Ordering a company valuation also supports the project’s development.
Help develop World Market Atlas — an independent project for individual investors. Your support helps us produce analyses and improve our tools.
Author’s explanatory illustration; boxes do not represent revenue shares. Cloud partners are suppliers and distribution channels, not free capacity.
Subscriptions for individuals and teams; enterprise access and workflows.
Tools for coding and knowledge work; value depends on usage, adoption and retention.
Consumption-based model access, including partner distribution. Revenue is not the same as token volume.
Reported USD billion. Full-year figures and half-year figures are displayed in separate panels. H1 2026 is unaudited.
| Period | Revenue | Operating income · EBIT | Net loss | Operating cash flow · CFO |
|---|---|---|---|---|
| FY2023 | 0.041 | -0.551 | -1.531 | -1.627 |
| FY2024 | 0.386 | -2.980 | -8.309 | -5.692 |
| FY2025 | 4.585 | -8.060 | -41.965 | -2.637 |
| H1_2025 | 1.163 | -2.135 | -12.969 | -1.513 |
| H1_2026 | 16.286 | -0.693 | -80.748 | 3.929 |
H1 2026. Derived bridge from operating income to net loss. Financing remeasurements are noncash and may recur before conversion.
Author scenarios, not company guidance or consensus. Financial anchor: 30 June 2026; information checked: 31 October 2026. All forecasts are nominal USD.
H2 2026 is a six-month stub; subsequent periods are full years. Revenue and FCFF share one linear scale. Forecasts are assumptions.
Costs are shown as negatives; capex includes opening capital-creditor settlement once. Tax remains a cash expense.
| Metric | H2 2026E | 2027E | 2028E | 2029E | 2030E | 2031E | 2032E | 2033E | 2034E | 2035E | 2036E | 2037E | 2038E | 2039E | 2040E | 2041E |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Revenue | 40,000 | 90,000 | 121,237 | 163,316 | 220,000 | 247,942 | 279,433 | 314,923 | 354,922 | 400,000 | 427,965 | 457,886 | 489,898 | 524,148 | 560,793 | 600,000 |
| Compute expense | -23,200 | -54,000 | -68,701 | -87,102 | -110,000 | -120,004 | -130,775 | -142,345 | -154,746 | -168,000 | -176,892 | -186,207 | -195,959 | -206,165 | -216,840 | -228,000 |
| Personnel expense | -8,000 | -15,300 | -19,398 | -24,497 | -30,800 | -33,720 | -36,885 | -40,310 | -44,010 | -48,000 | -49,929 | -51,894 | -53,889 | -55,909 | -57,949 | -60,000 |
| Sales, marketing and partnerships | -7,200 | -16,200 | -21,014 | -27,219 | -35,200 | -38,679 | -42,474 | -46,609 | -51,109 | -56,000 | -58,489 | -61,051 | -63,687 | -66,392 | -69,164 | -72,000 |
| Other expense | -2,800 | -6,300 | -8,082 | -10,343 | -13,200 | -14,381 | -15,648 | -17,006 | -18,456 | -20,000 | -21,042 | -22,131 | -23,270 | -24,460 | -25,703 | -27,000 |
| Depreciation − | -160 | -857 | -1,912 | -3,393 | -5,752 | -8,844 | -11,736 | -14,526 | -17,182 | -19,412 | -21,323 | -23,157 | -24,895 | -26,515 | -27,992 | -29,411 |
| Operating income · EBIT | -1,360 | -2,657 | 2,130 | 10,761 | 25,048 | 32,314 | 41,916 | 54,127 | 69,418 | 88,588 | 100,291 | 113,446 | 128,198 | 144,707 | 163,145 | 183,589 |
| Operating cash tax | -0 | -0 | -106 | -2,112 | -6,262 | -8,079 | -10,479 | -13,532 | -17,355 | -22,147 | -25,073 | -28,361 | -32,050 | -36,177 | -40,786 | -45,897 |
