“I spot bubbles the same way I spot bad deals: when the price is set by who else is in rather than what the company actually builds. We have made fast decisions, and most of the time they were the wrong ones, so I now ignore the Twitter charts and stay slow.”

“I don't think we're in a bubble now; my definition requires broad-based asset inflation, and our internal tracker of ~690 public tech companies sits at only the 60th percentile of the last two decades. In AI specifically, the bigger risk is easy capital dulling the discipline that produces real innovation, something I saw at PayPal when seven months of runway forced us to focus.”

“I closed a $100M fund into this exact froth, so I’m living the question. The bubble isn’t when valuations are high; it’s when revenue quality collapses beneath them. I watch for three things: POCs dressed up as contracted ARR, negative-margin growth celebrated as traction, and buyers who are curious rather than committed.”

“I look for the round-trip: when a startup's biggest customer is the same hyperscaler that invested in it, the revenue is often just the investment coming back home. Real AI value shows up where the customer actually pays for the output, not the compute.”

“I don't look at stock prices or private valuations to spot a bubble. During the dot-com bust, internet traffic kept climbing straight through the crash, and I'd bet the same on AI API calls today. Most AI investments will lose money, maybe 80% or more, but the winners will be so large that more money gets made than lost. The bubble, if there is one, is in fear-and-greed valuations, not in real demand.”

“I watch for the same pattern I saw in the last cycle: a startup with $180K MRR, fourteen logos, and a $7.5M seed ask where the monthly check is really paying for a team to keep the tool alive. That is not a software business, it is a services business wearing a dashboard, and the AI label is just the premium people are charging for the costume.”

“I don't ask whether AI is a bubble; I ask which bubble, and what stage we're in. The hardware buildout looks fragile to me, $600 billion in implied revenue chasing chips and data centres, while the companies I own are the ones with real free cash flow you'd want to buy more of if they fall.”

“I’ve lived through enough boom-bust cycles to know the bubble isn’t the technology, it’s the capital chasing half-formed ideas. Right now I think AI assets are in a bubble, but the AI market itself is not, and the real money will get made after this first wave crests, the way most internet value was created after the dot-com bust.”

“A bubble is when you can’t grow into the valuation: the company is still the discounted present value of future cash flows, and right now we’re paying two to four years ahead of where most of these businesses are. I watch for vibe revenue and experimental pilots that mask churn, because when growth slows, the retention problem becomes visible fast.”