The dot-com bubble, 2000 — when the technology was real and the prices were not
The answer
The Nasdaq Composite peaked at 5,048.62 on 10 March 2000 and fell to about 1,114 by 9 October 2002, a decline of roughly 78%. The internet turned out to be more important than almost anybody predicted. The prices were still wrong, and those 2 facts are not in conflict.
Why this costs you money
The dot-com bubble is the most dangerous history lesson in this cluster, because the story was correct.
People who bought internet shares in 1999 were right about the internet. They were right that it would change retail, media, advertising, banking and work. They were more right than the people who mocked them. And they lost 80% to 100% of their money.
Here is the mistake, stated precisely: being right about the industry tells you nothing about the return on the shares. The return depends on the price you pay relative to the cash the business eventually produces, and on whether the specific company you own is one of the survivors.
The way this costs you money today is simple. You identify a genuine technological change. You conclude, correctly, that it is enormous. You then buy whichever company is most associated with it, at whatever price, because the change is real. The first 2 steps were analysis. The third step contained no analysis at all, and the third step is where the money goes.
How it works
Four mechanisms turned a real technology into a set of impossible prices.
1. A new measurement replaced profit. Companies that had no profits were valued on "eyeballs", page views, registered users, or revenue growth alone. The argument was that profits would follow scale. For a few companies that turned out to be true. The measurement was not ridiculous. It was untested, and untested measures are exactly the ones that get adopted when the tested ones give an inconvenient answer.
2. Very small free float. An initial public offering in 1999 often sold a small fraction of the company — sometimes under 10%. A small supply of shares meeting large demand produces a high price. That price was then multiplied by the total share count to report a market capitalisation. So a company could be described as worth $10 billion on the basis of trading in a tenth of it.
3. Lock-up expiry. Insiders were usually barred from selling for around 180 days after the IPO . When that period ended, supply arrived. Many 1999 IPOs peaked before their lock-ups expired, which tells you what the price had really been measuring.
4. Analyst research that was not research. Investment banks earned fees from underwriting IPOs. The same banks published buy recommendations on the companies they had underwritten. In April 2003, 10 firms reached a settlement with US regulators totalling about $1.4 billion, including terms separating research from investment banking.
Then there was the ending, which was undramatic. There was no single crash day. The Nasdaq peaked on 10 March 2000, and fell for 2 and a half years. Many investors bought more on the way down, because each level looked cheap compared with the last one.
What it tells you, and what it does not
What it tells you. Separate 3 questions that people collapse into 1.
- Is the technology real?
- Will it produce large profits for somebody?
- Will it produce large profits for this company, at this price?
The 1999 answer to question 1 was a clear yes. The answer to question 2 was also yes. The answer to question 3, for most listed companies of the era, was no. Most of the profit from the internet went to companies that either did not exist yet, or were not the obvious candidates.
What it does not tell you. It does not tell you that high valuations are always wrong. Amazon in 1999 looked absurdly priced on every conventional measure, and an investor who bought at the 1999 peak, held through a fall of more than 90%, and never sold, ended up with an extraordinary result. That is a real outcome and it happened to almost nobody, because holding through a 90% fall requires a kind of certainty that is indistinguishable, in advance, from stubbornness.
It also does not tell you that "no profits" means "no value". A company deliberately spending its profits on growth is a different case from a company that cannot produce profits. Telling those 2 apart is the actual skill, and the test is whether the spending is optional.
The decision rule
When a story is obviously true, the story is not the analysis. Ask what has to be true about the numbers, and by when.
The conditional form: if a technology is genuinely transformative and the company you own has a defensible position and the price implies a share of the market that is smaller than the market itself, then a high multiple can be justified. If the price implies revenue larger than the entire industry it sells into, the story being correct will not save you.
A second, harder rule: assume you cannot pick the winner. In 1999 the eventual winners were not the most obvious names, and the most obvious names mostly did not survive. If your case depends on picking the 1 survivor out of 30, say so out loud, and size the position as the bet it is.
Try this now
Ten minutes. This is a comparison almost nobody makes, and it changes how you think about concentration.
