AI's Bubble. Ready to burst yet?

While politicians here in NZ bicker about who should get credit for an Amazon data centre that either is (or isn't) opening, over in the States they're already wondering whether these data centres are part of an AI bubble that's starting to show clear signs of being about to burst. 

"Even Open AI boss Sam Altman is now talking about an AI bubble," notes Ted Gioia. "Of course, he knows better than anyone because he is seeing it up close—the disappointing release of ChatGPT-5 played a key role in setting off the current turmoil."

Consider this: 

AI buildout is contributing more to measured US economic growth than all of consumer spending.

I want you look long and hard at this chart, and consider the implications.

Another sign? Mark Zuckerberg just paid US$14 billion for a stake in Scale AI, the data-labelling startup that's never made a dollar.

Meanwhile in the real world, McDonald’s CFO told Bloomberg that the company is struggling because many customers are now too poor to afford breakfast. And this isn’t some isolated anecdote—it’s a data-driven report from the biggest restaurant chain in the world. ...

There’s a mismatch here between two visions of the emerging economy. So which one is real? Are we entering an AI-driven boom time like an out-of-control Monopoly game? Or will [Americans] be too broke to eat breakfast?

Several signs, maybe, that both are happening —many signs of businesses struggling, closing, unemployment and debt rising, customers at any price simply disappearing. And meanwhile, 

• half the gains in the stock market are due to betting on the shares of five companies, who are betting everything on their spending up AI data centres
• consumers however are spending so little that this "investment" spending on AI by just four CEOs (two of whose money is made mostly by selling ads) totals more in the last 6 months than all the spending by all those consumers
• the energy grid simply can't support this growth in AI data centres, and there’s no indication that consumers are willing to pay for the enormous infrastructure. 

That last is the biggest sign right there. 

Fewer than 1% of ChatGPT users are paid business accounts. That total is no larger than the number of paid Substack subscribers (but what a difference in company valuation!).

In fact, most of ChatGPT’s traffic disappears when students go on summer vacation.

That tells you how wide the chasm is between reality and the crazy claims of AI fanboys—but many of them (I bet) are also reluctant to pay for AI. ... The tech simply doesn’t live up to the hype. The more people deal with it, the less they like it. That’s why AI companies must give it away (or bundle it into an already successful product) in order to gain any reasonable usage.

So everywhere I go online, companies are touting free AI. That’s funny. It doesn’t fit the narrative of a transformative technology.

But even four billionaires can’t change reality," warns Gioia. 

Yes, they are spending like drunken sailors, but that just makes the bubble bigger. It can’t stop it from bursting. The crazy level of investment only makes the eventual fallout all the worse.

How much longer can it last? Maybe a few weeks or a few months or a few quarters. Billionaires often throw good money after bad. But the whole economy is fragile—or beyond fragile—right now. And that’s the bigger reality.

By any reasonable measure, the current trend is unsustainable. And there’s one thing I know about unsustainable trends—there’s a day of reckoning, and it’s not a happy one for the people who caused it. But, even sadder, they take down a lot of others with them when the bubble bursts.

Read the whole thing here. (NB: He's opened up the article from behind the paywall.)

PS: How is this AI capital malinvestment even possible? Because of absurdly low interest rates set by the state's economic planners at the US Federal Reserve — rates that are so "economically absurd" they are only made possible "because the monetary fraudsters on the Fed's Open Market Committee (FOMC) had their big fat thumbs on the scales in the bond pits."

And we do mean fraud: The Fed’s balance sheet rose by $1.2 trillion or 17% during the 12-month period ending on July 7, 2021, and at a time, as we will amplify below, when the Fed’s balance sheet should have actually grown by essentially zero.

That is to say, the FOMC was buying government debt and GSE paper hand-over-fist with fiat credits snatched from thin digital air, thereby starkly falsifying yields and prices in the bond pits. There is not a chance in the hot place that tax-paying, real money savers left to their own devices would accept such niggardly real yields.

David Stockman explains the Fed's fraud.

Ludwig Von Mises explains the inevitable results of malinvestment — "meaning bad investment in
lines of production that would not otherwise take place."