| Depreciation + | 160 | 857 | 1,912 | 3,393 | 5,752 | 8,844 | 11,736 | 14,526 | 17,182 | 19,412 | 21,323 | 23,157 | 24,895 | 26,515 | 27,992 | 29,411 |
| Cash PP&E investment | -1,348 | -4,500 | -7,274 | -11,432 | -17,600 | -19,339 | -21,237 | -23,304 | -25,554 | -28,000 | -29,244 | -30,526 | -31,843 | -33,196 | -34,582 | -36,000 |
| Change in working capital | -1,423 | -300 | -937 | -1,262 | -1,701 | -838 | -945 | -1,065 | -1,200 | -1,352 | -839 | -898 | -960 | -1,028 | -1,099 | -1,176 |
| Legacy settlement cash payments | -1,200 | -0 | -0 | -0 | -0 | -0 | -0 | -0 | -0 | -0 | -0 | -0 | -0 | -0 | -0 | -0 |
| Free cash flow to the firm · FCFF | -5,171 | -6,600 | -4,277 | -652 | 5,238 | 12,902 | 20,991 | 30,752 | 42,492 | 56,501 | 66,458 | 76,818 | 88,240 | 100,821 | 114,669 | 129,927 |
| Discount factor | 0.943 | 0.839 | 0.746 | 0.663 | 0.590 | 0.525 | 0.467 | 0.415 | 0.369 | 0.328 | 0.292 | 0.260 | 0.231 | 0.205 | 0.183 | 0.162 |
| Present value of FCFF | -4,876 | -5,536 | -3,190 | -433 | 3,090 | 6,770 | 9,796 | 12,764 | 15,685 | 18,548 | 19,403 | 19,947 | 20,378 | 20,707 | 20,946 | 21,107 |
| Metric | H2 2026E | 2027E | 2028E | 2029E | 2030E | 2031E | 2032E | 2033E | 2034E | 2035E | 2036E | 2037E | 2038E | 2039E | 2040E | 2041E |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Compute expense / Revenue | 58.0% | 60.0% | 56.7% | 53.3% | 50.0% | 48.4% | 46.8% | 45.2% | 43.6% | 42.0% | 41.3% | 40.7% | 40.0% | 39.3% | 38.7% | 38.0% |
| Personnel expense / Revenue | 20.0% | 17.0% | 16.0% | 15.0% | 14.0% | 13.6% | 13.2% | 12.8% | 12.4% | 12.0% | 11.7% | 11.3% | 11.0% | 10.7% | 10.3% | 10.0% |
| Sales, marketing and partnerships / Revenue | 18.0% | 18.0% | 17.3% | 16.7% | 16.0% | 15.6% | 15.2% | 14.8% | 14.4% | 14.0% | 13.7% | 13.3% | 13.0% | 12.7% | 12.3% | 12.0% |
| Other expense / Revenue | 7.0% | 7.0% | 6.7% | 6.3% | 6.0% | 5.8% | 5.6% | 5.4% | 5.2% | 5.0% | 4.9% | 4.8% | 4.8% | 4.7% | 4.6% | 4.5% |
| Economic stock-compensation cost / Revenue | 12.0% | 10.0% | 9.0% | 8.0% | 7.0% | 6.6% | 6.2% | 5.8% | 5.4% | 5.0% | 4.8% | 4.7% | 4.5% | 4.3% | 4.2% | 4.0% |
| Cash PP&E investment / Revenue | 3.0% | 5.0% | 6.0% | 7.0% | 8.0% | 7.8% | 7.6% | 7.4% | 7.2% | 7.0% | 6.8% | 6.7% | 6.5% | 6.3% | 6.2% | 6.0% |
Assumes economic conversion of notes and preferred shares, without deducting them again as debt. Actual Amazon conversion may be deferred.
No final IPO share count is disclosed. Leave blank for no per-share output; entering a count is your analytical assumption. It must include the economic effects of preference/note conversions and dilution.
Final IPO share value is not available without a reliable denominator.
Same operating scenario; rows are WACC and columns are g. This is not sensitivity of an established market share price.
| WACC / g | 2.50% | 3.00% | 3.50% |
|---|---|---|---|
| 11.44% | 558.8 | 564.0 | 569.7 |
| 11.94% | 520.3 | 524.2 | 528.5 |
| 12.44% | 486.3 | 489.2 | 492.4 |
| 12.94% | 456.1 | 458.2 | 460.5 |
| 13.44% | 429.1 | 430.7 | 432.3 |
Missing IPO information matters The model does not invent an offer price, IPO proceeds, current trading quote or definitive fully diluted share count. Funding rounds and run-rate revenue are separately dated reference points.
