ANT · Nasdaq · IPO research

Anthropic: growth has a price

Independent investment assessment

Wait

What the US$83 offer requires

Offer reference · 2026-11-18

US$83.00

Provisional central intrinsic value

US$55.64

12-month fundamental value

US$60.33

Wait at the US$83 offer reference. Our provisional central intrinsic value is US$55.64; US$60.33 is twelve-month fundamental value, not a prediction of the trading price. The offer requires materially stronger operating economics, while closing capitalization and the cash bridge remain unverified.

Valuation cutoff 2026-11-18 · underwriting reviewed 2026-11-19 · USD. The offer is not a live quotation or guaranteed execution price. Fundamental value is not a forecast of the next year’s market price.

Valuation uncertainty: the base rate/growth sensitivity spans US$42.45–US$84.64 at 9–13% required return and 2–4% terminal growth. This is not a confidence interval or a complete downside bound.

Downside: the adverse operating case does not support positive sustainable terminal cash flow; no numerical bear value is assigned.

Evidence gap: final closing shares and cash are unverified. Market-price scenario: unavailable.

What would change the decision?

Reconcile final closing claims/cash and test whether customer collections and retained compute efficiency support stronger per-share economics.

Final capital and cash reconciliation

Test: Economic claims and available liquidity

Needed evidence: Final priced prospectus and completed closing/share/cash disclosures

Recalculate per-share value; keep provisional while material gaps remain.

Productive compute and cash conversion

Test: Utilization, unit costs, retained pricing and collections

Needed evidence: Reported compute payments, collections and comparable operating evidence

Replace proxy hypotheses with evidence and rerun price/compute stresses.

Actual entry economics

Test: Price versus revised intrinsic range and funding exposure

Needed evidence: Dated executable price and applicable investor costs/FX when considering an investment

Reassess expected return; no mechanical buy threshold or trading action.

Consider only after material claims and funding risks reconcile and a revised, stress-tested valuation supports an attractive actual entry. No automatic buy price or trade is authorized.

The strongest countercase

Research capability, enterprise integration and trust can support demand, but rivals and cloud partners can capture efficiency gains. The base explicitly assumes 3% annual price erosion, rising utilization and lower capacity/inference unit costs. These are testable analyst hypotheses, not reported operating metrics. The strongest countercase is a durable platform that retains more of its efficiency benefits.

Reverse DCF: the burden of proof

Each test solves one variable against the US$83 offer at an 11% USD required return and 3% terminal growth. These are alternative conditional hurdles, not uniquely identified market expectations.

Activity scale, all years

1.55× base

Activity changes in every forecast year; capacity and inference spending change with it. Realized pricing, utilization and unit-cost paths stay fixed. Training cash does not scale automatically.

Year-ten revenue US$856.9B · tested bounds 0.5–3×

Mature cash operating margin

42.9%

Only the mature cash margin changes. Explicit revenue, operating costs, cash flows, opening claims and terminal capital turnover stay fixed. The resulting margin must be justified independently.

Year-ten revenue US$551.9B · tested bounds 2.0%–60.0%

Operating drivers and valuation effects

Activity, utilization and unit-cost indices below are analyst hypotheses. They are not measured tasks, tokens, customers, channel segments or physical capacity. The first-year activity index is 100. Revenue is first-year revenue × activity growth × realized-price index. Capacity cash scales with activity ÷ utilization × capacity-unit cost; inference scales with activity × inference-unit cost; training is budgeted separately.

Central operating bridge
Forecast yearActivity indexRealized-price indexUtilization assumptionCapacity-unit cost indexRevenueCapacity / inference / training cashCompute / revenue
202710010065.0%100US$110BUS$60B / US$10B / US$10B72.7%
20281559768.0%94US$165.4BUS$83.6B / US$14B / US$10.8B65.5%
202922094.0971.0%88.36US$227.7BUS$106.8B / US$17.8B / US$11.7B59.8%
203029091.26774.0%83.058US$291.1BUS$126.9B / US$21.1B / US$12.6B55.2%
203137088.52977.0%78.075US$360.3BUS$146.3B / US$24.3B / US$13.6B51.1%
203244085.87379.0%73.39US$415.6BUS$159.4B / US$26B / US$14.7B48.1%
203351083.29781.0%68.987US$467.3BUS$169.4B / US$27.1B / US$15.9B45.4%
203457080.79883.0%64.848US$506.6BUS$173.7B / US$27.3B / US$17.1B43.0%
203562078.37484.0%60.957US$534.5BUS$175.5B / US$26.7B / US$18.5B41.3%
203666076.02385.0%57.299US$551.9BUS$173.5B / US$25.6B / US$20B39.7%

Cost indices begin at 100. Other central assumptions: operating cash 20% of revenue; future-service award economics 3%; positive modeled pretax tax rate 25%; owned capex 7%; working capital 5% of positive incremental revenue. None is issuer guidance. Channel-level reported margins are unavailable.

