Aswath Damodaran Values Anthropic’s IPO at Nearly $1.5 Trillion. Here’s Why I Think His Base Case Is Too Optimistic

I began reading Aswath Damodaran’s Anthropic valuation with one question: how could a company facing such enormous uncertainty deserve a discount rate of just 9.74%?

By the time I finished his argument, my disagreement had become broader. His estimate of around $1.48 trillion rests on a business generating around $1 trillion of annual revenue within a decade, earning a 27.5% operating margin and continuing to earn attractive returns on new investment long afterwards.

None of those outcomes is impossible. My concern is how much confidence I would place in all of them arriving together.

Mr. Damodaran makes his reasoning clear and invites readers to choose different assumptions. I have taken that invitation seriously. I want to explain his valuation first, then show where I would make different judgments and what those changes do to the answer.

Under my alternative operating assumptions, I arrive at around $583 billion of equity value, using his treatment of cash, financing obligations and assumed IPO proceeds.

That comparison includes a hypothetical $100 billion capital raise in both cases. It is an estimate under stated assumptions, not a prediction of the IPO price. The useful question is why the two sets of assumptions lead to such different answers.

First, What Does the $1.48 Trillion Valuation Mean?

It is not an announced IPO valuation. It is Professor Damodaran’s estimate of what the equity could be worth under his assumptions.

The preliminary S-1 leaves the final offer price and share count blank. As of this update, the final IPO terms remain unconfirmed. Mr. Damodaran’s assumed $100 billion of IPO proceeds is a valuation input, rather than an amount Anthropic has confirmed it will raise.

For the wider filing, I have explained the business and its financial statements in our Anthropic IPO S-1 breakdown. Here, I am asking a narrower question: what future would justify paying this much for the business?

Mr. Damodaran uses a discounted cash flow valuation, or DCF.

Imagine buying a small shop. You would want to know how much customers will spend, how much the shop will keep after expenses, and how much it must put back into stock, equipment and expansion. You would also care whether those future earnings are dependable.

A DCF asks those questions year by year. It then converts the cash expected in the future into an equivalent amount today.

For Anthropic, the sequence is:

1. Estimate revenue.
2. Subtract operating costs and taxes.
3. Subtract the investment needed to support the business and its growth.
4. Discount the remaining cash to today.
5. Adjust for failure, financial claims and cash to estimate equity value.

The math gives a precise answer to an uncertain question. That is why understanding the assumptions is more useful than memorising the final number.

How Aswath Damodaran Builds a Trillion-Dollar Revenue Business

The business he is valuing extends well beyond a chatbot

Mr. Damodaran starts with the work AI could perform for customers, rather than the number of people opening a chatbot.

A consumer paying for a chatbot is one source of revenue. A company using Claude to write software, process documents and execute recurring workflows creates another. If AI performs work that previously required expensive human effort, the potential spending pool could become much larger than today’s chatbot subscriptions.

His story places Anthropic towards the premium end of that market: customers paying for capability, reliability and useful work. It also allows for opportunities beyond selling access to models, including businesses built around what those models can accomplish. I treat those opportunities as part of his investment thesis, not as revenue Anthropic has already secured.

The underlying market opportunity is a forecast. A company does not automatically collect the economic value its technology creates. Customers, cloud providers, application developers and competitors can capture parts of it too.

These revenue streams are explained in greater detail in our Anthropic business model breakdown, including how subscriptions and usage-based payments work.

The revenue growth his valuation requires

Mr. Damodaran’s forecast gets Anthropic to more than $100 billion of revenue in the first forecast year. Revenue then grows by 42% annually through year five, before the pace gradually slows.

Source: Aswath Damodaran’s Anthropic valuation; author’s calculations. Forecast years are model periods, not reported financial results.

From years six to ten, annual growth steps down from 42% towards 4.9%. By the tenth year, roughly 2036 in the investment story, Anthropic is generating just over $1 trillion of annual revenue.

How much of that revenue becomes profit?

His profit forecast separates the market into two segments, each with its own assumed margin.

Source: Aswath Damodaran’s Anthropic valuation. These are his assumptions, not disclosed segment results.