Illustration of the valuation process. The preliminary prospectus leaves final IPO pricing and capitalization fields blank.
Market commentary
My perspective on markets, the economy and events that matter to investors.
Anthropic before the IPO: a DCF and the share price the prospectus cannot yet establish
Information cutoff: 31 October 2026. Preliminary S-1: 26 October; financial anchor: 30 June. Annual data are audited, interim data unaudited. Final IPO pricing and capitalization are unavailable. DCF forecasts and parameters are author assumptions; historical share illustrations are not the offering denominator.
Anthropic can be valued using DCF, but the preliminary prospectus does not yet support one definitive post-IPO share price. The S-1 filed on 26 October 2026 reveals a rapidly growing business, substantial financial resources and complex capital claims. Offer pricing, issuance size and final post-conversion capitalization fields remain blank. That limitation belongs beside the output, rather than hidden in a footnote.
Our base case produces a $489.2 billion conditional equity pool under assumed full conversion. This is an author forecast, not an issuer-confirmed value. We also show downside and upside cases, capital-cost sensitivity and a conditional per-share calculator. A unit value appears only after an explicit share-count assumption is entered or a clearly labeled historical illustration is selected.
1. What is known about the offering, and what is missing
Anthropic, PBC publicly filed its S-1 on 26 October. It intends to seek a Nasdaq Global Select Market listing as ANT. The document remains preliminary and does not establish that trading has started. The SEC filing list checked on 31 October showed no later public S-1/A amendment. SEC prospectus, Anthropic announcement.
Final pricing, offered shares, primary-versus-secondary allocation, net proceeds and post-offering capitalization are not yet known. Primary issuance brings cash into the company and increases shares. An existing investor’s secondary sale brings the company no new cash. These transactions are economically different.
We anchor the balance sheet at 30 June 2026, forecast the second half and 2027–2041, and date the information cutoff 31 October. We do not manufacture a September balance sheet or fill blank prospectus fields with press estimates.
2. The business: models, platform and applications
Anthropic was founded in 2021. It develops the Claude model family, platform and API access, and applications for users and organizations. Claude Code and Cowork support programming and knowledge work respectively. Revenue depends on paid consumption, subscriptions and deployments, rather than everyone who has ever tried a product. Prospectus: Business and Revenue.
Consumption-based revenue was $13.064 billion in H1 2026 and subscription revenue was $3.222 billion, approximately 80.2% and 19.8% of the total. These are billing categories, not official segments with separately reported margins. We do not assign every subscription dollar to one product or all usage revenue exclusively to enterprises. Revenue note.
Major cloud providers are distribution partners, infrastructure suppliers and investors simultaneously. Those relationships can ease customer and capacity access, but do not remove compute costs, dependency risk or financing claims. Technical progress becomes shareholder value only when revenue also covers research, people, selling costs and future cash requirements.
3. Prospectus figures: real growth, different periods
The full-year columns cover twelve months and the final two cover six months. We do not treat half-years and full years as equally long periods. Interim figures are unaudited. First-half revenue grew approximately 14-fold year over year, but that pace cannot simply be extrapolated indefinitely. Statements, F-7 and F-12–F-13.
Q2 2026 alone generated $11.555 billion of revenue and $0.974 billion of EBIT. This gives a more favorable operating picture than the half-year net loss, but one quarter does not establish durable profitability across the investment cycle. Quarterly Results of Operations.
At 30 September, the company reported more than 11,900 customers generating at least $100,000 of TTM revenue, over 1,900 generating at least $1 million and approximately 250 generating at least $10 million. These are nested groups and should not be added. TTM means the trailing twelve months, not guaranteed future annual contracts. Customer adoption.
4. Why an $80.7 billion net loss is not the same cash burn
H1 operating loss was $0.693 billion, while net income included a $72.388 billion noncash fair-value change in convertible notes, derivative liabilities and SAFEs. A rising company valuation can increase investors’ financial claims and generate an accounting expense. That is not a GPU bill paid in the same amount. Similar remeasurements occurred previously, so we do not unconditionally call them “one-time”. Income statement and financing note.
