The internet and World Wide Web may be a passing fad. Its commercial significance is uncertain.

Robert Metcalfe had invented Ethernet and cofounded 3Com. He was not a naive observer of the technology industry. In a December 1995 column in InfoWorld, he predicted that the internet would experience a "catastrophic collapse" sometime in 1996, traffic would exceed capacity, the infrastructure would fail, and the bubble would burst. He offered to eat his words if wrong.

He was wrong. He blended his printed column into a smoothie and drank it at the 1997 World Wide Web Conference, in an image that has circulated ever since as a small monument to the dangers of underestimating what was coming.

Metcalfe was not alone in his skepticism. Throughout 1995 and 1996, doubts about the internet's durability competed with the enthusiasm of its promoters. Critics pointed to real problems: the web was slow, connections were unreliable, security was minimal, and most of the businesses launching on it had no clear path to profitability. The dot-com valuations that would peak in 2000 hadn't fully inflated yet, and it was still possible to frame internet enthusiasm as media-constructed hype that would soon correct.

What the skeptics missed was the network effect Metcalfe himself had theorized years earlier. The value of a network scales with the square of its number of users, a principle now called Metcalfe's Law. As each new user, each new website, each new business joined the internet, the value to everyone else increased. The incentive to join became self-reinforcing. By 1996, Yahoo, Amazon, eBay, Hotmail, and Internet Explorer were all live and growing. The network had passed a threshold from which it would not recede. The fad critique was always applying static analysis to a dynamic system, measuring today's adoption against yesterday's infrastructure, instead of tracking the trajectory between them.