The four ‘O’s that shape a bubble

Open Editor’s Digest for free

Rula Khalaf, editor of the Financial Times, picks her favorite stories in this weekly newsletter.

The writer is president of the Rockefeller International Foundation. His latest book is What’s Wrong with Capitalism?

Amid the chatter about the AI ​​craze, people started joking about a “bubble in the speech bubble.” Google searches for AI with the “B” word are up and the mood in the markets is buoyant, but behind these weak indicators there is no standard measure of a bubble. My testing focuses on four systems: overvaluation, overownership, overinvestment, and overleverage. Here’s how AI can be achieved now:

Overvaluation In the great gold bubbles dating back to the 1970s and the dot-com boom of the late 1990s, inflation-adjusted prices rose tenfold over 10 to 15 years. US technology stocks recently reached this threshold. Moreover, a study of bubbles over the past century shows that the probability of collapse rises to more than 50 percent when the industry at the heart of the mania outperforms the market by more than 100 percent over two years. AI-related stocks are close to that tipping point as well.

These significant price increases have pushed long-term valuations for US stocks to near the highest levels in history. Some say it doesn’t matter, because AI will boost growth more than previous technological revolutions, and valuations were more extreme in 1999-2000. But if we are any guide to history, valuations and prices are flashing a warning of a deep red bubble.

Excessive ownership This measures the amount of money flowing into the hot new thing. Americans are aggressively chasing stocks, especially in technology. Households hold a record 52 percent of their wealth in stocks, higher than the peak in 2000 and well above levels in the European Union (30 percent), Japan (20 percent), and the United Kingdom (15 percent).

A closely related signal is overtrading. Over the past five years, the number of shares traded daily in the United States has risen by 60 percent to about 18 billion shares. The retail share of short-term stock options has grown from a third to more than half. Young people are succumbing to “financial nihilism” – and indulging in speculation because they have given up on buying a house.

If the stock market falls on a given day, retail investors buy impulsively the next day. Their favorites are clear: On the Robinhood platform, the five most-owned stocks are all in the Magnificent Seven. With $7.5 trillion in money market mutual funds, small American investors may have purchasing power left over.

Because financial conditions are still very loose, liquidity continues to lift stocks. This almost forces institutional investors to keep buying, including many skeptics of the AI ​​euphoria. The result is a strange new animal: the fully invested bear.

AI enthusiasts say the constant talk of bubbles proves that this is not a bubble, because the peak comes when anxiety disappears and optimism becomes universal. Maybe, but anxiety was actually growing before the dot-com crash. One year ago, the Federal Reserve Bank of San Francisco raised the specter of 1929. Many columnists and economists echoed these concerns, as did many institutional investors. Just like today.

Overinvestment Technology investment recently exceeded 6 percent of US GDP, surpassing the record set in 2000. Companies are pouring capital into artificial intelligence data centers and the power plants to power them, led by hyperscalers. Counting the Seven Wonders alone, spending on AI has doubled since 2023 to reach $380 billion this year, and is on track to exceed $660 billion by 2030. The potential returns are far from clear. For every survey that says demand for AI is skyrocketing, another shows the opposite: Less than 15 percent of US companies say they are using AI, amid multiple signs that the rate of adoption is slowing.

Technology optimists say investment in AI will pay for itself by lowering labor costs, replacing up to 40 percent of the tasks humans now do “in the not too distant future,” and raising unemployment rates to 20 percent. Will humans sit back while this unfolds? Disruption on this scale could lead to political backlash, limiting the degree to which investment in AI can be achieved.

Excessive influence So far, US companies and households do not appear to be overleveraged. But Seven Wonders is no longer the ATM machines it was a year ago. Amazon, Meta, and Microsoft are now net debtors, up from one debtor in 2023. Their profits continue to rise, but with so much flowing into AI, only Google and Nvidia are still generating piles of cash.

And this time, debt is piling up on the government’s ledger, thanks to record deficits — a big risk. If bond investors begin to question America’s fragile finances, they could push up long-term interest rates, which could reverberate throughout the economy.

At the same time, leverage in financial markets has surpassed traditional margin loans to purchase individual stocks. Now there are funds that borrow money to amplify their bets. These “leveraged ETFs” are easily accessible to retail investors and have seen their assets grow sevenfold over the past decade to about $140 billion.

To varying degrees, all four systems indicate that AI is a bubble and has reached an advanced stage. However, history also shows that there is no precise point at which a bubble bursts under its own weight.

The only consistent cause of the crash, which dates back to the railroad bubbles of the 19th century, was rising interest rates and tightening financial conditions. So, while we are clearly in a bubble, it is possible that it will continue to grow until the money inflating it begins to dry up.