The right person to read to learn about this is Andrew Odlyzko at UMN: http://www.dtc.umn.edu/~odlyzko http://www.dtc.umn.edu/~odlyzko . He draws some amazing c

The right person to read to learn about this is Andrew Odlyzko at UMN: http://www.dtc.umn.edu/~odlyzko http://www.dtc.umn.edu/~odlyzko . He draws some amazing comparisons between the dot-com bubble and the railway mania of the 1850s. The dot com boom was based heavily upon a myth: that the internet was doubling every 100 days. Various executives at WorldCom / UUnet kept repeating this for years, even though it was only true for a few months in 1996. Since WorldCom / UUnet was the biggest ISP at the time, the business world (and especially investors) believed this. At the same time Lucent was doing some creative accounting where they would sell equipment super-cheap on long-term repayment plans with a buy-back offer. They showed inflated sales and no liabilities: this too made it look like their networking business was growing at an astounding rate, which supported the WorldCom claim. Now, if it had been true (that the internet was doubling every 100 days continuously for 4 years), then it would have made perfect sense to invest in bizarre dot com stocks: there was so much potential upside that there was no way you could possibly be paying too much for any stock. A company that has an exponential growth with no end in sight is a very good buy, regardless of how silly it might look now. So that's precisely what investors did. Unfortunately, the myth wasn't true, and it was becoming obvious by 2000. At this point there was no reason to continue investing in business solely because they had a strong internet presence. If internet growth was "only" 100% per 18 months (which was probably closer to the truth) then even established bricks-and-mortar companies would be able to move fast enough to capture significant internet sales. This is indeed what has happened since. Without fresh cash injections, the crazier dot coms died, taking with them large numbers of companies that supported them with hardware, development and (especially) network infrastructure.