Why Tech Bubbles Keep Happening

The Pets.com sock puppet had button eyes, a microphone and a watch on its collar, and for about eighteen months it was one of the most recognisable characters in America. It turned up on morning television. It ran in Super Bowl XXXIV in January 2000, one of twelve dot-com adverts in a broadcast of sixty-one. Pets.com floated on the NASDAQ the following month at $11 a share.

On 9 November 2000 it stopped trading. Nine months.

The numbers underneath were not subtle. In its first fiscal year the company took $619,000 in revenue and spent $11.8 million on advertising. It was selling goods at roughly a third of what it paid for them, and later at about 27 percent below cost, which means every order made the hole bigger. Amazon was one of its backers.

So the sock puppet became the headstone for the whole era, the thing people point at to prove everyone had lost their minds.

Here is the part that spoils the joke. Shipping heavy bags of dog food to people's houses is now a perfectly good business. It is called Chewy, it was founded in 2011, PetSmart bought it for $3.35 billion in 2017, and in its 2024 financial year it reported net sales of $11.86 billion.

The idea was right. The date was wrong. Almost everything interesting about bubbles is in that gap.

The Victorians did this with railways

In the 1840s, Britain went mad for railway shares. The peak was 1846, when Parliament passed 263 Acts authorising new railway companies, covering 9,500 miles of proposed line. Ordinary middle-class people who had never owned a share bought into schemes for routes that had not been surveyed.

Late in 1845 the Bank of England raised interest rates, money moved elsewhere, and the whole thing stopped almost overnight. Share prices collapsed. Companies that had taken deposits could not raise the rest. Plenty of people were ruined.

Then look at what was standing afterwards. Of the schemes authorised between 1844 and 1846, 6,220 miles of railway actually got built. Britain's network today is around 11,000 miles. More than half the railway the country still runs on was financed by a mania, and the failed lines were mostly bought up cheap by the survivors and made to work.

Turner painted Rain, Steam and Speed in 1844, right as it was starting. The painting is in the National Gallery. Most of the companies are not in anything.

Why they form at all

A bubble needs a technology that is genuinely new, genuinely important, and genuinely impossible to value.

That last part does the work. If you are pricing a water utility you can look at what it earns and what it will earn. If you are pricing a company that has invented a category eighteen months ago, there are no earnings to discount and no comparable to check against. Nobody knows which three of the two hundred entrants will still exist. In that vacuum the price stops being an estimate of value and becomes an estimate of what other people will pay.

Keynes described this in 1936 with a newspaper competition where readers pick the six prettiest faces from a hundred photographs, and the prize goes to whoever picks closest to the average choice. The sensible entrant does not pick the faces they find prettiest. They pick the ones they think everyone else will pick, and the really sophisticated ones go a level deeper and anticipate what average opinion expects average opinion to be. Prices in a new sector work the same way, and cheap credit turns the volume up.

The economist Carlota Perez put a shape on the whole cycle. Financial capital rushes into a new technology and funds a build-out far beyond anything current demand justifies. That frenzy ends in a crash. Then, after the wreckage is cleared, the technology gets deployed properly by people who want to use it rather than trade it. On her account the bubble is not a malfunction of the process. It is the part of the process that pays for the infrastructure.

Why people buy in when they know

This is the question that actually bothers people, and there are three good answers.

The first is that knowing is not the same as selling. Markus Brunnermeier and Stefan Nagel went through hedge fund holdings from 1998 to 2000 and found that hedge funds, the supposedly clever money, were heavily overweight in exactly the overpriced technology stocks. They did not lean against the bubble. They rode it, and then trimmed their positions in individual stocks shortly before those stocks fell. They were not fooled. They were surfing, and they got off in time.

The second is that fund managers are employees. A manager who sits out a rising market underperforms every quarter, has to explain it to clients every quarter, and gets redemptions long before being proved right. Chuck Prince, then running Citigroup, put it plainly to the Financial Times in July 2007: "When the music stops, in terms of liquidity, things will be complicated. But as long as the music is playing, you've got to get up and dance. We're still dancing."

The third is timing, and it is brutal. Julian Robertson ran Tiger Management, one of the great hedge funds, and he refused to buy technology stocks because he thought the prices were nonsense. He was right. He also bled investors for two years while being right, and he announced he was closing the fund in March 2000, the same month the NASDAQ peaked. He had called it correctly and it finished him anyway, because on a monthly statement being early and being wrong look exactly alike.