The math is straightforward:

(75% × 40%) + (25% × 10%) − 5 percentage points = 27.5%.

For every $100 of revenue in the mature business, the model therefore leaves $27.50 of operating profit before tax.

The regulatory adjustment is a reduction of five percentage points, not a 5% reduction in the original margin.

The model starts with a 5% operating margin in year one and reaches 27.5% in year five. The actual intermediate margins are 14%, 18.5% and 23%.

By year ten, $1.011 trillion of revenue multiplied by 27.5% produces approximately $278 billion of annual operating profit.

Profit is not yet cash that investors can take home

This is where the shop analogy becomes useful again.

Suppose the shop earns a profit but needs a larger refrigerator, more inventory and another location. The money spent on those things cannot also be paid to its owners.

Anthropic has a much more expensive version of the same problem. Growth requires compute, research, infrastructure and working capital. Some investment maintains the business; some expands it.

What a sales-to-capital ratio of three means

Mr. Damodaran uses a sales-to-capital ratio of 3 for forecast reinvestment.

In this model, each additional $3 of annual revenue requires approximately $1 of additional net investment.

If revenue expands by $60 billion, the implied net reinvestment is $20 billion. A lower ratio means the business needs more investment for the same additional revenue.

This is a shortcut for estimating net reinvestment. It is not a claim that chips cost one-third of sales, or that every dollar of existing revenue needs another 33 cents of investment every year. Replacement spending, depreciation and research treatment still have to be consistent with the profit and reinvestment forecasts.

Mr. Damodaran also assumes investment precedes the revenue it supports by one year. The first forecast-year reinvestment is calculated from the increase between year-one and year-two revenue:

($146.32 billion − $103.04 billion) ÷ 3 = $14.43 billion.

Year-one operating profit is only $5.15 billion. With no modelled tax in that year, the resulting free cash flow is approximately negative $9.27 billion.

Rapid growth can consume cash even when the income statement shows a profit.

This timing assumes the capacity supporting the initial leap in revenue has already been funded or captured elsewhere in the valuation. Moving investment back one year does not make its cost disappear.

Why he adjusts research and compute commitments

Mr. Damodaran treats research spending as an investment with a multi-year life, rather than assuming every research dollar benefits only the year in which it is spent.

The idea is reasonable. If a training programme produces models that earn revenue over several years, charging its entire cost against one year can distort comparisons with a business that buys a physical asset and depreciates it over time.

His research adjustment uses a three-year life. Past expenditure is built into a research asset, and a portion is amortised each year.

He also capitalises a schedule of compute and lease commitments. That means estimating their value today and treating them as debt-like obligations, rather than ignoring them because they are not all conventional borrowings on the balance sheet.

The model discounts around $454.43 billion of scheduled payments into a present value of around $363.02 billion.

Both adjustments recognise something real: frontier AI needs substantial investment. Capitalising a cost spreads its economic effect across the periods that benefit from it. The cost still has to be paid, so profit and cash flow can look very different during the build-out.

How taxes enter his forecast

To estimate the cash generated by Anthropic’s operating business, Mr. Damodaran first calculates operating profit after an assumed tax charge. He uses the following tax rates for this valuation calculation. These are assumptions for estimating operating cash flows, rather than a forecast of Anthropic’s actual tax payments.

Source: Aswath Damodaran’s Anthropic valuation. These are forecast effective rates, not statutory tax rates.

Past losses, deductions and credits can delay cash taxes. Their eventual effect depends on tax rules, the timing of investment and where income is earned. For this comparison, I retain Mr. Damodaran’s tax timetable; my disagreement will focus on the operating business and its long-term economics.

How Future Cash Becomes $1.48 Trillion Today

What WACC actually means

WACC, or weighted average cost of capital, combines the return required by shareholders and lenders, weighted by their shares of the business’s financing. Debt is generally measured after its assumed tax benefit.

It is used here because the cash flow being valued belongs to the operating business before payments to debt and equity providers.

The practical idea is simple. A dollar received next year is worth less today because I must wait, and because receiving it is uncertain. At a 10% discount rate, $110 expected one year from now has a present value of $100.

A higher discount rate lowers the present value of the same projected cash flows. It is not an extra operating expense and it is not the company’s borrowing rate.

