Anthropic IPO Valuation: Core Assumptions and Cost of Capital

This article revisits a valuation of Anthropic, an AI company, ahead of its IPO, focusing on the importance of feedback and challenging one's own assumptions, particularly regarding the cost of capital.

The Anthropic Valuation Revisited

The initial valuation of Anthropic, conducted a week prior, generated significant discussion. The valuation process emphasizes an open feedback loop, listening to disagreements, and learning from them without necessarily adopting every suggestion. The goal is to value companies for an audience of one, providing a transparent spreadsheet for others to modify inputs and arrive at their own conclusions.

The initial valuation projected Anthropic's value between $1.45 trillion and $1.5 trillion, which was still lower than the expected IPO pricing range of $1.5 trillion to $2 trillion. This valuation was based on a story where Anthropic becomes an incredibly successful company in a rapidly expanding AI space, achieving:

• Revenues: $1 trillion by the end of the next decade.
• Product Mix: More premium-priced products than mass-market.
• Operating Margin: 32.5% (adjusted to 27.5% after a 5% regulatory drag).
• Reinvestment: $3 of revenues for every dollar of capital, reflecting easier access to compute power.
• Cost of Capital: 9.74%.

Key Details and Assumptions

Several details were highlighted:

• Value per Share: A placeholder, as the share count is not yet finalized. The aggregate equity value remains constant, but per-share value will change. Restricted shares are included.
• Cost of Capital (9.74%): This includes a cost of debt based on a "double B plus" rating, a below-investment-grade rating, as suggested by news reports.
• Compute Commitments: Approximately $450 billion in contractual commitments and several hundred billion more in flexible commitments. The present value of contractual commitments ($373 billion) is included as debt.
• Dilution Risk: Addressed by counting restricted shares, valuing employee options (reducing equity value), and assuming convertible bonds are either captured in their value or converted at IPO. Future capital raises due to negative free cash flows are already incorporated into the discounted cash flow (DCF) valuation.

Challenging the Valuation: Weakest Links

The valuation process involves playing devil's advocate with one's own assumptions, identifying the most fragile storylines.

1. Revenue Assumption: $1 Trillion by End of Next Decade

This is considered the biggest assumption, as it would make Anthropic the largest company in the world by revenue today. This projection relies on several sub-assumptions:

• Total AI Market: Growing to $5-$10 trillion by the 10th year, requiring widespread business adoption, more employee replacement than mere tools, and high customer willingness to pay.
  • Risk: The market might be more tool-centric, with limited adoption or lower pricing.
• Premium Product Market: A significant portion of the AI market will be for premium, frontier models with high prices.
  • Risk: Open-source models could perform similarly, reducing the premium market's size.
• Anthropic's Market Share: Anthropic will be a dominant winner in the premium AI market due to its current lead.
  • Risk: The premium market could become splintered or disrupted, leading to a smaller market share for Anthropic.
• Add-on Markets: Anthropic can build on add-on markets (e.g., "Anthropic drugs"), similar to how Google built self-driving cars.
  • Risk: These new businesses might require skills Anthropic lacks (e.g., logistics for ride-sharing), hindering entry.

Failure of any of these layered assumptions would result in revenues below $1 trillion.

2. Governance Structure

The second biggest assumption is that governance will not hinder the company. Anthropic's story requires management to be adaptable and willing to reverse policies. The concern is that the power concentrated in seven founders, who may share a similar worldview and operate in an echo chamber, could make change difficult, especially when admitting mistakes. This risk might not show up directly in numbers but affects the acceptable price-to-value ratio for investment.

3. Profitability Assumption: 32.5% Operating Margin

Achieving this margin requires several conditions:

• High Gross Margins (80%+): Assumes compute costs shift from variable to fixed, maintaining high gross margins.
  • Risk: Pricing pressures could drive down AI product prices, or input costs could rise, reducing gross margins.
• Economies of Scale: As revenues grow, economies of scale, particularly in compute costs, will kick in.
  • Risk: Compute and employee costs could grow faster than expected, weakening economies of scale.
• Sustainable Margins (Moats): Anthropic can maintain these margins long-term, implying a "moat." Technology moats often involve first-mover advantage and product stickiness (e.g., Microsoft Windows/Office).
  • Risk: There might be no moats, or stickiness might not develop, making profits unsustainable.

If these assumptions break down, the target margin will be significantly lower.

4. Reinvestment Assumption: $3 Revenues per $1 Capital

This assumption for compute power, the primary investment, relies on:

• Locked-in Capacity: Existing compute agreements provide enough capacity for the near future.
• Decreasing Compute Costs: As revenues grow, Anthropic can add compute power at lower costs due to cheaper chips.
  • Risk: Chip costs might decrease, but other AI agent power costs could offset this.
• New Business Reinvestment: Reinvestment needs for new ventures (like drugs) are unknown.
  • Risk: Reinvestment could be higher or lower than anticipated.

If these conditions are not met, Anthropic would need to reinvest more to achieve the same revenues.

5. Risk Factors

Three types of risk were incorporated:

• Cost of Capital (9.74%): Discussed in detail later.
• Failure Risk (10%): Acknowledges Anthropic is a young company with inherent risks, but argues it's not as high as for some other companies (e.g., CoreWeave).
• Regulatory Risk (5% margin reduction): This is the most uncomfortable assumption, as political pressures could make regulations far more onerous.

