Thomas Laffont: The $4T AI IPO Wave Is Coming… and We’ve Never Seen Anything Like It

Why does it matter? Because a $100B company is statistically a better 10x bet than a startup — and three IPOs are about to prove it

Thomas Laffont of Coatue — $55 billion under management — spent two weeks building these slides. The data runs directly against the conventional framing of venture risk: that early-stage exposure is where the real upside lives, and that mega-caps are where returns compress.

• A $100B+ company has a 31% chance of delivering another 10x — nearly four times the odds of a unicorn reaching decacorn
• SpaceX, OpenAI, and Anthropic's combined IPOs will return more than the entire prior decade of venture exits in a single wave
• 25% of Meta and Google's ad inventory is AI-enabled today, heading to 100% — a $150B layer most AI market sizing ignores entirely
• 479 companies became unicorns in 2021; fewer than 20% have exited or raised since — and the 2024 AI cohort's fate hangs on which line it follows

The 31% rule: centacorns are the best risk-adjusted 10x bet in the market

Unicorn becoming decacorn: 8% odds. Decacorn to centacorn: 13%. Centacorn — $100B or more — achieving a 10x from there: 31%. "This kind of flies in my opinion in different than maybe we would have expected," Laffont says.

Sacks immediately extrapolates to trillion-dollar companies, guessing north of 30% there too. Laffont agrees: each threshold is a filter for compounding durability. "To get to that level that's called the trillion dollar club, you have to have a dominant business. And then the question is just at what point do you hit saturation." So far, every market these companies entered has proven larger than anyone predicted.

The allocation implication cuts hard: concentrating in existing centacorns is statistically superior to diversifying across early-stage bets hoping to manufacture the next one.

SpaceX's valuation per launch rises as it launches more — and the market is pricing the right thing

Valuation divided by number of launches was flat for years. Then it broke sharply upward. Laffont's CODE framework explains the inflection: each launch phase transforms what the business actually is.

Pre-constellation, SpaceX sells government contracts — one-time, unpredictable. Deploy a constellation and the model flips: recurring subscribers, recurring revenue. Multiple constellations become a platform. And the platform's addressable market isn't the launch business. It's the global telco profit pool — "anywhere between 2 to 400 billion" — that Starlink is attacking with a product requiring no radio towers and delivering coverage everywhere.

Analysts pricing SpaceX on launch cadence are using the wrong denominator. The right one is the broadband and wireless profit pool it's systematically moving into.

Anthropic passed Azure 18 months in — and could exceed all of Microsoft by 2028

Azure fell. Laffont's revenue chart starts in January 2025, and within months Anthropic and OpenAI had already lapped Workday, then ServiceNow, then Adobe, then Salesforce in January, now Google Cloud and Azure both. By end of 2026, potentially bigger than AWS. By 2028, potentially bigger than all of Microsoft.

One product moment bent the entire curve: "Anthropic pre-cloud code was a completely different company than post-cloud code. So one event completely dented the trajectory of almost that entire industry."

Pricing these companies on current revenue multiples is a category error. The denominator is growing faster than the multiple can expand — which means the multiple only looks stretched to people using a denominator that's already stale.

AI has a $150B revenue layer hiding inside Meta and Google's existing P&Ls

$140 billion in AI revenue today, $300 billion by year-end, doubling again in 2027. Laffont breaks the ecosystem into three pillars: consumer subscriptions, enterprise cloud code and Codex, and the one almost everyone skips — advertising.

"One that I think a lot of people forget, but it's ads." About 25% of Meta and Google's current inventory is AI-enabled. Penetration, Laffont argues, will eventually hit 100%. That's $150 billion sitting inside existing platform financials — never labeled AI revenue, never surfacing in headline market sizing reports. Add it back and the real AI economy is roughly twice what most coverage claims.

479 zombie unicorns from the 2021 vintage are stranded — don't confuse that cohort with the AI wave

Pre-ZERP: 73 unicorns, 80% had exited or raised a new round within 20 quarters. The 2021 cohort: 479 companies — six times larger — with fewer than 20% at the same mark. The red line barely moves.

The 2024 AI cohort is arriving now and Laffont asks directly which line it follows. His funding data implies the answer: capital per unicorn is up 5x since 2021, concentrated into a small number of names. "The winners are compounding faster than ever, which means the costs of not being in a winner are higher than ever." The 2021 vintage isn't context for the AI moment — it's a cautionary tale parked right beside it.

A rational price war between OpenAI and Anthropic is the wildcard nobody's modeling

Chamath raises the ridesharing analogy and Laffont lands cleanly on the conclusion: if either lab enters public markets with hundreds of billions in fresh capital, the playbook says you use it to price out the competition. "Rationally, they should," Laffont agrees. The only brake right now is infrastructure burn — they're spending too much to make the math work. Post-IPO, that constraint loosens. Every enterprise AI contract signed at current rates assumes a pricing floor that a single competitive move could eliminate. The companies that built the AI era have every incentive to torch its margins once they can afford to.

---

Topics: AI, venture capital, private markets, IPO, SpaceX, Anthropic, OpenAI, unicorn economy, hedge funds, Coatue, centacorn, power law, AI revenue, capital allocation