The $4T private Magnificent 8: Coatue's Thomas Laffont makes the bull case for AI compounding

On 5 June 2026, at the All-In Summit in Los Angeles, Coatue Management co-founder Thomas Laffont walked a room of venture capitalists, allocators, and operators through a $4 trillion thesis: the most valuable companies of the next decade are private, AI-driven, and compounding at a speed the public markets have never seen. Coatue, the $55 billion hedge fund, is raising an additional $1 billion specifically to add to those private AI positions. The framing — drawn from decades of public-equity data cross-referenced against Coatue's private portfolio book — amounted to a counter-narrative to a year of bubble-comparison headlines.

Laffont's argument is not that everything is fine. It is that dispersion inside the private market has widened so dramatically — and is now legible in operating data rather than pitch decks — that the disciplined play is to concentrate capital at the top, not retreat from private. The "Magnificent 8," a Coatue-constructed index of eight private companies now worth nearly $4 trillion combined, has outperformed the public Magnificent 7. That gap, Laffont argued, will close the moment these companies begin to list — and the IPO wave from SpaceX, OpenAI, and Anthropic alone will deliver more proceeds than the prior ten years of public listings combined.

The Magnificent 8 and a private index that dwarfs the public one

Laffont's proposed index — SpaceX, OpenAI, Anthropic, Stripe, Databricks, Revolut, ByteDance, and Anduril — represents roughly $4 trillion in private value. "These are not fake companies. These are companies generating substantial revenue at scale that are growing faster than anything we've ever seen," Laffont told the audience. The index, he said, has crushed the public Magnificent 7 over the relevant lookback window — a back-of-envelope claim that the public-market cohort has now lost its monopoly on leadership returns.

That outperformance is the data point Coatue is betting on. Chamath Palihapitiya, on the same panel, cited a separate backtest showing a strategy that rebalances annually into the top 10 Nasdaq names by market cap outperformed a passive alternative by roughly 3x over a decade. The implication is uncomfortable for diversified public-equity allocators: the same concentration mechanic now operates inside private markets, where Coatue has structural access. "The new unicorn economy is healthier and we really have AI to thank for that. The winners are compounding faster than ever which means the costs of not being in a winner are higher than ever," Laffont said.

The unicorn cohort is healthier than the ZIRP wreckage — but with a sharp edge

Laffont's most pointed chart was historical. He split the unicorn universe into two cohorts: 73 pre-ZIRP companies, and 479 unicorns minted in the 2021 free-money vintage. Twenty quarters later, fewer than 20% of the 2021 cohort had raised a new round or exited. The pre-ZIRP cohort cleared 80%. The implication is not subtle: roughly four out of every five 2021 unicorns are zombies or worse, while the survivors that did make it through have become the AI compounder base.

Unicorn creation has since normalised to pre-COVID levels. Average unicorn markups are up roughly 70% since September 2024, mirroring the public-market recovery. But funding per unicorn has increased 5x since 2021 — fewer companies, much larger cheques. That, Laffont argued, is the healthier shape: capital is concentrating where the operating data is concentrating, and the long tail is being allowed to die quietly. "The power law rules our lives. All the great gains are being consolidated into small numbers of companies but we're still seeing strength in those," he said.

Anthropic, the AI revenue stack, and the case for compounders

The most aggressive claim in Laffont's deck was the trajectory of Anthropic. According to a chart he presented, Anthropic has, in roughly twelve months, blown past the annualised revenue of Workday, ServiceNow, Adobe, Salesforce, Google Cloud, and Azure in sequence. He projected Anthropic could surpass AWS in annualised run-rate revenue by the end of 2025, and, if growth holds, exceed all of Microsoft by 2028. "Anthropic pre-Claude Code was a completely different company than post-Claude Code — one event completely dented the trajectory of almost that entire industry," he said, framing Claude Code as the inflection point.