One-driver effects on central intrinsic value
Change from baseCurrent intrinsicDifferenceAnnual model funding
Activity −20%US$45.74US$-9.90Within modeled buffer
Activity +20%US$65.55US$9.90Within modeled buffer
Realized price −20%US$30.97US$-24.67Within modeled buffer
Compute cash +10%US$47.62US$-8.02Within modeled buffer
Future award cost 5% of revenueUS$52.03US$-3.61Within modeled buffer
Activity and compute-cost matrix
Activity versus baseCompute cash versus linked operating modelIntrinsic valueAnnual model funding
0.8×0.8×US$58.78Within modeled buffer
0.8×1.0×US$45.74Within modeled buffer
0.8×1.2×US$32.50Within modeled buffer
1.0×0.8×US$71.59Within modeled buffer
1.0×1.0×US$55.64Within modeled buffer
1.0×1.2×US$39.52Within modeled buffer
1.2×0.8×US$84.41Within modeled buffer
1.2×1.0×US$65.55Within modeled buffer
1.2×1.2×US$46.54Within modeled buffer

Activity changes revenue and linked capacity/inference costs. The compute multiplier then changes capacity, inference and training cash. These are stress combinations, not probabilities or measured efficiency forecasts.

Cash funding and quarterly liquidity

Eight equal-duration forecast quarters from the research date; analyst timing allocation, not disclosed calendar-quarter or covenant compliance. No option cash or new financing. Economic equity-award charge is noncash in this budget.

Conditional opening cash excluding future option exercises: US$232.1B. Base minimum modeled quarterly cash: US$228B. This does not establish intra-quarter solvency.

Central quarterly cash budget
Period endOpening cashCollectionsCompute cashOther cash usesNoncash award cost — excludedEnding cashBuffer shortfall
2027-02-17US$232.1BUS$27.5BUS$20BUS$8.5BUS$0.8BUS$231.1BUS$0B
2027-05-19US$231.1BUS$27.5BUS$20BUS$8.5BUS$0.8BUS$230.1BUS$0B
2027-08-19US$230.1BUS$27.5BUS$20BUS$8.5BUS$0.8BUS$229BUS$0B
2027-11-18US$229BUS$27.5BUS$20BUS$8.5BUS$0.8BUS$228BUS$0B
2028-02-18US$228BUS$41.3BUS$27.1BUS$13BUS$1.2BUS$229.2BUS$0B
2028-05-19US$229.2BUS$41.3BUS$27.1BUS$13BUS$1.2BUS$230.4BUS$0B
2028-08-18US$230.4BUS$41.3BUS$27.1BUS$13BUS$1.2BUS$231.7BUS$0B
2028-11-18US$231.7BUS$41.3BUS$27.1BUS$13BUS$1.2BUS$232.9BUS$0B

Cash uses include operating cash, taxes, owned capex and growth working capital. Award economics remain in valuation but are not counted as cash payments here. The US$5B buffer is an analytical threshold, not a covenant certification.

Funding stresses — no automatic rescue financing
StressMinimum quarterly cashFirst buffer breachMaximum buffer shortfallAnnual funding test
Base timingUS$228BNo modeled breachUS$0BNo annual breach
Price pressure: realized prices −20%US$200.3BNo modeled breachUS$0BNo annual breach
Lower utilization: 20% below baseUS$202.9BNo modeled breachUS$0BNo annual breach
Deployment spend front-loadedUS$207.5BNo modeled breachUS$0BNo annual breach
Collection delay: 50% of each quarter revenueUS$208.5BNo modeled breachUS$0BNo annual breach
Cash bridge US$25B lowerUS$203BNo modeled breachUS$0BNo annual breach
IPO and placement proceeds unavailableUS$123.4BNo modeled breachUS$0BNo annual breach
Combined severe funding stressUS$-128.3B2027-08-19US$133.3BFails in forecast years 2, 3, 4, 5, 6, 7, 8, 9, 10

Pricing stresses change collections and taxes; lower-utilization stress increases capacity spending without assuming more revenue; deployment stress front-loads compute payments; delayed collections carry receivables into later quarters. Combined stress includes all of these, unavailable offering proceeds and an illustrative US$25B adverse cash bridge. Severities are analyst tests, not predicted outcomes.

The severe case has an unfinanced buffer shortfall of US$133.3B. No investable stress valuation or assumed capital raise hides that failure.