DCF starts with operating economics, so we do not later add back a financing remeasurement that never entered EBIT. Positive CFO of $3.929 billion is not itself sustainable FCFF either: it includes noncash adjustments and changes in receivables, prepayments and settlements. An explicit cash forecast is necessary.
5. AI costs: two presentations of the same spending
The main income statement classifies expense by nature. H1 2026 compute and infrastructure expense was $8.945 billion, personnel $3.152 billion, sales, marketing and partnerships $3.168 billion, and other expense $1.714 billion. Total operating expenses were $16.979 billion against $16.286 billion of revenue. Income statement.
A supplemental functional note also reports $2.380 billion of cost of revenue and $9.741 billion of research and development. Gross margin derived as (16.286 − 2.380) / 16.286 is approximately 85.4%, but it excludes research costs allocated to R&D. It does not mean 85% of sales becomes shareholder cash. Total compute expense is not interchangeable with functional cost of revenue. Supplemental Financial Information.
The model uses a consistent expense architecture. Forecast margins and efficiency are author assumptions, not reported facts. Stock compensation remains an economic cost; adding it back without accounting for ownership claims would overstate an existing share’s value.
6. Cloud, leases and long-term commitments
At June-end, minimum purchase-service commitments were approximately $265.6 billion, with terms of one to twenty years. The prospectus separately discloses approximately $321.4 billion of estimated future undiscounted obligations for leases not yet commenced; some arrangements remain conditional. Some compute and infrastructure agreements have no fixed minimum dollar amount; eventual payments depend on delivered capacity and usage terms. One number is therefore not a complete price for all future infrastructure. Note 13 and lease note.
A cloud service is not automatically a purchased server. If its expense and cash payment already enter the forecast, subtracting the entire contract again from equity would double count the same obligation. Ignoring payments merely because historical PP&E capex is small would be equally problematic.
The model treats future hosted and leased capacity as cash service costs, with owned assets captured separately as capex. Base compute expense from H2 2026 through 2030 totals approximately $343.0 billion, but an aggregate budget does not prove coverage of every contract in every year. A complete timetable of all conditional payments is unavailable. We therefore show a five-percentage-point increase in compute costs relative to revenue: the equity pool falls from $489.2 billion to approximately $390.6 billion. This is an economic stress, not certification of contractual compliance.
7. DCF assumptions and mechanics
We calculate FCFF = EBIT − operating cash tax + depreciation − investment − increase in working capital. The historical consumption/subscription split provides context; the forecast models aggregate revenue. A complete comparable history of token volume, pricing and paid seats is unavailable; revenue paths are explicit analytical assumptions, rather than a unit model pretending to greater precision.
The base assumes revenue of $40 billion in H2 2026, $90 billion in 2027, $220 billion in 2030 and $600 billion in 2041. Compute costs decline from 60% of sales in 2027 to 38% in 2041; personnel from 17% to 10%, sales and partnerships from 18% to 12%, and other costs from 7% to 4.5%. This requires substantial economic improvement. Depreciation and owned capex are separately forecast, with owned capex reaching 6% of revenue in 2041. These ratios are not issuer guidance. The base first achieves positive annual FCFF in 2030.
The risk-free rate is 4.94%, the 30 October US Treasury ten-year yield. Beta, equity risk premium, target financing structure and terminal growth are author assumptions. Base WACC is 12.44%, g 3.0% and terminal ROIC 15.0%. Without a trading history, beta is not a reliable observed beta for a listed ANT share. Treasury.
Terminal value pays for continued growth: after-tax operating income is reduced by reinvestment based on g/ROIC and required maintenance spending. The final forecast year does not continue forever without further capital needs. Tax and working capital are separately scheduled.
8. Scenario values and the equity bridge
No scenario probabilities are assigned. Base enterprise value is $368.2 billion, with terminal value supplying 52.4% of EV. A larger terminal share makes mature economics and capital costs more influential, while short-term apparent precision matters less.
