The Unit-Economics Bubble Test
A real bubble is when the businesses don't work, not when prices are high; and if everyone calls it a bubble, it isn't one
- Difficulty
- Moderate
- Time to result
- ~ongoing to results
- Steps
- 3
- Confidence
- 85%
To judge whether a market is a genuine bubble, don't look at how high prices are; look at whether the underlying businesses' unit economics actually work. The dot-com crash was a bubble because the math didn't work and there was no revenue. A market where companies are growing real revenue fast and being priced to that growth is not a bubble, even at high prices. A second signal: bubbles require capitulation, so as long as many people are loudly calling something a bubble, it usually isn't one yet.
Origin
Horowitz's framework contrasting the late-1990s dot-com bubble he lived through as Netscape/Loudcloud CEO with the 2025 AI market, responding to Sam Altman calling AI a bubble; references his ~2011-12 Economist debate with Steve Blank.
Core principles
- 01A bubble is defined by broken unit economics, not by high prices
- 02Real revenue growth priced appropriately to that growth is not a bubble
- 03Bubbles need capitulation, so universal 'it's a bubble' talk means it isn't one
- 04In any hot run, both the great company and the crap company get funded; that alone isn't a bubble
How to run it
- 1
Check the unit economics, not the price level
Ask whether the businesses actually work: is real revenue coming in, and do the per-unit economics hold? High prices with working economics is not a bubble.
Pro tip Contrast with dot-com: Evite needed 300 engineers and there were only ~55M internet users, half on dial-up, so the math couldn't work regardless of investor enthusiasm.
Watch out Don't confuse 'the internet is real / AI is real' with 'therefore not a bubble'; the dot-com era was real technology and still a financial bubble because the businesses didn't work.
- 2
Test whether price is justified by growth
Compare valuations to actual revenue growth. Companies going 0 to $800M in a year, priced to that trajectory, are being valued appropriately rather than irrationally.
Pro tip The strongest bear case isn't 'prices are too high' but 'the growth won't continue,' so evaluate the durability of growth directly.
Watch out High technological immaturity means today's leaders' positions may not be sustainable; that's a real risk, but it's a competitive-dislocation risk, not proof of a financial bubble.
- 3
Read the capitulation signal
Gauge sentiment: a true bubble needs near-universal belief that it's NOT a bubble, which pushes prices out of control. Widespread bubble-warnings mean capitulation hasn't happened.
Pro tip Horowitz knew ~2011-12 wasn't a tech bubble precisely because ~1,400 articles were calling it one.
Watch out Accept that in any hot market both great and worthless companies get funded; that's normal venture behavior, not the bubble test.
In the wild
Horowitz was a CEO through the dot-com crash, where the technology was real but the unit economics didn't work, so once investors saw no revenue, funding stopped and companies went bankrupt. He contrasts this with 2025 AI: products working better than any technology ever built, companies scaling from 0 to $800M in a year, priced to that growth.
→ Horowitz concludes AI is likely not a bubble in the dot-com sense, expecting competitive dislocation rather than an economics-driven collapse, though he grants growth may not continue.
Common mistakes
Equating high prices with a bubble
Prices can be high because revenue is genuinely growing fast; the dot-com bubble was defined by broken economics and no revenue, not merely by elevated valuations.
Ignoring the capitulation signal
Calling something a bubble while everyone else is also calling it one misreads the setup; real bubbles form only when nearly everyone believes it's not a bubble.
Is it for you?
Best for
Investors and operators trying to judge whether a hot, high-priced market (like AI) is a genuine bubble before allocating or committing
Not ideal for
Making short-term price-timing calls; the framework assesses bubble structure, not when prices will move
From the transcript
“the one thing about bubbles is anytime everybody thinks it's a bubble, it's not a bubble”
“But the thing that made it a bubble was like the unit economics didn't work, the businesses didn't work.”
“Like these businesses are all working and they're being priced appropriately for how they're growing.”
From the episode
$46B of hard truths from Ben Horowitz: Why founders fail and why you need to run toward fear (a16z co-founder)