The AI revenue stack, as Laffont sized it, runs from roughly $140 billion in current AI-attributable revenue to $300 billion by year-end 2026, and doubles again in 2027. He broke it into three pillars: consumer subscriptions priced on ARPU, AI-enabled advertising — currently about 25% of Meta and Google ad inventory, with 100% penetration implying a $150 billion opportunity — and enterprise software anchored by Claude Code and OpenAI's Codex. The ad-pillar math is mechanical: if every ad impression becomes AI-personalised, the addressable market doubles in unit economics even without pricing changes.

He also pointed to public reports that Anthropic had a profitable month recently — a fact he raised explicitly to push back against bubble comparisons. "The public market is the great test equalizer, the great antiseptic. It will not care about my bullshit presentation," he said, anticipating the discipline an IPO will impose.

The CODO framework, Starlink, and the telco profit pool

For SpaceX, Laffont introduced what he called a "CODO" — an internal Coatue valuation framework anchored to launch cadence and constellation platform effects. The addressable profit pool, he argued, is not the launch-services market but the global telco broadband and wireless market, estimated at $200–$400 billion. "Starlink works all the time, no radio towers. We think that's a solved problem, but every time we get a dropped call, we get reminded there's a better technology out there," he said.

The framework reframes SpaceX at roughly $1.75 trillion as a telco-displacement play that happens to be bundled with a launch business, rather than the reverse. The profit-pool displacement logic extends well beyond SpaceX: Laffont flagged major displacements ahead across telecommunications, semiconductors, autos, and consumer health, as AI-driven cost curves force incumbents to reprice. The semiconductor case is asymmetric. Paraphrasing a contact, Laffont noted: "If I want to design a chip like OpenAI, I can go to TSMC and I know it's hard but at least I have TSMC to help me. If I want to make memory, well, there is no TSMC." That bottleneck, he argued, will be a defining constraint on how fast the AI revenue stack can scale.

Power law math, the bubble question, and the IPO test

The most uncomfortable part of the deck was the returns distribution. Laffont laid out the odds of a 10x return: roughly 8% for unicorns becoming decacorns, 8–13% for decacorns becoming centicorns ($100 billion+), and 31% for centicorns achieving a 10x from that base. The math implies that an LP diversifying across early-stage unicorns is, on average, underwriting a long-tail loss; an LP that concentrates at the top decile captures the 31% bracket. Three companies crossed from $500 billion to $1 trillion in market cap in the same calendar year, with two doing it in a matter of weeks — a clustering the historical record has no analogue for.

Laffont is not blind to the bubble framing. He concedes the 2021 cohort is a graveyard and acknowledges the ZIRP-era wreckage. But the case he built is that the survivors of that wreckage — and the AI cohort that has emerged in their place — are operating businesses with revenue, gross margin, and growth profiles that the 2000 and 2021 vintages never had. An OpenAI–Anthropic price war, he conceded, is rationally likely given their capital bases, even if infrastructure spend limits how far either can push the lever. The public market, in his framing, will arbitrate.

The bet Coatue is making — with an extra $1 billion in fresh allocation — is that the Magnificent 8, listed or not, will compound through the next leg. The risk is that the concentration trade cuts both ways: the same 31% centicorn bracket that Laffont is leaning into is the one a 2000-style repricing would hit first. The All-In Summit was a data-rich argument that the second outcome is unlikely. The public listings, beginning as soon as the regulatory windows open, will be the verdict. As Palihapitiya put it: "When three or four trillion dollars gets put back to GPs, then to LPs, and then the recycling happens... time to sell San Francisco homes, anybody interested in a 40,000 square foot mausoleum?"

Desk note: Where wire coverage has framed Laffont's panel as a private-AI hype session, Monexus treats it as a structured counter-bubble argument — anchored to the 73-vs-479 cohort split, the 8%/8–13%/31% return brackets, and the $140B-to-$300B revenue stack — and separates Coatue's specific private bets from the broader private-market data it cited.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://www.youtube.com/watch?v=UIoV8rG_25s