Did the AI Bubble Just Pop?! Anthropic's Leaked Numbers are INSANE

Finance-Focused Summary

The discussion argues that the AI boom may be vulnerable to a debt- and liquidity-driven downturn, even if AI proves transformative and some companies ultimately become highly profitable. The timing and severity are uncertain. The central risk is that spending and borrowing may outpace revenue for too long, causing lenders to pull back before projected profits arrive.

Anthropic Figures Discussed

The speakers describe Reuters as having reported leaked, preliminary figures from an Anthropic filing. They emphasize that the figures are unaudited and should be treated cautiously.

• Revenue: About $4.6 billion in 2025, reportedly up 12× year over year.
• Operating losses: About $8 billion. The speaker characterizes this as a meaningful cash-flow concern, distinct from accounting losses.
• Expenses: Approximately $12 billion, against roughly $4.6 billion in revenue.
• Total loss: About $42 billion in 2025, including substantial non-cash, mark-to-market, or “paper” losses related to equity and financing arrangements. The speakers caution that these losses may not represent cash leaving the company immediately, though they could still create future obligations.
• Committed spending: Approximately $518 billion over several years, not all due in a single year. The speaker presents this as a major financing burden and suggests the figure could rise.
• Valuation and fundraising: The discussion cites a prior valuation of roughly $950 billion, about $80 billion raised in a previous round, and a possible $2 trillion IPO valuation. These are presented as reported or prospective figures, not confirmed outcomes.

The speakers contrast the scale of these obligations with Anthropic’s current revenue, arguing that the valuation depends on sustained, exceptional growth and a future point at which revenue overtakes costs.

Main Investment and Market Risks

• Debt and liquidity: AI firms’ heavy capital requirements could make them dependent on continued borrowing. If lenders become skeptical, companies may have difficulty refinancing or meeting obligations—even if their businesses are growing rapidly.
• Revenue-growth assumptions: The investment case relies on revenue eventually growing faster than spending. Delays, lower-than-expected adoption, cheaper competing models, or a macroeconomic slowdown could push profitability further into the future.
• Rising costs and competition: Companies are competing for chips, data-center capacity, and energy. The discussion argues that competition could make stated spending commitments a lower bound rather than a ceiling.
• Interconnected financing: The speakers describe private-credit deals, special-purpose vehicles (SPVs), equipment collateral, and supplier guarantees as creating links among AI companies, chip suppliers, lenders, and investors. They warn that this complexity may make risks difficult to track.
• Supplier-backed financing: Nvidia and Broadcom are discussed as examples of suppliers providing guarantees or residual-value support that can help customers obtain cheaper financing for equipment purchases. The speakers compare this to earlier tech cycles in which suppliers helped finance customers who then bought their products.
• Historical analogies: The discussion compares the boom with the dot-com period and, in some financing respects, the 2006–08 credit cycle. It notes that the dot-com bust involved losses of roughly 90% in some areas and cites Nortel and Lucent as examples of companies damaged when customers and financing faltered. The speakers explicitly say this is not a prediction that 2008 will repeat.
• Rates and government borrowing: The speakers say 10-year Treasury yields are rising and describe competition for capital between AI borrowing and U.S. government debt issuance. They cite roughly $2 trillion in annual government borrowing and about $40 trillion in accumulated government debt as part of the broader backdrop.
• Potential wider effects: The discussion raises questions about AI’s effects on employment, consumer spending, and the broader economy, while emphasizing that the outcomes are uncertain.

Framework and Investing Considerations

• Compare revenue with spending and obligations: Track whether revenue growth is actually approaching the level needed to cover operating costs, capital investment, and debt service.
• Distinguish cash losses from accounting losses: Treat operating cash deficits as different from non-cash valuation adjustments, while still examining whether those adjustments could create future liabilities.
• Follow the funding chain: Identify who supplies capital, who guarantees the debt, what assets secure it, and whether the same suppliers benefit from the proceeds.
• Test the growth assumptions: Consider how lower-cost models, open-source competition, weaker demand, or a downturn could affect pricing, adoption, and the timeline to profitability.
• Consider where value may accrue: The speakers contrast investing in specific AI companies with investing in infrastructure. They suggest infrastructure may be more durable if AI remains widely used, but caution that infrastructure investments can still carry concentrated risks—for example, exposure to a specific data center or financing project.
• Avoid relying on a single outcome: The discussion says investors should not assume that all AI companies will succeed or that the projected growth curve is guaranteed.

Explicit Cautions and Views

The guest’s view is that a downturn sooner could limit the eventual damage, while the host argues that a gradual deflation would be preferable to a disorderly crash. Both warn that a crash could destroy substantial investor value and have prolonged economic consequences. They also stress that AI itself may remain important and that some long-term winners could emerge after a shakeout.

The host advises viewers to limit risk rather than assume the boom cannot fail and says investors may want to avoid having zero exposure to a major technological shift. He mentions Visa and Mastercard as “toll-collector” businesses that may be less dependent on which AI company wins. These are presented as opinions and examples, not individualized investment advice.

No explicit “not financial advice” disclaimer is heard. The episode includes sponsor advertisements for Incogni and Ketone IQ; these are not presented as investment recommendations.

Assets, Companies, Sectors, and Instruments Mentioned

• Companies and platforms: Anthropic, OpenAI, Nvidia, Broadcom, Oracle, Google, IBM, Cisco, Nortel, Lucent, Visa, Mastercard.
• Financial firms and private-credit participants: Blue Owl, BlackRock, Apollo, Blackstone, Morgan Stanley.
• Assets and instruments: Equities, IPOs, private-company shares, corporate debt, private credit, bonds, U.S. Treasuries, SPVs, equipment-backed financing, and chip or data-center collateral.
• Sectors and themes: Artificial intelligence, semiconductors, cloud and data-center infrastructure, energy, open-source AI, private lending, and technology infrastructure.
• Tickers: No ticker symbols are stated in the subtitles.

Presenters and Sources

Presenters: Tom Bilyeu and Jeff Snider.

Sources or commentators cited in the discussion: Reuters; Michael Burry; Ed Zitron; the CEOs of IBM and Cisco; Elon Musk; Bill Ackman; and historical examples involving Lehman Brothers, Nortel, and Lucent.