Anthropic Opened Its Books. What Is It Really Worth?

Anthropic IPO: Financial and Investment Summary

Valuation and Financial Performance

• Anthropic’s prospectus reportedly describes a prospective valuation of $1.5–$2 trillion. The video says expectations had already shifted from $2 trillion to $1.5–$1.75 trillion in a few weeks.
• For the first half of the year, Anthropic reported a net loss of $80.7 billion. According to the video, more than $72 billion came from quarterly fair-value remeasurement of convertible financing agreements, mainly involving Amazon and Google. Because those instruments are treated as liabilities until conversion, Anthropic records a loss when the value of its shares rises. The presenter stresses that this was not $72 billion of cash burned.
• Excluding that revaluation, Anthropic reported about $2.7 billion in adjusted profit for the half year. Its actual operating loss was just under $700 million, compared with $2 billion in the same period a year earlier. Anthropic reached operating profit in Q2 and generated nearly $4 billion in positive operating cash flow; the subtitle does not specify the period for that cash-flow figure.
• Computing costs fell from 251% of sales at the beginning of the previous year to 48% in Q2. The presenter emphasizes that Anthropic did not exclude model-training costs from its reporting.
• The prospectus’s financial data ends on June 30, leaving investors without public Q3 results. Steve Eisman estimated sequential revenue growth at 135% in Q1 and 147% in Q2, then around 45% in Q3, based on figures Anthropic reportedly shared privately. The video notes that 45% quarterly growth would still imply roughly 250% annual growth, but the sharp slowdown could affect IPO sentiment.
• Anthropic reportedly cut the cost of a typical task by 25%. The presenter describes intensifying price competition from Google’s Gemini and cheaper OpenAI models, as well as lower-cost open-weight models from China.

Compute Commitments and Funding Needs

• Anthropic reportedly signed new compute-capacity deals with minimum commitments of $417 billion in roughly the ten months since the end of 2025. Total commitments through 2045 exceed $800 billion.
• The cited commitments include:
  • Amazon: $110 billion take-or-pay commitment.
  • Broadcom: Approximately $161 billion in chip leases.
  • SpaceX: Up to $120.5 billion, or about $3.1 billion per month once capacity is running.
  • Additional capacity from Google, Microsoft, Nscale, and Lambda.
• These are presented as purchase and lease obligations disclosed in the prospectus notes, not conventional debt shown on the balance sheet.
• The presenter argues that Anthropic is moving from a relatively asset-light model—leasing compute—to a capital-intensive business requiring spending on data centers, chips, cooling, and ongoing replacement of obsolete equipment. That could make recent operating profit difficult to sustain and helps explain the company’s need to raise capital.

Growth, Market Demand, and Valuation Uncertainty

• The video describes a short-lived “token maxing” boom in which companies encouraged employees to use AI tools heavily, sometimes tracking usage by tokens. Meta employees reportedly used more than 60 trillion tokens in 30 days, and one heavy user could cost over $1 million. Amazon later cautioned against using token consumption as a productivity measure and shut down an internal leaderboard.
• Software engineers are described as early adopters. A much larger potential market is office workers using AI to process and summarize documents. The presenter’s central uncertainty is whether that market can generate durable revenue at current prices.
• Illustrative extrapolations show why forecasts are fragile: if revenue simply doubled every year, Anthropic would reach $630 billion in revenue by 2031; maintaining half its current growth rate would imply $584 trillion. The presenter treats these as examples of how quickly extrapolations become implausible, not as reliable forecasts.
• The conclusion is that Anthropic appears to be a real business with paying customers and strong growth, but its $1.5–$2 trillion valuation is highly speculative. The price is described as a “guess,” with bullish and bearish cases depending on uncertain growth, monetization, competition, and capital requirements.

Governance and Other Risks

• Anthropic is a public benefit corporation, allowing its board to weigh its stated public-benefit mission alongside shareholder interests.
• The video describes a governance structure in which a long-term-benefit trust nominates four of seven directors, while the seven founders hold a special class of shares with 50.1% of voting power, regardless of how many shares they retain. The founders select the other three directors. The presenter says founder influence over the nomination process and an employee vote in the event of an impasse further limit ordinary shareholders’ control.
• Amazon and Google are described as major investors whose shares have no voting rights. The presenter contrasts this structure with the traditional “one share, one vote” model and notes that markets have often tolerated restricted shareholder rights in high-profile technology companies.
• The prospectus warns of AI-related risks, including self-preservation behavior, manipulation, blackmail-like outputs, assistance to attackers, and catastrophic harm potentially beyond the insurance industry’s capacity to cover.
• The company also flags regulatory and government risk. The video says government contracts account for about 1% of revenue today, but argues that the larger risk is potential state control of powerful AI technology. It recounts the Department of Defense’s supply-chain-risk designation and the company’s unsuccessful appeal, while noting that a court had found earlier government action retaliatory.
• The prospectus reportedly allows insiders to sell shares soon after the IPO. The presenter also notes that Meta was raising $95 billion and Alphabet about $90 billion, framing the broader fundraising environment as a reason to scrutinize who is selling and who is raising capital.

Assets, Companies, and Sectors Mentioned

• AI companies and models: Anthropic, Claude, OpenAI, Google Gemini, Grok.
• Technology and infrastructure companies: Amazon, Alphabet/Google, Meta, Microsoft, Broadcom, SpaceX, Nscale, Lambda, Tesla, Snap.
• Sectors and instruments: Artificial intelligence, cloud computing, semiconductors and chips, data centers, convertible debt or financing instruments, equity shares, chip leases, and take-or-pay compute agreements.
• No stock ticker symbols are stated in the subtitles.

Explicit Investor Takeaway and Disclosures

• The presenter offers no firm buy or sell recommendation. The caution is that investors are effectively betting AI growth and monetization will arrive fast enough to justify the valuation and meet the $800 billion-plus compute commitments.
• The video is sponsored by Revolut. The sponsor segment advertises a $20 sign-up offer and up to 4% annual interest on savings, alongside currency and transfer services. No explicit “not financial advice” disclaimer appears in the subtitles.

Presenters and Sources Mentioned

• Presenter: Patrick Boyle.
• Sources cited or discussed: Matt Levine; Jonathan Veale of The Wall Street Journal; Steve Eisman on Prof G Markets; Paul Kedrosky on Hidden Forces; Robert Armstrong and John Foley of the Financial Times; Michael Burry; Scott McNeely. The video also refers to reporting by Fortune.