The people who called the dot-com crash are now calling the AI crash — and they're giving you a date.
The Summary
- Panmure Liberum warns AI bubble could burst within two years, targeting 2028 as the potential crash point
- Ray Dalio labels AI a "classic bubble" nearing burst, citing rising interest rates and liquidity pressure as catalysts
- Two heavyweight financial voices converging on the same thesis within 48 hours is the signal
- The trigger isn't technology failure — it's money getting expensive
The Signal
When Joachim Klement from Panmure Liberum puts a two-year clock on the AI bubble, he's not predicting AI will fail. He's predicting AI companies will run out of runway before they prove they can actually make money at scale. That's a different bet entirely. And when Ray Dalio calls it a "classic bubble" in the same week, you're watching pattern recognition from people who've seen this movie before.
The setup is textbook: massive capital inflows, sky-high valuations disconnected from revenue, and a belief that "this time is different" because the technology is genuinely revolutionary. All of that can be true while also being a bubble. The 1999 internet really did change everything. That didn't stop the Nasdaq from losing 78% of its value.
"The trigger isn't whether AI works. It's whether companies can turn inference costs into profit margins before the money runs out."
What makes this different from 2000 is the interest rate environment. Dalio points to rising rates forcing wealth into cash, which means the free money that funded AI's first wave is evaporating. Companies that raised at 40x revenue now have to show a path to profitability, not just a path to AGI. The math gets ugly fast when:
- Training runs cost $100M+ and need to happen quarterly
- Inference costs per query haven't dropped fast enough to support free or cheap tiers
- Enterprise customers are still in pilot phase, not production deployment phase
- The delta between "AI is amazing" and "AI increased our profit margin" remains wide
The Panmure thesis centers on a severe market crash specifically triggered by AI names. That's not generic bearishness. That's a view that AI stocks have become so concentrated in major indices that their collapse would drag everything down with them. Look at how much of the S&P 500's gains over the past two years came from seven stocks. Now imagine four of them cut in half.
The Implication
If you're building in this space, the clock matters more than the technology roadmap. You have maybe 18-24 months to get from "we're AI-powered" to "we're profitable and AI is why." Companies that can't show margin improvement or real cost savings for customers before 2028 are playing musical chairs.
For workers, this means the AI job boom might plateau faster than expected. If funding dries up, the model trainer and prompt engineer roles proliferate less. The winners will be people who can use AI to make existing businesses more efficient, not people building the 47th AI wrapper startup. Watch where the talent migrates in the next six months. That's your signal for where the real value is accumulating.