and are now worth $35 trillion, nearly half the total market.

Meanwhile, there are 20,000 cryptocurrencies worth $5.8 trillion, with Bitcoin representing more than half of them.

Total cash in AI and crypto betting is more than a quarter of global GDP; It is probably the largest tech investment boom/bubble in history.

In my opinion it’s probably not about that. There is no need for an all-out collapse to send a shock to the financial and real economies.

Jobs in exchange for doomsday

But Kohler’s Doomsday conclusion is that the situation is worse if investment is justified by the profits generated; because it would create massive long-term unemployment for the majority, while the very few at the top continue on their current path of becoming unimaginably rich.

Gerard Minack’s note has a smaller focus and therefore a more concrete consequence: A lot of capital is being burned.

Minack limits its count to Alphabet, Oracle, Microsoft, Amazon, Meta, Broadcom, Nvidia and Palantir, which are on the “AI8” list. During this bubble phase, investors continue to reward companies that increase planned capital expenditures related to AI. However:

“The greater the capex, the greater the revenue that must be generated to generate adequate returns on that investment. In my opinion, the revenue hurdle is already incredibly high. Whatever technical wonders AI will create, the investment returns will be disappointing. This will inevitably lead to significant market losses.”

Greatly simplifying Minack’s analysis, AI8 conservatively projects more than $1 trillion in investment stock by the end of next year. This allows for a 20 percent depreciation on capex of more than $2 trillion in 2027. (And a reminder that this ignores the capex of unlisted companies like Open AI.)

$1 trillion for 10% return

So how much revenue will AI businesses need to generate to generate a reasonable return on this invested capital stock? I’ll skip the details of Minack’s calculation, but AI would need about $925 billion in annual revenue to generate a modest 10 percent return on invested capital. And that return compares with hyperscalers’ current ROIC of 25 percent.

In contrast, Minack cites Praetorian Capital’s Harry Kupperman’s observation that the wildly successful Microsoft Office 365 subscription services generated $94 billion in revenue last year.

“In other words, to achieve an average return on investment, the AI ​​industry will need to support 7-10 companies with widely distributed and widely subscribed businesses like Office 365.”

And then it gets even harder. If hyperscalers’ current 50 percent gross margin approaches the S&P500 average, they would need $1.2 to $1.6 trillion in additional revenue. As Minack concluded, “good luck with that.”

Another complication is that AI will cannibalize much of hyperscalers’ existing business.

I will speculate further on how competition between so many players will affect margins. There is also the well-reported phenomenon that most AI spending is cyclical; the big players hang out and buy most of each other’s laundry.

show me the money

It’s all good fun until investors hit the impending “show me the money” phase and the music stops.

That’s when the big boys get hit and the fringe players, the bubble startups, lose their shirts.

As for the simultaneous crypto bubble, RBA Governor Bullock last week noted the challenges to the security of the financial system if quantum computing works effectively.

“If you believe what they say about quantum computing, what took 200 years to decrypt will take a few minutes to break. So that’s a big threat,” he said.

Decrypted crypto is not crypto at all.

And the Little King

There’s also the small matter of the world’s largest economy being run by a hot-headed toddler king and a gang of self-enriching accomplices.

All of this comes together to reveal the biggest mystery of all: How markets so willingly charge higher fees to hit more records, regardless of risk, and reap the biggest gains at the riskiest end.

On the one hand, there is a view that the reckoning is always a little further ahead. On the other hand, the second law of the “old vineyard dog” Anthony Peters comes to mind:

“No one who takes a long position in a falling market will be fired, but anyone who goes short in a rising market will be in trouble.”

Good luck with that too.

With Donald J. Albo supporting mass murder