Doom Bubble Lens for Tech Cycles
The technology can win even as the stocks lose.
The Doom Bubble Lens separates two questions investors routinely conflate: will the technology matter, and will the stocks priced as 'the technology' deliver returns? Coffin uses the dot-com era to show that the underlying technology — the internet — succeeded enormously while the average startup attached to it failed. Survivors like Amazon and Google are rarities, not the base rate.
Applied to AI, the lens says it is entirely consistent for AI capabilities to keep getting tremendously better while a meaningful slice of AI-branded stocks turns out to be 80–90% drawdowns. Bank of America's poll showing 40% of fund managers calling AI a bubble captures exactly this asymmetry.
The lens's value is suppressing the urge to bet the portfolio on a thematic story, while still respecting the technology's real-world impact.
- Underlying technology success and stock-market success are different questions.
- Survivor bias makes Amazon and Google look like the base rate when they are outliers.
- Hype cycles overestimate near-term impact and underestimate long-term diffusion (Amara's Law).
- Themes should not drive position sizing — fundamentals should.
- Be more sceptical the closer a company's pitch is to a single hot keyword.
- Separate the two betsWrite down two distinct claims: (a) this technology will matter in 10 years; (b) this specific stock will deliver returns. Treat them as independent.WarningIf you can't justify (b) without (a), the position is a theme, not a thesis.
- Map attrition riskList the categories of companies competing in the wave (foundation models, infra, applications, hardware) and estimate how many survive. Assume most don't.
- Look for picks-and-shovels with durable economicsIdentify whether the company benefits regardless of which application layer wins, and whether the economics are defensible (pricing power, switching costs).Pro tipAsk whether the moat survives if the hot keyword goes out of fashion.
- Stress-test against survivor biasForce yourself to name the dead companies of the comparable prior cycle (e.g. Globe.com, Pets.com) and ask which of today's names you'd be looking back at in the same way.
- Anchor sizing in fundamentals, not narrativeSet position sizes based on business quality, valuation, and diversification — not on how 'inevitable' the narrative feels.WarningConviction in the technology is not conviction in the stock.
Globe.com was an early social-media name many thought was the future. It no longer exists, while the internet itself reshaped the world. The technology won; the early stocks largely didn't.
Around Canadian marijuana legalisation, themed stocks skyrocketed. Coffin's firm avoided them entirely; most are down 80–90% from those peaks.
Nvidia's run looks 'obvious' as the picks-and-shovels seller of the AI gold rush, but Coffin notes this clarity is hindsight bias and not a common occurrence across all peers.
Coffin frames this around Amara's Law — we overestimate tech in the short term and underestimate it in the long run — and his colleague Ben Felix's research that technological revolutions have not been great for stock-market returns once attrition is counted. The dot-com Globe.com example anchors the framing.