- Find today's 5 largest companies in the world by market capitalisation. Any financial site lists this. Write them down with their market caps.
- Now find the 5 largest companies by market capitalisation 20 years ago. Search for "largest companies by market cap" with that year. Write those down too.
- Do the same for India: the 5 largest companies by market cap on the NSE today, and the 5 largest 20 years ago.
- Count how many names appear on both lists, in each country.
- Now check your own holdings. How many of them are in today's top 5, or are funds whose largest holdings are in today's top 5? Add up that money as a percentage of your portfolio.
What you should see. In the United States, the overlap between the 2 lists is usually small — often 1 name, sometimes none. Twenty years is long enough for the top of the market to be replaced. In India the overlap is usually larger, because the largest Indian companies are older and more concentrated in energy, banking and consumer goods. That difference is itself informative.
Then look at step 5 and notice something. If you own an index fund, you already own a large position in today's top 5, whether you chose them or not. The concentration in your portfolio is probably higher than you believe, and it is concentrated in exactly the names that a 20-year comparison says are least likely to still be there.
This is not an argument to sell them. It is an argument to know the number.
Three real cases
1. Pets.com, 2000 (United States) — the whole cycle in 9 months An online pet supplies retailer. It listed on the Nasdaq in February 2000, and announced it was closing in November 2000, about 9 months later. What it turned out to be: a business that lost money on nearly every order, because shipping heavy, low-value goods to individual homes cost more than the goods sold for. The unit economics were visible in the prospectus. The advertising was better than the business.
2. Cisco Systems, March 2000 (United States) — a great company at an impossible price Cisco briefly became the most valuable company in the world, at a valuation of roughly $500 billion. Cisco was not a fraud, not a fad, and is still a substantial business decades later. Its shares fell about 80% from the peak and took a very long time to approach it again. What it turned out to be: proof that "good company" and "good investment" are different judgements, and the price is the whole difference.
3. Indian IT, 1999 to 2001 — the real version and the imitation version Indian software services companies were genuinely growing, genuinely profitable, and genuinely exporting. Their shares rose sharply through 1999 and 2000, and then fell hard in 2000 and 2001 along with the global technology market. Alongside them, a number of small Indian companies added a technology-sounding name or announced an internet division, and their prices rose with the sector. What it turned out to be: the substantial companies survived and grew for the next 2 decades; the renamed ones mostly did not. The name change was the signal, and it was free to observe.
The question that resolves it
A novice asks: is this technology going to be big?
An expert asks: who captures the profit, and what is already in the price?
Transformative technologies frequently make their customers rich and their providers poor. Air travel transformed the world and airlines have destroyed enormous amounts of capital. The question is never whether the change is real. It is where the money settles when the change is finished.
What would make this wrong
The falsification condition. If price paid did not matter when the underlying story was correct, then buying the leading technology companies at any price in 1999 would have produced good long-run returns. Track the actual survivors from the March 2000 peak and most of them took many years to recover, several never did, and the index itself took about 15 years to regain its March 2000 level . If a study showed that entry price was irrelevant over 20 years for correctly identified technologies, this article would be wrong.
The honest limit. Some genuinely expensive-looking companies were correctly priced. Amazon is the standing example. The article is not claiming that a high multiple is proof of a bubble. It is claiming that a correct story is not evidence about the price.
The hindsight trap. "Everybody knew Pets.com was ridiculous" is false, and believing it will make you worse at this.
Consider what a careful investor in late 1999 actually saw. Internet usage was doubling roughly every year. Companies that listed in 1995 and 1996 had produced returns of many multiples, so scepticism had already been punished for 4 years. Serious institutions were buying. Serious academics published arguments that the equity risk premium had permanently fallen. And productivity statistics were genuinely improving, which supported the "new economy" argument at the level of national data, not just anecdote.
Even the sceptics were mostly wrong about the reason. Many argued the internet would not matter commercially. That was a bad prediction.
So what could a careful person genuinely have seen? Two things, both public.