Mr. Damodaran’s forecast uses 9.74% for the first five years. It then gradually falls to 9.13% in year ten, with 9.13% used for the mature business thereafter.

So there are two relevant rates. Changing the first one while leaving the second untouched does not fully test the model’s long-term risk assumption.

The business continues after year ten

A ten-year forecast does not imply Anthropic closes down in the eleventh year.

The model adds a terminal value for the cash generated beyond the explicit forecast. This is the largest part of the calculation.

It assumes revenue and cash flow eventually grow at 4.9% a year indefinitely. That rate follows the model’s risk-free-rate input. It is nominal dollar growth, meaning it includes inflation.

It also assumes a mature return on invested capital of 15%.

Return on invested capital, or ROIC, measures operating profit after tax relative to the capital invested in the business. It helps answer another question: how much profit must Anthropic reinvest to keep growing?

The model uses:

Required reinvestment share = long-term growth ÷ long-term return on capital.

At 4.9% growth and a 15% return on capital, Anthropic must reinvest approximately 32.7% of its after-tax operating profit indefinitely. The remainder becomes free cash flow.

Here is the terminal calculation:

Source: Aswath Damodaran’s Anthropic valuation; author’s calculations. Rounded figures may not sum exactly.

The $3.48 trillion amount is a value at the end of the forecast, not a value today. Discounting it back produces approximately $1.40 trillion today.

The present value of the first ten years’ cash flows contributes another $424.66 billion.

Consequently, approximately 77% of the operating value before the failure adjustment comes from cash flows beyond year ten.

That is not automatically a flaw. A business expected to operate for decades should have substantial value beyond its first ten years. But it tells me where the valuation needs the most scrutiny: the durability of the mature business.

The final bridge from business value to equity value

Mr. Damodaran separately assumes a 10% probability of failure, with 50% recovery of the estimated operating-business value in that outcome.

The math retains 95% of the calculated operating value:

90% × full value + 10% × half value = 95% of full value.

It is a 5% reduction in operating value. It is not a 10% reduction in equity value.

Source: Aswath Damodaran’s Anthropic valuation; author’s calculations. The $100 billion proceeds are hypothetical. Equity values shown are before allocating employee option claims.

This gives the whole-company equity estimate before allocating employee option claims. Converting a total equity estimate into value for each public share requires the relevant share classes, conversions and employee awards to be accounted for consistently.

This is the complete route to the headline. It is not simply a trillion dollars of future revenue multiplied by a convenient sales multiple.

Why I Would Use a Higher Cost of Capital

Mr. Damodaran’s case deserves to be stated fairly. He does not choose 9.74% because he thinks Anthropic has no risk. He views it through the risk of large technology businesses, with an additional burden from its compute commitments, and allows separately for financing failure and regulatory costs.

He also argues that discount rates must relate to the returns available in public markets. I agree. A rate chosen simply because a company feels exciting or dangerous would be difficult to defend.

My disagreement is about how much confidence I would place in Anthropic having the economic resilience of an established technology business.

Usage can be sticky without being recession-proof

Embedding Claude inside a customer’s workflow can make it harder to replace. That is a reason to expect recurring demand.

But a customer can keep using Claude while doing less work with it. If it delays software projects, reduces experimentation or runs fewer AI tasks, usage-based spending can slow without the customer leaving.

Anthropic’s S-1 explicitly describes variability in customer usage. It says total customer spending in a two-week holiday period fell approximately 20% against the preceding two weeks. That is evidence that activity affects spending; it does not establish what would happen in a recession.

The unanswered question is whether spending would prove resilient because Claude saves money, or cyclical because businesses cut projects and activity. Both effects are possible. I would want a longer record before giving the first explanation most of the weight.

The commitments to computing capacity make that question more consequential. Capacity can be secured before the revenue arrives, and those obligations do not necessarily shrink when usage slows.

I reflect the financing component through the same obligation treatment as Mr. Damodaran. My higher discount rate concerns the sensitivity of the operating business to broad economic conditions; it is not a second deduction of the compute bill.

Why I use 11.5% initially

I use 11.5% for the first five years, gradually declining to 9.5% by year ten. To explain that choice, I need to show what return shareholders might require, what borrowing costs, and how the two fit together.

