The current AI-driven frenzy feels more like 1997–1998 than 1999 or 2000 of the dot-com era “ This will not end well, but not yet. It’s still early in the bubble.”

Technical deteriorations in the U.S. stock market were apparent last week, but Nvidia Corp.’s NVDA, +4.00% blowout earnings report saved the day. The tech giant’s results beat Wall Street’s expectations on all metrics, and the company also guided upwards. There wasn’t anything to dislike about the report. As a consequence, the PHLX Semiconductor Index SOX, which is a bellwether for artificial-intelligence-related plays, rallied strongly.

Yet much like the dot-com bubble of the late 1990s, signs of excesses are evident. Still, I reiterate my view that the AI bubble has far more room to run before it reaches the phase of “magnificent exuberance.”

This is how a bubble builds. Consider General Electric Co. GE, +1.12%, which was one of the non-tech darlings of the late 1990s. Jack Welch, who was regarded as the superstar CEO of the day, demanded that division managers become either No. 1 or 2 in their business lines. If it failed to achieve those goals, the division was either shut down or sold.

Division managers who couldn’t make the numbers tried the old trick of financialization: GE would lend a customer money to buy its products and round-tripped the funds to inflate sales. The maneuver worked so well that GE Capital was born. GE Capital, in its efforts to achieve top-two status in its industry, lent to anything that moved. It wasn’t just aircraft engines, but emerging-market loans and subprime mortgages. GE Capital became bigger than the industrial divisions of the company and eventually blew up; GE CEO Jeff Immelt announced the divestment of GE Capital in 2014, and eventually its parts were sold off over the next two years.

“Today’s tech giants are buying equity in companies and round-tripping the funds to boost sales. ”

It’s happening again. Instead of vendor financing, today’s tech giants are buying equity in companies and round-tripping the funds to boost sales. Amazon.com Inc.’s AMZN, +0.83% investment in Anthropic and Nvidia’s investment in CoreWeave are just two of the most visible examples of financialization. As today’s market is focused on the prospect of total addressable market and sales growth, tech executives, who are mostly paid with stock-based compensation, are scrambling to boost sales growth at any cost. 

This will not end well, but not yet. It’s still early in the bubble.

There have been some concerns raised about U.S. market-index concentration, as well. Those worries are mitigated by two factors. Concentration peaked in 1931–1932, instead of at the time of the October 1929 crash. Additionally, the current episode of increasing index concentration has been gentler and less sharp than the last two instances.

Valuations are also more reasonable than in the late-1990s dot-com bubble experience. The prices of technology and communication-services stocks are still closely tracking their earnings. 

Yet there are a number of warnings of excesses to be worried about. Amazon is about to replace Walgreens Boots Alliance Inc. WBA, +1.08% in the Dow Jones Industrial Average DJIA.  SentimenTrader has documented how being removed from the Dow has been, on average, a contrarian buy signal and how new additions have tended to lag the market.

The last time this happened was in 2020, when Salesforce Inc. CRM, +2.61% replaced Exxon Mobil Corp. XOM, +1.26%, which has been great for Exxon shareholders and not so much for Salesforce investors. So this latest change in the composition of the Dow may be a sign that Amazon, and growth stocks generally, are about to lag the market.