Is AI a Bubble? Lessons from Past Market Cycles

If you're raising capital for an AI company right now, you've probably noticed that some investors are eager and others are cautious. If you're deploying capital into AI, you're likely wrestling with the same question everyone else is: are we in a bubble, or is this real?

A Q4 2025 analysis from GMO—a firm known for calling bubbles before they pop—offers a useful framework. Their argument: you don't actually need to answer the bubble question to make smart decisions. What matters is understanding what kind of market we're in.

This Looks Like 2000, Not 2007

GMO compares today's AI market to three previous cycles: the Internet Bubble of 2000, the Everything Bubble of 2007-8, and the Duration Bubble of 2021. Their conclusion is that the current environment most closely resembles 2000—a bubble concentrated in a narrow segment of the market, while the rest remained reasonably valued.

That's actually useful information. In 2000, returns were concentrated in a small number of large-cap tech companies. Valuations for AI-related stocks today carry significant premiums over the broader market. Expectations for future earnings growth are baked into current prices. If you're building or investing outside the Magnificent Seven, you're operating in a different valuation environment than the headlines suggest.

What This Means for Founders

If the bubble is concentrated—not broad—then being outside the concentration isn't a disadvantage. Founders building AI infrastructure, vertical applications, or enterprise tools may find that their valuations aren't as inflated as the market leaders, which means there's room to grow into reasonable multiples rather than needing to justify stretched ones.

It also means that if there's a correction, it may not hit everyone equally. The founders who survive corrections aren't the ones with the highest valuations—they're the ones with real revenue, real margins, and real customer retention.

"When a bubble is concentrated to one part of the market, you can express relative views without having to have a strong view on the bubble overall."

— GMO, "Bubble, Not Bubble," Q4 2025

What This Means for Investors

GMO's framework suggests that investors don't need to make a binary bet. If you believe AI is concentrated like 2000, non-AI segments may offer better risk-adjusted returns. International markets may provide exposure to growth without the same valuation premiums. And quality—sustainable competitive advantages—may outperform momentum when the music stops.

For early-stage investors, this is a reminder that entry price matters. The best AI startups at reasonable valuations may outperform mediocre ones at stretched valuations, even if the sector as a whole continues to rise.

The Takeaway

You don't need to know if AI is a bubble to invest wisely. What you need to know is whether valuations are concentrated or broad—and right now, they're concentrated. That means there's opportunity outside the obvious names, and there's risk in chasing the ones everyone already owns.