AI Bubble

The comparison everyone reaches for is the dot-com era. The internet was real and changed everything, but in the late 1990s investors priced in that change years too early, and when the market fell in 2000 many companies disappeared. A bubble, in that sense, can sit on top of a genuinely important technology. The argument is about timing and price, not about whether the technology works.

A business version: a startup raises money at a valuation that assumes AI agents will replace whole departments within two years. Customers run pilots, most stall before production, and the next funding round comes in far lower. The product may still be useful. The expectations were the bubble.

The hype cycle, a model popularized by the research firm Gartner, gives the pattern names: a peak of inflated expectations, a trough of disillusionment, then a slow climb to a plateau of productivity. Not every technology follows it, and it describes moods rather than measuring them.

Concern isn’t only coming from skeptics. In October 2025 the Bank of England’s Financial Policy Committee said the risk of a sharp market correction had increased and that disappointing AI progress could hurt the valuations of companies whose revenue depends on AI infrastructure spending. That is a sourced statement of risk, not a prediction that a crash will happen.

Guests disagree about how deep it goes. In episode 74, at 20:07, Laurie Voss argues there is both a valuation bubble and a hype bubble, and that unless someone builds AGI, AI will behave like a normal technology. At 44:10, Voss describes the trough of disappointment in a hype cycle: valuations collapse, then a slower climb to where people expected to be, only years later. Episode 74 is a sponsored episode (G2i), and its opening read credits Inngest, Svix, Walrus Memory and G2i as the season’s sponsors. In episode 15, at 1:19, Kelly Vaughn answers “yes” to whether agentic AI is a bubble, saying the industry is “aboard the hype train” of agents for everything. The episode was sponsored by Galileo. Both are opinions.

Editorially, the useful question for a team isn’t “is this a bubble?” but “would this project still make sense if the hype ended tomorrow?” That means measuring returns from an AI pilot before scaling it. For more, see Is the AI Agent Bubble Real? and Why Enterprise AI Projects Fail to Show ROI. Past slumps are covered under AI winter.