25 år efter dotcom-bubblan – lärdomar för AI-investerare

This week marks the 25th anniversary of the Nasdaq peak, but all wasn’t lost in the ensuing downturn.

Twenty-five years ago this week, the Nasdaq Composite Index hit its dot-com-era peak after soaring more than 500% in five years. Its subsequent collapse was swift and brutal.

Small investors lured by a promising new technology called the internet lost fortunes. The economy stumbled. Highflying companies like Pets.com, TheGlobe.com and Webvan collapsed.

Today, some investors are worried the same cycle might be playing out when it comes to artificial intelligence. Even if that is the case—a big if—there is an important lesson for investors from the dot-com collapse: Ultimately, the early internet hype proved correct.

It’s easy to understand the fear, and the echo of the dot-com boom. Leading AI companies are valued in the tens or hundreds of billions of dollars, some of them with little prospect of generating meaningful sales. And investors are racing to give the companies still more money at ever-higher prices to build even bigger clusters of AI chips to fill out new, cavernous data centers. 

Yet the dot.com boom and bust showed that big bets on ambitious technologies can pay off in the long run. The five most valuable listed companies globally—and six of the top seven—are tech companies from that era or ones that grew from seeds planted then.

In other words, the dot-com bubble had elements of what some investors call “good bubbles” that fuel rapid adoption of revolutionary technology. That is opposed to “bad bubbles” in which people speculate on assets that don’t make the economy more productive—things like tulip bulbs, Beanie Babies or houses in the Arizona desert.

“It’s very difficult to be a radical innovator,” says Carlota Perez, author of “Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages.” To create a world that doesn’t exist, such innovators convince suppliers, workers and financiers that they should march simultaneously toward an imagined future of clamoring consumers.

While people race to cash in, ideas are tried and infrastructure is built. Many fail who nevertheless lay important groundwork. The fiber-optic lines of 2000 were the equivalent of the electrical grids of the early 1900s, the railroad tracks of the 1800s, the canals of the late 1700s, says Perez. Busts followed those booms, yet the networks fertilized new markets. 

AI innovators are making their own gargantuan capital investments, primarily in specialized semiconductors known as graphics processing units, or GPUs, made by Nvidia. Today Nvidia is among the most valuable companies in the world, with a market capitalization of $2.7 trillion.

The question is whether those investments will lead to productivity advances that power the economy, or elements of a good bubble.

The jury will be out for some time, but there are a number of tangible advances already. Search is smarter. AI bots can write software code, cover letters and more. AI agents booking flights, filing taxes, scheduling meetings and acting as smart assistants could boost productivity in years to come.

Not that there won’t be losers. Some AI companies are already melting down. Sequoia Capital’s David Cahn has written of the massive revenue hole that AI companies need to fill to justify their data-center spending, which could lead to a speculative shakeout. Yet he’s optimistic that a “huge amount of economic value” will be created.

To distinguish bad bubbles from good bubbles, look at the assets people are betting on, says Bill Janeway, former vice chairman of Warburg Pincus who has studied speculative eras.

The worldwide banking system collapsed in 2008 when houses people couldn’t afford were financed with risky mortgages sliced into highly leveraged securities and derivatives. In contrast, Janeway points to Tesla. Some investors think its shares are overvalued, but Tesla is using its windfall to deliver a future of electric vehicles, solar power, as well as self-driving cars and robots powered by AI.

Many overhyped AI startups may flame out. Yet some will have had brilliant ideas that get picked up by others.

An early version of a smartphone was released in 1994 by a company called General Magic. Co-founder Marc Porat envisioned digital touch-screen phones, but his device arrived years early. There were no digital cellular networks, Porat recalls, and to retrieve digital content, of which there was little, it had to be plugged into a dial-up modem.

By the time Steve Jobs unveiled the first iPhone in 2007, wireless phones and the internet were widely used, flash memory was cheap, computer chips were smaller and faster, and responsive touch screens had been invented. 

And when the smartphone revolution came, it was General Magic alums Tony Fadell and Andy Rubin, who helped deliver it. Fadell helped develop the iPod and iPhone at Apple; Rubin founded Android, the startup behind the world’s largest mobile operating system. 

As hundreds of billions of dollars pour into AI, some are calling it another bubble. True or not, Perez forecasts that AI will boost productivity as the first electrical lines did when they replaced steam power. 

General Magic went bankrupt in 2002. Its devices were so far ahead of their time few consumers bought them. Another prescient idea it was working on would also have to wait: AI agents built to complete tasks for people. 

The engineer working on that project, John Giannandrea, now heads AI at Apple.