The relentless, frenetic excitement surrounding Artificial Intelligence feels familiar. For anyone who remembers the turn of the millennium, it’s a clear echo of the dot-com bubble, a time of speculation about a new technology completely detached from business fundamentals.

Back then, optimism for an internet-driven economy sent stock prices for companies like AOL to astronomical levels. The problem was that few of these companies were actually profitable. When the bubble burst, the losses were staggering. AOL Time Warner posted a staggering $98.7 billion annual loss in 2002, the largest in U.S. corporate history at the time, largely from writing down the value of its internet division.

Today, we see a parallel. The market is flooded with AI hype, yet very few companies—outside of key infrastructure players like chipmaker Nvidia—are turning a profit. This is unsustainable. These unprofitable AI app-developing “miners” are the biggest customers for Nvidia’s “shovels.” If they run out of cash and collapse, the demand will inevitably falter.

This doesn’t mean the AI revolution isn’t real. It is. But, as with the internet, the first wave is rarely the one that lasts. The true, world-changing success of the internet didn’t come from the bubble-era giants; it came from the companies built after the crash, like Google and Facebook.

Bill Gates articulated this exact point in a recent interview, directly comparing the current “frenzy” to the internet bubble. He clarified that while AI is “profound”—the “biggest technical thing ever” in his lifetime—it will follow a similar pattern. Just as with the internet, many companies will fail, and a “ton of these investments… will be dead ends.”

History suggests we are in the speculative phase. The real, lasting AI revolution will likely be built on the ashes of this first, over-hyped wave.