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

2026-06-04 - 33 min - source - Read full transcript

Thomas LaffontChamath Palihapitiya (host)Jason Calacanis (host)David Sacks (host)

Key insights

AI is concentrating venture funding into fewer, much larger rounds rather than creating more unicorns.

Coatue's data shows unicorn creation has normalized back to pre-COVID levels after the 2021 ZIRP peak, but funding per unicorn has increased 5x since 2021 because AI companies are capturing an outsized and growing share of total venture dollars.

power-law-concentration

A private-company 'Magnificent Eight' - SpaceX, Stripe, Anthropic, Databricks, Revolut, ByteDance, Anduril, and one more - represents roughly $4 trillion of value and has outperformed the public Magnificent Seven.

Laffont frames this basket as the next-generation index: diversified across internet, AI, fintech, and space tech, with almost every constituent outperforming the public Mag 7 even before any of them have gone public.

ai-ipo-wave

The unicorn ecosystem's cash-return profile is improving even before the largest expected 2026 liquidity events land.

2026's ratio of cash returned to cash consumed by the unicorn economy is already tracking close to the 2021 peak, and that is before SpaceX's expected IPO in the coming weeks and Anthropic's confidentially-filed S-1 are counted - adding those three companies alone would roughly match the prior decade's combined liquidity.

ai-ipo-wave

OpenAI and Anthropic's revenue growth has outpaced every major SaaS company that came before them and is projected to overtake AWS and all of Microsoft's revenue within a few years.

Starting from a January 2025 baseline, the two companies' revenue trajectories passed Workday, then ServiceNow, then Adobe, then Salesforce within about 18 months, and Coatue's forecast has them passing Google Cloud and Azure now, AWS by the end of the year, and potentially all of Microsoft by 2028.

ai-revenue-growth

SpaceX's valuation is best explained by a four-phase business-model framework tied to launch cadence, not by current revenue alone.

Coatue's framework moves a space company through pre-constellation (one-off, unpredictable government revenue), ramp (first constellation, recurring subscriber revenue), scale (multiple constellations serving governments and militaries that want to control their own infrastructure), and platform (new businesses like space data centers) - each phase increases the market's willingness to pay a higher valuation per launch.

valuation-frameworks

The odds of a company achieving a 10x outcome rise sharply once it reaches the $100B+ 'centacorn' tier, suggesting scale itself becomes a filter for durability.

Coatue's data shows roughly an 8% chance a unicorn becomes a decacorn, and a similar 8-13% chance a decacorn becomes a centacorn, but once a company crosses $100B in value (public or private), its odds of a further 10x jump to 31% - Laffont attributes this to compounding advantage and durability of earnings surviving each successive filter.

power-law-concentration

The total AI ecosystem revenue is roughly $140B today, projected to reach ~$300B this year and double again by 2027, split across consumer, advertising, and enterprise.

Coatue breaks the revenue into three pillars: consumer subscriptions (subs times ARPU), AI-enabled advertising (currently about a quarter of Meta and Google ad inventory, expected to eventually reach 100% penetration and add roughly $150B), and enterprise tools like Claude Code and Codex.

ai-revenue-growth

No new $100B+ 'centacorn' has emerged in the private markets in the past couple of years, which Laffont flags as a potential warning sign for ecosystem health.

The count of companies at the centacorn tier has been roughly flat for a while; if that count stays flat over the next decade rather than growing, Laffont says it would signal the private-market power-law dynamic is narrowing rather than producing new scaled winners.

power-law-concentration

Today's trillion-dollar AI companies are structurally different from the 2000 dot-com bubble or 2021 ZIRP-era companies because they generate substantial, fast-growing, and in Anthropic's case reportedly profitable, revenue.

Laffont pushes back on valuation skepticism by noting these are decades-old technology categories (not first-time-revenue dot-com names), growing faster than anything seen before, and that Anthropic was reported to have had a profitable month - evidence he says the growth is not just narrative.

ai-revenue-growth

Going public will be the 'great equalizer' for AI valuations, but heavy passive/index buying may delay when the market actually stress-tests them.

Chamath and Laffont agree the public market disciplines valuations that private markets can't, but note that unlike a traditional IPO where price discovery happens on day one, high demand for passive inclusion will push out the real test of these valuations to roughly six months post-listing, once initial supply/demand imbalances wash through.

ai-ipo-wave

A strategy of annually rebalancing into the top 10 Nasdaq companies by market cap has historically outperformed a static Nasdaq buy-and-hold by roughly 3x over a decade.

Cited in the Q&A as supporting evidence for staying concentrated in compounding winners rather than diversifying broadly, since the largest companies by market cap tend to keep compounding faster than the index average.

valuation-frameworks

AI is disrupting essentially every sector of the economy simultaneously, a breadth Laffont says he hasn't seen in prior technology cycles.

He points to telecom (Starlink addressing the full global broadband/wireless profit pool), semiconductors (data-center buildout reshaping state energy economics), autos (Ferrari's stumble introducing EV and autonomous technology), and consumer behavior (GLP-1 drugs changing food and alcohol consumption) as concurrent, unrelated sectors all being reshaped by the same underlying capital and technology wave.

ai-revenue-growth

Books referenced

Foundation series - Isaac Asimov - Laffont compares Claude Code's effect on Anthropic's trajectory to the Mule in Asimov's Foundation - an unpredictable single event that bends an entire system's forecasted path.