I retain Mr. Damodaran’s borrowing-cost and financing assumptions for this comparison. The change I make is to the risk assumed for shareholders.

The cost of equity starts with a risk-free return, then adds compensation for taking market risk. Using a 4.9% risk-free rate, an approximately 5% equity risk premium and an assumed equity beta of 1.7 gives:

4.9% + 1.7 × 5% = 13.4% cost of equity.

Beta measures sensitivity to broad market movements. My 1.7 assumption reflects the greater economic sensitivity I would allow for a young business whose revenue depends heavily on customer usage. It is a judgment for this scenario, rather than a measured Anthropic share-price beta.

Next comes the cost of borrowing.

Mr. Damodaran’s accompanying cost-of-capital calculation uses a borrowing cost of around 6.28% before tax. That consists of a 4.9% risk-free rate plus approximately 1.38 percentage points of additional compensation for credit risk.

Interest deductions can reduce a company’s tax bill. Assuming a 25% tax benefit, the estimated borrowing cost becomes:

6.28% × (1 − 25%) = approximately 4.71% after tax.

That is where the 4.7% comes from. The calculation assumes the interest deduction can be used; it does not mean Anthropic currently pays an effective tax rate of 25%.

Finally, I need to decide how much weight to give each funding source.

Mr. Damodaran’s supporting calculation uses approximately $1.61 trillion of equity value and $454 billion of debt-like value, including capitalised lease and compute obligations.

Together, those amounts represent approximately $2.07 trillion of capital:

Source: Aswath Damodaran’s accompanying valuation calculations. Figures are rounded.

These are provisional inputs used to establish financing weights. The $1.61 trillion equity figure is separate from his final $1.48 trillion DCF estimate, and the weights do not represent a confirmed IPO financing split. I retain them to isolate the effect of my higher assumed equity risk.

Combining the two required returns gives:

78% × 13.4% + 22% × 4.7% = approximately 11.5% WACC.

This makes 11.5% a transparent alternative scenario. It does not establish that Anthropic’s true cost of capital is precisely 11.5%.

As the business matures, I allow the rate to decline to 9.5%. That assumes greater stability over time, while retaining a modest premium over Mr. Damodaran’s 9.13% mature rate.

I agree with Mr. Damodaran to avoid charging for the same risk repeatedly. Lower pricing belongs in revenue or margins, higher infrastructure spending belongs in reinvestment, and financing failure belongs in a failure scenario. The higher WACC reflects my judgment about exposure to broad economic conditions.

A higher WACC alone still leaves a trillion-dollar valuation

Source: Aswath Damodaran’s valuation assumptions and author’s recalculations. All other assumptions are held constant. Values are before allocating employee option claims.

My two rate changes reduce equity value by approximately $288 billion, or about 19.5%. That is substantial. It also leaves the valuation above $1 trillion.

This supports part of Mr. Damodaran’s argument: revenue, profitability and reinvestment deserve at least as much attention as the discount rate. My lower final estimate therefore needs a broader business case.

The Other Assumptions I Would Change

1. I would start with $750 billion of year-ten revenue

I can see a route to Anthropic becoming an exceptionally large business. The filing shows substantial enterprise adoption and expanding usage. Its approximately 900% net revenue retention at September 30 is striking evidence of spending expansion in the defined customer cohort.

But that figure compares the cohort’s revenue across two quarters a year apart. It is not a promise that the same customers multiply their spending ninefold every year.

I am also cautious about converting labour savings into supplier revenue. If Claude helps a company save $10 million, Anthropic does not necessarily collect $10 million. The customer needs a reason to adopt it, competitors can offer alternatives, and other businesses in the delivery chain need to earn a return.

I therefore use $750 billion of revenue in year ten as my central scenario, rather than approximately $1 trillion.

There is no filing that proves $750 billion is correct. It is a deliberate reduction in the market share and monetisation I am prepared to assume, while retaining substantial success.

I retain his first forecast-year revenue of $103.04 billion. Getting from there to $750 billion still requires revenue to grow more than sevenfold over the following nine years.

