Synthesis

Infinity or zero

Sam Lessin's shorthand for how investors model AI companies: with no year-over-year growth model, every model collapses to a probability of infinity (or zero), so tails dominate a…

Type
Synthesis
Confidence
Low
Sources
11 videos
First published
Updated

Across 11 episodes, Sam Lessin argues that AI companies are valued as a probability-weighted bet on infinity with no growth model underneath, so any signal gets overread (the talent-war reaction, per Jess’s and Sam’s comments in dbdH6dPMDxk). He links this to the ‘Hail Mary for infinity’ mood among investors who think compounding about 10% a year is not enough, to seed outcomes being top-heavy (so he favours companies with a story of infinity), and to his rule of not shorting anything where infinity is possible. In the St Petersburg framing from an unnamed host, an infinite payout has no price. Sam also argues median and mean diverge: the mean outcome can be hugely positive while the median is zero. Counter-notes: Jess says the likelier result is a valuation reset rather than death and that valuations assume impossible permanent growth, and Sam concedes lab leaders might be right although history suggests otherwise. Sam says pricing AI labs is impossible given easy switching. All of this is opinion, not modelling.

Key points

  • With no growth model, valuations collapse to infinity or zero and signals get overweighted — 2 episodes. (Sam · Chinese AI Model GLM 5.2 Beati (Jun 2026) ▶ 24:19)
  • Infinite outcomes cannot be priced; St Petersburg paradox framing (an unnamed host) — 1 episode. (one of the hosts · Matt Mazzeo (Sep 2026) ▶ 27:07)
  • Investors are on a ‘Hail Mary for infinity’ because ordinary compounding is felt as not enough — 1 episode. (Sam · Elon’s Trillion-Dollar Move, W (Feb 2026) ▶ 14:56)
  • Sam favours stories of infinity over fine money-making machines, and will not short where infinity is possible — 2 episodes. (Sam · Founders Beware: Illiquid Valu (Jan 2026) ▶ 23:15)
  • Median and mean differ: a mean can be hugely positive when the median is zero, which draws gamblers and retail — 1 episode. (Sam · SpaceX Hit $2.5 Trillion. Now (Jun 2026) ▶ 14:41)
  • Sam says AI-lab valuations cannot be financially priced and are high-variance narrative bets — 1 episode. (Sam · Stripe Buys OpenRouter, OpenAI (Aug 2026) ▶ 30:48)
  • A likelier end than death is a valuation reset; valuations assume permanent growth across industries (Jess) — 1 episode. (Jess · Adrian (Oct 2026) ▶ 16:37)
  • The ‘infinity story’ is the labs’ pitch that one company takes all, said to be fading — 2 episodes. (Sam · The Internet Is Fighting Back (Oct 2026) ▶ 11:26)

Where guests agree

  • The hosts treat AI-lab valuations as bets on the tails rather than as projections from cash flow (one show, mainly Sam’s framing) (Sam · Chinese AI Model GLM 5.2 Beati (Jun 2026) ▶ 24:19)

Frameworks

  • Infinity or zero (zero, infinity and the mean) — Sam’s description of how investors currently model AI companies: a probability-weighted bet on infinity with no growth model beneath it. (Sam · Chinese AI Model GLM 5.2 Beati (Jun 2026) ▶ 24:19)
    • No year-over-year growth to anchor a model
    • Any incremental signal is traded on faster than others
    • So signals cycle into a bigger story than they deserve
  • Infinity or zero (zero, infinity and the mean) — Sam’s account of how the Valley’s growth narrative moved from Google and Facebook’s geometric returns to crypto and AI stories where the tails (infinity or zero) dominate; the inverse is infinite negative liability. (Sam · Silicon Valley Is Great at Tec (Oct 2026) ▶ 15:46)
    • Google and Facebook: unexpectedly large outcomes
    • Bitcoin, Ripple, Solana, GameStop: belief as self-perpetuating, average of zero and infinity is infinity
    • AI labs: own everything or nothing; religious framing
    • Inverse tail: infinite liability, P(doom), which says spend everything to stop it
  • Infinity or zero (zero, infinity and the mean) — Sam argues that if the market is effectively infinite, the mean outcome can be hugely positive even if the median outcome is zero or negative, which is why gamblers and retail buy in. (Sam · SpaceX Hit $2.5 Trillion. Now (Jun 2026) ▶ 14:41)
    • median outcome might be negative
    • believe the total addressable market is infinite
    • mean of zero and infinity is infinity
  • Two ways to be red-pilled on AI and the asymmetry card — Sam’s explanation of why AI infrastructure leaders talk about infinite demand. (Sam · Agentic AI Will Break the Inte (Jan 2026) ▶ 27:52)
    • Actually red-pilled: you genuinely believe the incredible claims
    • Not quite red-pilled, but the cost-benefit of going all in narratively favours doing so
    • It does not have to be right, only what people continue to believe