So is it even a bubble?

The word implies there is nothing inside. Usually there is something inside.

Take the telecoms build-out of the late 1990s. Companies put more than $500 billion, most of it borrowed, into laying fibre optic cable, installing switches and building wireless networks, on demand forecasts that turned out to be fantasy. When the money stopped, the bankruptcies were enormous and bondholders recovered just over 20 cents in the dollar. As an investment it was a catastrophe.

The cable stayed in the ground. Much of it sat unused for years, and then the next decade of cheap bandwidth, video and cloud services ran across capacity that nobody would have built if they had known the real near-term demand.

This is why some economists dislike the term. Eugene Fama's objection is that if nobody can identify a bubble before it pops, calling something a bubble afterwards is storytelling rather than analysis. And the honest version of the dot-com story is not that the internet was overhyped. The internet was, if anything, underhyped. What was wrong was the belief that any given company would be the one to capture it, on that timetable.

A more useful reading is that a bubble is a financing pattern, not a verdict on the technology. It is the mechanism by which a society funds infrastructure that no rational investor could justify in advance, and the bill lands on whoever is holding at the end.

Who ends up holding it

That last clause is the part that deserves anger rather than admiration.

The NASDAQ Composite peaked at 5,048.62 on 10 March 2000. By 9 October 2002 the NASDAQ-100 was down 78 percent. Something like $5 trillion of market value was gone. Cisco, a real company selling real equipment that people genuinely needed, lost 80 percent of its share price and has spent the decades since as a perfectly sound business that never got back to that number.

And the distribution of the damage is not random. The professionals reduced their positions before the falls, because that is what the data shows they did. The people still holding at the bottom were disproportionately the ones who arrived last, having read the same headlines that the early money had already acted on.

The infrastructure is a genuine public gain. The losses are a private and quite specific transfer, and the two facts do not cancel each other out just because they happened in the same market.

What the sock puppet was actually wrong about

Pets.com did not fail because people did not want pet food delivered. It failed because in 2000 not enough households had broadband, the parcel logistics to move heavy low-margin goods profitably did not exist yet, and the company was burning cash on television to acquire customers it lost money on.

Chewy solved none of those problems by being cleverer. It arrived when they had already been solved by somebody else, mostly by the warehouses and delivery networks built in the intervening decade.

The sock puppet was not stupid. It was eleven years early, which in a market is a distinction without a difference.

Sources

• Wikipedia, "Pets.com" (the sock puppet, the January 2000 Super Bowl advert, the February 2000 IPO at $11, the $619,000 of first-year revenue against $11.8 million of advertising, selling below cost, and the 9 November 2000 shutdown nine months after the float).
• Wikipedia, "Dot-com bubble" (twelve of 61 Super Bowl XXXIV adverts bought by dot-coms, the 5,048.62 NASDAQ peak on 10 March 2000, the 78 percent fall by 9 October 2002, roughly $5 trillion of lost market capitalisation, Cisco's 80 percent decline, the $500 billion telecoms build-out, and bondholders recovering just over 20 percent).
• Wikipedia, "Railway Mania" (the 263 Acts of 1846 and 9,500 miles proposed, the Bank of England rate rise that ended it, and the 6,220 miles actually built from the 1844 to 1846 authorisations).
• Wikipedia, "Carlota Perez" (the installation, frenzy, turning point and deployment sequence, and the argument that the crash is part of the mechanism rather than a failure of it).
• Wikipedia, "Keynesian beauty contest" (the 1936 analogy and the regress of anticipating what average opinion expects average opinion to be).
• Brunnermeier and Nagel, "Hedge Funds and the Technology Bubble", The Journal of Finance, 2004 (hedge funds heavily tilted towards overpriced technology stocks from 1998 to 2000, capturing the rise and cutting positions before the falls).
• Wikipedia, "Tiger Management" (Julian Robertson's refusal to hold technology stocks and the closure of the fund in March 2000).
• Wikipedia, "Chewy" (founded 2011, bought by PetSmart for $3.35 billion in 2017, listed in 2019, and net sales of $11.86 billion in the 2024 financial year).