To reach that target, growth in years two to five becomes approximately 35.86%, before declining towards my 4% mature growth rate. I am allowing a very successful business, while assuming less long-term scale than Mr. Damodaran.

2. I would use a 25% mature operating margin

A premium product can deserve a premium price. But maintaining that position in frontier AI requires repeated investment.

The S-1 says models released within the previous six months generated more than 80% of revenue in the September quarter. That supports the case for rapid innovation. It also shows how actively the company must refresh what customers buy.

The filing describes reductions in customer cost per task from newer models, while saying efficiency improvements preserved margins. That is evidence against assuming every price reduction destroys profitability. It does not establish that the same balance will hold for a decade.

My 25% operating margin allows Anthropic to remain a highly profitable business. It gives somewhat less credit to sustained premium pricing after research, distribution and the continuing cost of competing.

This is an all-in forecast margin under the same broad accounting framework. I do not deduct another five percentage points for regulation after it. That would count the concern twice.

3. I would require more investment for each extra dollar of sales

I reduce the sales-to-capital ratio from 3 to 2.5.

For an additional $100 of annual revenue, that increases modelled net reinvestment from approximately $33.33 to $40. It is 20% more investment for the same additional revenue.

My reason is the combination of rapid model replacement and the commitment to build and replenish computing capacity. Anthropic may become far more efficient, but I would not assume those gains translate fully into lighter investment needs. More capable products can create more demanding workloads.

I preserve the one-year investment lead in the original model. I also retain Mr. Damodaran’s existing capitalised compute amount for the comparison. The 2.5 ratio describes additional net investment in the forecast; it is not another deduction of every contract already included in the opening obligations.

This is a cautious efficiency assumption, not a directly observed long-run ratio. A full schedule of asset lives, replacement costs and contract payments would be better than either of our shortcuts.

4. I would reduce perpetual growth to 4%

The original 4.9% rate is not inherently impossible. It is a nominal rate in a dollar valuation, and there is a rationale for tying long-term growth to a nominal risk-free rate.

My concern is how much favourable economics is already assumed to survive indefinitely.

I use 4% nominal perpetual growth to put less weight on expansion beyond the explicit forecast. It still allows the mature company to grow every year.

This change also affects the growth slowdown in years six to ten. I do not change the terminal denominator while pretending the surrounding forecast remains completely untouched.

5. I would use a 12% mature return on capital

This is one of the least visible assumptions and one of the most consequential.

Mr. Damodaran’s 15% mature return on capital exceeds his 9.13% mature cost of capital by almost six percentage points. It assumes Anthropic continues creating substantial economic value from additional investment indefinitely.

That could happen. But permanent excess returns need durable advantages in customer relationships, technology, distribution or cost. Fast adoption is evidence of current strength; it does not settle the durability question.

I use 12%, against my mature cost of capital of 9.5%. Anthropic still earns more than its cost of capital. I am allowing it a lasting advantage, just a smaller one.

At my 4% growth rate, the mature business reinvests 4% ÷ 12% = one-third of after-tax operating profit. The model therefore pays for the growth I am assuming.

What I leave unchanged

I retain Mr. Damodaran’s 10% failure probability and 50% recovery assumption. I do not have enough evidence to claim that 20% or 30% would be a better estimate.

I also retain the 4.9% risk-free-rate input for a controlled comparison. Replacing one market input with a newer number would require reconsidering the associated risk premiums and debt costs, not simply choosing whichever number lowers value.

I also retain his cash, financing and tax treatment. That makes the resulting valuation gap a comparison of business assumptions. I am not claiming that different classifications of those items explain why his estimate is too high.

My Revised Valuation, With the Math Visible

Here are the assumptions I use together. The right-hand column contains my scenario choices, not facts claimed by Anthropic.

Source: Aswath Damodaran’s valuation assumptions and author’s alternative assumptions and calculations. Both cases use the same investment timing and margin-convergence structure.

Here is the terminal calculation, one step at a time.

First, year-ten revenue of $750 billion grows by 4% to $780 billion in year eleven.

At a 25% operating margin, that produces $195 billion of operating profit.

After 25% tax, it leaves $146.25 billion.

To grow at 4% while earning 12% on new capital, the business reinvests one-third of that amount, or $48.75 billion.