Navigating Uncertainty in Investment

When faced with significant uncertainty, investors have three main pathways:

1. Avoidance: Conservative investing suggests avoiding companies with high uncertainty.
2. Margin of Safety: A traditional value investing approach where the purchase price must be a certain percentage below the estimated value (e.g., 40% below a $1.5 trillion valuation).
   • Trade-off: Increasing the margin of safety reduces Type 1 errors (buying the wrong stock) but increases Type 2 errors (missing out on good investments). The optimal margin depends on risk aversion and portfolio composition. A high cash position might indicate an excessively high margin of safety.
3. Law of Large Numbers: Spreading risk across a diversified portfolio (e.g., Anthropic as one of 20-25 stocks). This allows for investment in uncertain companies, as individual errors average out across the portfolio.

Addressing Pushback: The Cost of Capital

A significant portion of the feedback indicated that the 9.74% cost of capital was too low. While this is a common focus, it's argued that the cost of capital should not be the biggest driver of valuation.

What is the Cost of Capital?

The cost of capital is the "Swiss army knife of finance," serving as:

• Hurdle Rate: In corporate finance, it determines whether to undertake a project.
• Discount Rate: In valuation, it's used to discount future cash flows.

It's a weighted average of the cost of equity and the cost of debt.

• Cost of Equity: Reflects what an investor can earn on an investment of equivalent risk elsewhere. It's an opportunity cost and can be subjective but must pass reality checks.
• Cost of Debt: The long-term borrowing rate, adjusted for taxes.

Reality Checks for Cost of Equity

1. History:

   • Over the last 25 years, average US equity returns were 8.74%.
   • Over the last 50 years, 11.92%.
   • Over the last 100 years, 10.13%.
   • The equity risk premium (over T-bond rate) over 100 years was 5.54%. With a current T-bond rate of 4.9%, this suggests an average cost of equity around 10.44%.
   • Claiming a 15% cost of equity for an average firm is inconsistent with historical returns, especially given the US market's success.

2. Implied Equity Risk Premium:

   • Calculated by backing out the expected return from the current S&P 500 level and expected cash flows.
   • As of October 2026, the implied expected return for US equities was 8.99% (8.83% in November).
   • Historically, US equities have not delivered an expected return greater than 10% in 32 years, except for periods in the late 1970s and early 1980s when T-bond rates were very high.

3. Consistency with PE Ratios:

   • Using a dividend discount model, a 15% cost of equity, a 5% nominal growth rate, and a 20% return on equity would imply a market P/E ratio of 7.27.
   • Given current P/E ratios closer to 20, a 15% cost of equity is inconsistent with investing in the current market.

Cost of Debt

The cost of debt is the rate at which a company can borrow long-term today. For Anthropic, a "double BB plus" rating suggests a pre-tax cost of debt around 6.5%. High costs of debt (10-20%) often reflect failure risk, which should be separated and accounted for explicitly (e.g., through a failure risk percentage) rather than solely inflating the discount rate.

Cost of Capital Ranges

A histogram of cost of capital for US companies shows tight ranges:

• Energy Companies: 80% have a cost of capital between 7% and 11%.
• Technology Companies: 80% have a cost of capital between 9% and 12.3%.

This suggests that finessing the cost of capital within these narrow ranges has limited impact on valuation precision. While Anthropic might be riskier, placing it at the 90th percentile (12% cost of capital) requires adjusting the failure rate to avoid double-counting risk. Additionally, the cost of capital should change over time as the company matures and its risk profile evolves.

Consequences of Misestimating Hurdle Rates

Setting an arbitrarily high hurdle rate, especially for institutional investors and publicly traded companies, has significant consequences:

• Professional Investors (VCs, Private Equity):
  • Often claim target returns of 25-50%, but actual returns are often closer to or below market averages (S&P 500/NASDAQ).
  • This "leakage" comes from inflating discount rates with failure risk.
  • Overreaching to meet high promised returns can lead to risky investment decisions.
• Publicly Traded Companies:
  • CFOs often set hurdle rates 2-3% higher than market-justified rates, believing it's "conservative."
  • Consequence 1: Underinvestment: Too-high hurdle rates turn away good projects, leading to underinvestment and loss of market share, especially as companies mature.
  • Consequence 2: Gaming the Numbers: Managers inflate projected cash flows to meet high hurdle rates, leading to "made-up cash flows discounted at a made-up discount rate," rendering NPV calculations meaningless.
• Individual Investors:
  • Demanding a 20% cost of equity means no S&P 500 company would be undervalued, and only "basket case" companies would appear attractive.
  • This leads to holding excessive cash or becoming vulnerable to scams promising unrealistic returns.
• Privately Owned Businesses:
  • High hurdle rates (18-25%) lead to underinvestment, loss of market share, and undervaluation of the business when it's time to sell.
  • Appraisers often contribute to this by using "build-up approaches" that add thinly sourced premiums (e.g., small-cap premium, liquidity premium, company-specific risk premium) to inflate the cost of capital, effectively giving a "gift" to public companies that acquire them.

Conclusion

The Anthropic valuation is not a defense but an opportunity to discuss valuation principles. The key assumption remains revenue growth and the broader AI market's evolution. While individual investors can set their own hurdle rates, it's crucial to understand the consequences. The cost of equity must pass reality checks: historical returns (9-10%), implied expected returns (not in double digits since 1994), and consistency with market P/E ratios. Arbitrarily high rates lead to poor investment decisions and missed opportunities.