First, cash burn against cash on hand. Every listed company filed accounts. For a large number of 1999 internet companies, the cash flow statement showed cash going out faster than it came in, with a balance that would run out in 12 to 24 months and a plan that depended on raising more. That is not an opinion. It is a division. Companies that needed the capital markets to stay open were the ones that died when the markets closed.
Second, the identity of the sellers. Insiders and venture investors were selling into the IPOs and after the lock-ups, and that was disclosed. A company whose founders are selling at the offer price is telling you something about the offer price.
What a careful person could not have seen was which companies would survive. In 1999, choosing between the leading online bookseller and the leading online pet store from published information was genuinely hard. Anybody who claims the winners were obvious should be asked for their 1999 list.
In India
India's dot-com episode was real, and it had a second act that most accounts skip.
The first act, 1999 to 2001. Indian software exports were growing rapidly and the sector had genuine advantages: cost, English, and a supply of engineers. Share prices of the exporters rose sharply. The bubble part was not the sector; it was the price and, especially, the imitators. Companies with no software business renamed themselves or announced internet subsidiaries. The "dot-com" suffix worked in Mumbai exactly as it worked in New York.
The correction in 2000 and 2001 coincided with the unwinding of the Ketan Parekh positions, several of which were in technology, media and telecommunications stocks. India had a global technology decline and a domestic manipulation collapse at the same time, which is one reason the Indian fall was severe.
The second act, and this is the honest part. The Indian IT sector recovered and grew for 2 decades. The large listed software exporters became some of India's most valuable businesses. Somebody who bought the substantial companies at the 2000 peak waited a long time and eventually did well. Somebody who bought the renamed ones lost everything. Both were "buying Indian IT in 2000".
The modern Indian version. Since 2021 a set of Indian technology companies have listed while still loss-making. SEBI now requires the offer document to disclose key performance indicators and to compare the issue price with the prices at which the company issued shares to investors in the preceding 18 months . That requirement is unusual internationally and directly useful: it shows what professional investors paid shortly before you were asked to pay more.
In the United States
The American record from this period is unusually well documented, and 3 mechanisms are worth carrying forward.
Lock-up expiry. Insiders typically cannot sell for about 180 days after an IPO. The expiry date is in the prospectus. A surge of supply on a known future date is one of the few genuinely predictable events in markets.
The research settlement. In April 2003, 10 major firms settled with the SEC, the New York Attorney General and others, agreeing to pay about $1.4 billion and to separate research from investment banking. Two prominent analysts were barred from the securities industry. The rules that came out of this are why US research reports now carry explicit disclosure of banking relationships.
Pro-forma earnings. Many companies reported earnings that excluded costs they described as unusual, most commonly the cost of paying employees in shares. The SEC's Regulation G, adopted in 2003, requires reconciliation of such non-standard figures to standard accounting. When a company today reports "adjusted" profit far above its reported profit, the reconciliation table is where you find out what was removed.
Where they differ, and what that tells you
The difference is what happened to the sector afterwards, and it is instructive.
In the United States, the dot-com era companies that survived went on to become the largest companies in the world. The bubble was in the price, and the underlying industry became more important than even the optimists claimed.
In India, the software services sector also survived and grew — but it grew as a services export business with steady margins, not as a set of global platform monopolies. The Indian companies that emerged from 2000 are large, profitable and still growing. They did not become the most valuable companies on earth.
What that tells you is the difference between a good industry and a winner-takes-most industry. American internet platforms had network effects: each user made the product better for the next user, so 1 company took most of the market. Indian IT services has no such effect. A larger vendor is not automatically better for the client.
The consequence is how you size a bet. In a winner-takes-most industry the extreme outcomes are real, so a small position in several candidates is rational. In a services industry with normal competition, no company takes 60% of the market, so paying a winner-takes-most price is an arithmetic mistake. Ask which kind of industry you are in before deciding what multiple is defensible.
Carry this
- Being right about the industry says nothing about the return on the shares.
- 3 questions, not 1: is it real, will it pay, will it pay this company at this price.
- Cash out ÷ cash on hand tells you who survives a closed capital market.