That leaves $97.50 billion of mature-year free cash flow.

The terminal value is then:

$97.50 billion ÷ (9.5% − 4%) = $1,772.73 billion at the end of year ten.

Discounting it to today gives approximately $630.11 billion. The first ten years’ cash flows contribute another $252.84 billion.

Applying the same 95% survival/recovery factor gives:

($630.11 billion + $252.84 billion) × 95% = $838.81 billion.

That is my estimated value of the operating business under these assumptions.

From the operating business to equity value

I then use the same bridge from operating value to equity that Mr. Damodaran uses. This retains his capitalised obligations, cash and other non-operating assets, together with the assumed new capital.

Source: Aswath Damodaran’s cash and financing assumptions; author’s alternative operating assumptions and calculations. Figures are rounded.

My central estimate is therefore approximately $583 billion, or about $580 billion when rounded to avoid suggesting more certainty than the forecast deserves.

The $100 billion capital raise is hypothetical in both cases. Holding everything else constant, excluding that new cash would reduce my estimate to approximately $483 billion. A different funding plan could also change the business’s ability to invest, so that subtraction alone is not a complete alternative financing scenario.

New capital also comes with new owners. A $100 billion raise does not create $100 billion of free wealth for the existing shareholders.

These are estimates of total equity before employee-option allocation, rather than a target price for a public share. The final capital structure and IPO terms would be needed to translate them into a consistent per-share comparison.

How much does my answer depend on the revenue assumption?

Quite a lot. Holding my other scenario assumptions and equity-bridge conventions unchanged:

Source: Author’s calculations using alternative revenue targets and otherwise unchanged assumptions from my central case. Growth rates are recalibrated to each target. These are scenarios, not a statistical confidence interval.

Even retaining a trillion dollars of year-ten revenue does not recover his $1.48 trillion result under my other assumptions.

Equally, someone who believes Anthropic can sustain higher margins, earn more on new investment or require less capital can obtain a materially higher value without making an arithmetic mistake.

My own model still gets approximately 71% of its pre-failure operating value from the terminal period. A lower answer has not made the distant future more observable. It has changed what I am willing to assume about it.

What I Agree With, and What Would Change My Mind

I agree with Mr. Damodaran’s decision to value the business through cash flows. Consumer popularity, a spectacular revenue run rate or a large private funding round cannot answer how much cash Anthropic will eventually generate for its capital providers.

I also agree that compute commitments belong in the analysis, research can have a multi-year economic life, and the company should become less risky if it matures successfully. Those are useful foundations.

Where I disagree is the amount of favourable economics I would put into the central case before the evidence is stronger.

Governance deserves attention as well. The proposed structure described in the S-1 gives the founders 50.1% of general voting power through a special class, while the Long-Term Benefit Trust has the right to appoint a board majority. Public-benefit commitments and possible future share contributions can affect what public shareholders receive. I would track their cash-flow and dilution consequences explicitly rather than inventing an unexplained governance discount.

Lock-up expiries may affect the supply of shares available for trading. They do not, by themselves, alter the cash generated by Claude. I would keep that trading question separate from this DCF.

The evidence that would move my valuation upwards is practical:

• Established customer cohorts continuing to expand spending after the initial adoption surge.
• Revenue growth translating into cash after research, infrastructure and the full cost of compute commitments.
• Strong margins surviving competition and successive model launches.
• A clear contract schedule showing less capital required for each additional dollar of revenue.
• A completed capitalisation table showing exactly how the business’s value is divided.

I would move the valuation down if spending expansion weakens, replenishing compute proves more expensive, or a greater share of the economic benefit flows to customers and suppliers rather than Anthropic.

Mr. Damodaran explicitly leaves room for readers to change his assumptions. I take that invitation seriously. Respecting the framework means being willing to challenge the inputs, including my own.

My criticism is about the central business forecast. Mr. Damodaran’s assumptions describe a company that reaches enormous scale, retains attractive margins and earns strong returns on new capital indefinitely. I would give less weight to that combination until its durability is better established.

I am prepared to believe Anthropic can become an extraordinary business. I am not yet prepared to make all of those outcomes my base case at nearly $1.5 trillion.