At 30 June, the company held $112.389 billion of cash and $19.262 billion of short-term investments, a combined $131.651 billion. Restricted cash of $2.529 billion is not added as freely available liquidity. We subtract $0.443 billion of finance-lease obligations, a $0.265 billion unfunded investment commitment and an analytical $10 billion operating reserve from liquid assets. The net adjustment is $120.943 billion. Under assumed full conversion, we do not subtract $108.566 billion of convertible-note carrying value or preferred capital and then count the same claims again as shares. This is a conditional post-conversion claim pool, not a valuation waterfall for today’s preference rights. The known remaining $1.2 billion settlement and $0.148 billion of capital creditors are charged once in forecast cash flows.
Undrawn funding facilities and hypothetical IPO proceeds are not free cash today. Drawing them may create debt or new ownership claims. The table’s pool is not the value of existing legal common shares alone before applying preference rights.
9. What could one share be worth?
Assuming full conversion of historical preferred stock and the specified potential Amazon conversion gives approximately 12.5874 billion economic shares on a June basis. Adding all historical options and RSUs gives an illustrative 15.4341 billion claims; full option exercise would bring approximately $5.194 billion of gross cash. These are historical calculations, not the final IPO share count. Using the 15.4341 billion historical claims and adding full option-exercise cash, the base illustration gives approximately $32.03 per share. This is only a conditional June-data example, excluding later changes, not an IPO target price. Notes 9–10.
Amazon’s ownership cap can defer some note conversion. Exercise prices, RSU conditions, tax settlement, later grants, primary issuance and other transactions change the denominator. Weighted-average EPS shares are not a substitute for a current diluted count. The calculator leaves shares blank by default; readers can supply a denominator or select an explicitly dated illustration.
The general post-IPO formula is (pre-issuance value + net primary proceeds − new claim adjustments) / final economic shares. Knowing only company value is insufficient. Until offering data are complete, we do not publish an unconditional ANT price target.
10. Public benefit, control and future dilution
Classes A and N carry economic rights; special classes T, F and E do not carry the same economic entitlement. Control also differs from ordinary capital-proportional voting: Class F is intended to retain 50.1% of votes and the Long-Term Benefit Trust has significant board-election rights. Public benefit corporation status requires consideration of a specified public purpose as well. Governance and Description of Capital Stock.
The controlled Public Benefit Fund received 846.841 million N shares. While held by a wholly owned subsidiary, they are not treated as externally outstanding for accounting purposes; later external transfers can affect economic ownership. The board may also discretionarily issue up to another 2% at successive 50% share-price milestones, subject to observation conditions. This is neither an automatic guaranteed issuance nor a single fixed cap on all future dilution. Founders donating personally held shares is different from the company issuing new shares. Public Benefit Issuances.
11. Reference points, risks and what changes value
On 28 May, Anthropic announced a $65 billion Series H round at a $965 billion post-money valuation. This is a dated private financing term, not a current public market capitalization or directly comparable common-share quote. The announced revenue run rate exceeding $47 billion was a pace measure, not revenue recognized over a completed year. Official Series H announcement.
The main risks include competition and pricing pressure, the cost of frontier-model development, infrastructure availability, distribution concentration, copyright and settlements, regulation, system safety and funding commitments. The model does not predict unknown court outcomes or treat supplier plans as guaranteed profitable sales.
Value increases with stronger retention, profitable consumption and subscription growth, lower unit costs and less capital intensity. It falls with a lost product advantage, underused infrastructure commitments or greater dilution. A good product and rapidly growing revenue do not automatically make a share’s purchase price attractive.
12. Sources and confidence
The foundation is the public 26 October S-1, its statements and notes, SEC filing index, issuer announcements and official Treasury data. The model, charts and explanatory illustrations are our own. All forecasts are nominal USD; detailed model tables use millions and summaries use billions.
This is scenario analysis requiring final offering terms, rather than a finished investment recommendation. The most important next evidence is an amended prospectus, full post-conversion capitalization, IPO pricing and net proceeds, infrastructure schedules and further comparable results. Inputs and calculations are inspectable in the model and downloadable data.
Posts reflect the author’s personal views as of publication. They are not investment recommendations.