Companies

Coatue Management - Laffont's $55B hedge fund; the presentation is Coatue's annual private-market data update to the besties.
SpaceX - Expected to IPO within weeks; Laffont uses it as the case study for a launch-cadence-based valuation framework.
Anthropic - Confidentially filed its S-1 the same day as the taping; reportedly posted its first profitable month; central example of AI revenue growth outpacing prior SaaS companies.
OpenAI - Compared against Anthropic on growth trajectory; floated as a candidate for a future price war given both companies' cash reserves.
Stripe - Named as one of eight private companies in Coatue's ~$4T 'Magnificent Eight' index.
Databricks - Named as one of eight private companies in Coatue's 'Magnificent Eight' index.
Revolut - Named as one of eight private companies in Coatue's 'Magnificent Eight' index.
ByteDance - Named as one of eight private companies in Coatue's 'Magnificent Eight' index.
Anduril - Named as one of eight private companies in Coatue's 'Magnificent Eight' index.
Cerebras - Just completed its IPO; Laffont was a long-time board member and led its Series B, citing years of slow funding before a major OpenAI contract quintupled its value.
Ferrari - Cited as an example of a legacy company struggling to reposition around EV and autonomous technology.
Starlink - Argued to be addressing the entire global telecom/broadband profit pool, not just a niche satellite business.
Micron - Raised in Q&A as a hypothetical alternative allocation versus Coatue's diversified private bets.
Andreessen Horowitz - Cited as an example of a venture firm moving toward broad, index-like venture investing rather than concentrated bets.

Techniques and frameworks

Launch-cadence valuation framework - Coatue's four-phase model for SpaceX-like businesses: pre-constellation (one-off government revenue), ramp (one recurring-revenue constellation), scale (multiple constellations), platform (new adjacent businesses like space data centers) - each phase raises the quality and durability of the business model.
Cohort step-up probability analysis - Coatue's data showing the odds a company advances tiers: unicorn to decacorn ~8%, decacorn to centacorn ~8-13%, but centacorn (100B+) to a 10x return jumps to 31%, implying scale itself becomes a filter for durable compounding advantage.
Ben Graham-style durability filter - Framework the besties apply on stage for judging which companies reach 'trillion-dollar club' status: does the company have a compounding advantage and durable earnings at each step, similar to classic value-investing quality screens.
Top-10 Nasdaq rebalancing strategy - Cited data point: annually rebalancing into the top 10 Nasdaq companies by market cap outperformed a static Nasdaq buy-and-hold by roughly 3x over a decade.

Summary

Thomas Laffont, co-founder of Coatue Management, delivered his annual private-market data presentation to the besties on the main stage of the All-In Summit, walking through the state of the "unicorn economy" as AI reshapes both public and private markets. His headline chart shows funding concentrating into fewer, larger AI rounds: unicorn creation has normalized back to pre-COVID levels after the 2021 ZIRP peak, but funding per unicorn has risen 5x since then. He introduced a private-company "Magnificent Eight" - SpaceX, Stripe, Anthropic, Databricks, Revolut, ByteDance, Anduril, and an eighth name - worth roughly $4 trillion combined and outperforming the public Magnificent Seven, framing it as the next-generation index an investor would want to hold for the next decade.

Laffont spent significant time on a framework for valuing SpaceX that ties its valuation not to current revenue but to launch cadence, moving the company through four phases - pre-constellation, ramp, scale, and platform - each of which adds recurring revenue and optionality (space data centers, lunar/Mars applications). He paired this with cohort data on step-up probabilities: roughly an 8% chance a unicorn becomes a decacorn, a similar chance a decacorn becomes a centacorn, but a 31% chance a $100B+ centacorn achieves a further 10x - evidence, he argued, that scale itself filters for durable compounding advantage, echoed by the besties' comparison to Ben Graham-style earnings-durability screens.

On revenue, Coatue's data shows OpenAI and Anthropic's growth trajectories have already overtaken Workday, ServiceNow, Adobe, and Salesforce since January 2025, are now passing Google Cloud and Azure, and are forecast to surpass AWS by year-end and all of Microsoft by 2028. Laffont sized the total AI ecosystem at roughly $140B today, growing to $300B this year and doubling again by 2027, split across consumer subscriptions, AI-enabled advertising (currently about a quarter of Meta/Google ad inventory), and enterprise tools like Claude Code and Codex. He pushed back directly on bubble comparisons, noting these are decades-old businesses with real, fast-growing revenue - and that Anthropic reportedly posted its first profitable month - unlike dot-com-era companies with no revenue at all.

The conversation turned to what happens once these companies actually go public. Laffont and Chamath agreed the public market is the "great equalizer" that will subject SpaceX, OpenAI, and Anthropic to real scrutiny for the first time, though heavy passive-index buying may push the real valuation stress-test out to roughly six months post-IPO rather than day one. Laffont also flagged a possible AI price war between OpenAI and Anthropic as a rational, if unpredictable, use of their large cash balances, and noted the private markets have not produced a new $100B+ centacorn in the past couple of years - something he called a potential warning sign if the trend continues.

He closed by arguing the breadth of disruption in this cycle is unlike anything he's tracked before: Starlink addressing the entire global telecom profit pool, data centers reshaping state-level energy economics, Ferrari struggling with EV and autonomy, and GLP-1 drugs changing consumer food and alcohol spending - all happening concurrently, and all before, as he pointed out, the industry even has superintelligence.

Notable Quotes

"The public market is the great equalizer... it will be the great antiseptic. It will not care about my presentation." - Chamath Palihapitiya

"These are not fake companies... they trade at the lowest multiple of earnings of the S&P 500 of almost any other company." - Thomas Laffont

"Anthropic pre-Claude code was a completely different company than post-Claude code... it's hard for me to know whether that's truly... like the mule in the Foundation series, something that could never be predicted." - Thomas Laffont

"If I want to design a chip like OpenAI, I can go to TSMC... If I want to make memory, well there is no TSMC." - Thomas Laffont