Revenue growth, presumably. If you linearly extrapolate cash/burnrate, but revenue grows exponentially and eventually makes burn rate positive, then the prediction will be off.
It occurs to me that a lot of this bubble-like behavior may come from an initial market leader obscuring what they're actually doing and then lots of followers jumping on. Amazon was profitable on a unit-sold basis from Day 1, even before taking investment, but they invested all the profits back into capital improvements for the business, which made it look like they were hemorrhaging large amounts of money. Investors and other startups saw this and their success and thought "The way to succeed in the dot-com era is to burn lots of money!", completely overlooking that Amazon had validated their business model beforehand and showed it to be profitable, and then only expanded afterwards.
I believe that the trend toward "Just get lots of users, and the money will follow!" in the mid-2000s had a similar origin. People saw the Google founders say "We didn't know how we were going to make money when we started", it worked out for them, and so they all started businesses where they had no idea how to make money. I'm about 98% sure that the Google founders had maybe a dozen ideas about how to make money at the time they incorporated, they just needed to test them (so their statement was technically true), and it was highly likely that at least one would work out.