Gary Marcus, the cognitive scientist and longtime AI skeptic, has published a pointed critique of Anthropic's reported $30 trillion revenue target, calling it a fantasy that reveals more about Silicon Valley hubris than genuine business planning. The piece, published on his Substack, lands at a sensitive moment for the company as scrutiny over its financial projections intensifies ahead of a likely public market debut.
What Marcus Actually Argues
Marcus does not dispute that Anthropic has built capable AI systems. His argument is narrower and more pointed: that projecting $30 trillion in revenue requires assumptions about AI adoption and economic restructuring that are, in his view, nowhere close to being validated by current evidence. He draws a line between what AI can do in controlled demonstrations and what it will actually deliver at scale across the global economy. For context, $30 trillion represents roughly the entire annual GDP of the United States and European Union combined. Marcus argues that framing such a number as a revenue target, rather than a theoretical ceiling, reflects motivated reasoning rather than sober forecasting.
Key Facts
- $30 trillion is the reported revenue figure attributed to Anthropic's internal projections
- Gary Marcus is a professor emeritus at NYU and a prominent AI critic
- Anthropic is currently valued at close to $1 trillion ahead of a potential IPO
- The critique follows similar concerns raised by investors about AI sector overvaluation
- Marcus has previously challenged projection-driven narratives across the broader AI industry
The timing of the piece is notable. Anthropic has filed for an IPO as its valuation approaches $1 trillion, and public investors will soon be asked to price that valuation with real money. Marcus is essentially arguing that the revenue projections underpinning that valuation deserve the same skepticism applied to any pre-revenue startup claiming world-historical numbers. He is not alone in that view. Michael Burry, the investor famous for shorting the housing market, has already drawn parallels between Anthropic's IPO ambitions and the dot-com era, suggesting the pattern of hype outrunning fundamentals looks familiar.
"The $30 trillion figure isn't a forecast. It's a wish."Gary Marcus, Marcus on AI Substack
Why This Critique Matters Beyond the Numbers
Marcus's challenge is not purely about arithmetic. It touches on a deeper tension inside the AI industry between the long-term case for transformative technology and the near-term reality of building sustainable revenue. Anthropic has positioned itself as a safety-focused lab that can also win commercially, and that dual identity requires it to convince investors it can scale revenue while maintaining its research mission. Projections in the tens of trillions make the commercial case sound bulletproof, but Marcus argues they do the opposite by straining credibility.
The company has been expanding its commercial footprint in meaningful ways. Its push into specialized verticals, including a recent move to target the pharmaceutical sector, signals an awareness that general-purpose AI revenue has limits. Whether those moves can close the gap between today's business and the projections Marcus finds fantastical is the question investors will have to answer for themselves when shares eventually begin trading.
For now, the critique adds another voice to a growing chorus questioning whether the entire AI sector has priced in outcomes that may take decades to materialize, if they arrive at all. Marcus has been making versions of this argument for years, and the industry has often dismissed him. But with IPO filings now translating hype into prospectus language, the stakes for getting the numbers right have gone up considerably. Whether Anthropic's projections hold up to that scrutiny will matter well beyond one Substack post.