Tactics

  • Choose companies with a story of infinity, not only a solid money-making business. (Sam · Founders Beware: Illiquid Valu (Jan 2026) ▶ 23:15)

Stories & examples

  • St Petersburg paradox: A coin-flip game whose expected payout is infinite, so the logical price to play is infinity even though no one would pay it; used to show that infinite outcomes cannot be priced. (one of the hosts · Matt Mazzeo (Sep 2026) ▶ 26:27)

Numbers

  • Normal-stock compounding return Sam says is not good enough: 10 percent per year, net of taxes and inflation (host’s framing, approx.) (Sam · Elon’s Trillion-Dollar Move, W (Feb 2026) ▶ 14:56)
  • Illustrative P(doom) probabilities in Sam’s argument: 1 to 10 percent (host’s framing, approx.) (Sam · Silicon Valley Is Great at Tec (Oct 2026) ▶ 19:24)
  • Coin-flip first payout in the St Petersburg example: 2 dollars (self-reported) (one of the hosts · Matt Mazzeo (Sep 2026) ▶ 26:12)

Caveats & warnings

  • Sam concedes the AI leaders might be right even though history says otherwise. (Sam · Agentic AI Will Break the Inte (Jan 2026) ▶ 26:20)
  • An unnamed host calls the St Petersburg analogy ‘not like normal math’; it is a framing, not a model. (one of the hosts · Matt Mazzeo (Sep 2026) ▶ 27:40)
  • Sam notes the open-versus-closed model chart can be read two ways because the overall pie is getting bigger. (Sam · Adrian (Oct 2026) ▶ 18:22)

How much to trust this

  • Evidence: 11 episodes, 1 independent speakers (the four hosts and unidentified host speech count as one voice, the show; guests counted individually).

Blueprints that use this

Sources

  1. (01)

    OpenAI vs Google vs Meta: Business Model War

    YouTube

    1:04:19 into “OpenAI vs Google vs Meta: Business Model War”

  2. (02)

    Founders Beware: Illiquid Valuations vs. Real‑World Taxes

    YouTube

    23:15 into “Founders Beware: Illiquid Valuations vs. Real‑World Taxes”

  3. (03)

    AI Can Change Everything and Still Be a Terrible Business + Peptide 101 from Matt Mazzeo

    YouTube

    26:12 into “AI Can Change Everything and Still Be a Terrible Business + Peptide 101 from Matt Mazzeo”

  4. (04)

    Agentic AI Will Break the Internet (ClawdBot, OpenAI & the AI Demand Delusion)

    YouTube

    26:20 into “Agentic AI Will Break the Internet (ClawdBot, OpenAI & the AI Demand Delusion)”

  5. (05)

    The Internet Is Fighting Back Against Your AI Agent

    YouTube

    11:26 into “The Internet Is Fighting Back Against Your AI Agent”

  6. (06)

    Stripe Buys OpenRouter, OpenAI vs. Anthropic & the AI Backlash

    YouTube

    30:48 into “Stripe Buys OpenRouter, OpenAI vs. Anthropic & the AI Backlash”

  7. (07)

    SpaceX Hit $2.5 Trillion. Now Watch What Elon Buys Next.

    YouTube

    10:39 into “SpaceX Hit $2.5 Trillion. Now Watch What Elon Buys Next.”

  8. (08)

    Chinese AI Model GLM 5.2 Beating Frontier Models | Meta Glasses, Polymarket Scandal, AI Talent War

    YouTube

    23:34 into “Chinese AI Model GLM 5.2 Beating Frontier Models | Meta Glasses, Polymarket Scandal, AI Talent War”

  9. (09)

    Meta Muse vs Instinct vs OpenAI Dots: Do AI Agents Have Any Moat?

    YouTube

    16:37 into “Meta Muse vs Instinct vs OpenAI Dots: Do AI Agents Have Any Moat?”

  10. (10)

    Silicon Valley Is Great at Tech and Terrible at Business

    YouTube

    13:33 into “Silicon Valley Is Great at Tech and Terrible at Business”

  11. (11)

    Elon’s Trillion-Dollar Move, WaPo Layoffs & The Super Bowl

    YouTube

    14:56 into “Elon’s Trillion-Dollar Move, WaPo Layoffs & The Super Bowl”

(By email)

Compounding Insights