Commitment register — dates and overlap matter
Disclosed categoryAmountTiming / basisOverlap and interpretation
Scheduled long-term obligationsUS$320BAt September30; table p180Includes operating leases4224,financeleases1950,purchases313819; do not add its components again.
Scheduled remaining2026US$5.2BAt September30; remaining calendar2026Subset of319993; not a full-year or November-to-November amount.
Scheduled2027US$29.2BCalendar2027Subset of319993; does not include all excluded commitments.
Scheduled2028US$44.4BCalendar2028Subset of319993.
Short-term minimums excluded from tableUS$17.4BSeptember30 contractsNot included in319993; timing differs.
Uncommenced leases excluded from tableUS$451.2BSeptember30; long datedNot immediately payable; do not add all to first-year cash use.
Post-September minimum obligationsUS$105.7BSubsequent eventsDifferent contract set and timing; excludes additional short-term6600.
Post-September short-term minimumsUS$6.6BSubsequent eventsDo not confuse with September17.4Bshortterm pool.
Approximate next12month paymentsUS$70.7BOctober1,2026–September30,2027Cross-cutting timing estimate including4200conditional/terminable; not additive to above pools and not an exact November forecast-year floor.
Conditional / precommencement termination arrangementsUS$186.2BAggregate arrangements described p179Overlap with other categories; not an extra liability.

Source: amended prospectus, MD&A Contractual Obligations and Notes 7, 13 and 17. Subsets and cross-cutting timing estimates must not be added together. The October 2026–September 2027 US$70.7B disclosure is not an exact floor for a different November-to-November forecast period.

Capitalization, cash and terminal reconciliation

Conditional legal pro-forma A+N shares: 13,203.574 million. Conservative economic claims: 18,199.67 million. Neither is a verified closing share register. Class T/F control rights are separate from economic participation.

Economic share bridge
AdjustmentChange · million sharesRunning economic shares · millionsEvidence status
Pre-offer A + N, after specified conversions and net settlement11,933.9711,933.97conditional
New issuer shares at final offer price956.35112,890.321conditional
Concurrent placement shares at US$83313.25313,203.574conditional
Existing gross options1,774.34314,977.917derived
Existing unvested RSUs1,026.6616,004.577derived
Post-September RSUs47.04116,051.618derived
Acquisition issuances and holdback3.00116,054.619derived
Post-September matching issuance27.99216,082.611derived
Maximum existing vested matching obligation553.44116,636.052derived
Full existing-note conversion1,840.44518,476.497conditional
Less conversion already in pre-offer count-276.82818,199.67derived

Reported, estimated and conditional amounts have different evidence status. Running totals are a reconciliation, not a verified closing register or bank statement.

Opening cash and investment bridge
AdjustmentChange · US$ millionsRunning cash · US$ millionsEvidence status
September unrestricted cash and short-term investments131,845131,845disclosed
Gross primary issuance79,377.162211,222.162conditional
Estimated issuance fees-793.772210,428.391estimate
Concurrent private placement26,000236,428.391conditional
Estimated initial RSU withholding-4,281.055232,147.336estimate
Future option exercises — valuation convention only5,163.337237,310.673conditional
Post-September cash adjustment — unresolved0237,310.673unavailable

Reported, estimated and conditional amounts have different evidence status. Running totals are a reconciliation, not a verified closing register or bank statement.

Acquisition issuances include an additional 1.590190 million holdback shares omitted previously. Gross option shares and exercise proceeds are paired once for valuation; liquidity excludes those prospective proceeds. Initial public benefit shares held in a subsidiary are conservatively treated as future economic claims. Existing awards are represented in shares; future service awards receive one economic cost. No ungranted reserve is automatically treated as outstanding.

From forecast cash flow to mature economics
CaseYear-ten revenue / FCFFGrowth capex releasedExplicit working capital releasedMature cash marginTerminal cash before growthIncremental growth investmentTerminal FCFF
bearUS$264.7B / US$-12.5BUS$7.9BUS$0.2B-1.6%US$-4.5BUS$4BUS$-8.5B
baseUS$551.9B / US$114.9BUS$11BUS$0.9B23.0%US$130.6BUS$6.6BUS$124B
bullUS$1,254.4B / US$404.8BUS$12.5BUS$5.7B33.7%US$435.7BUS$12.5BUS$423.2B

Mature margin uses year-ten operating cost/tax ratios and 5% maintenance capex. Terminal revenue grows 3%; growth reinvestment is incremental revenue divided by sales-to-capital (2× bear, 2.5× base, 3× bull). A cash-flow step-up can be legitimate when growth investment falls. Negative terminal cash is not repaired to manufacture a positive value.

Evidence, verification standard and model download

Every material assumption below is classified. Disclosed facts are linked to source locations; derived figures expose their bridge; estimates require rationale and sensitivity; conditional financing remains conditional. Independent arithmetic and source reconciliation are distinct from human analyst approval.

Material evidence and assumption ledger
Input / assumptionValueEvidence typeBasis / calculationSource location
Offer price83disclosedOffer reference, not executable market quote.November 18 final pricing announcement · November18 pricing announcement
Historical legal Class A2,168.788disclosedMillion shares at September30, before pro-forma adjustments.November 17 amended prospectus · p26 capitalization reconciliation
Historical legal Class N30.643disclosedMillion shares at September30.November 17 amended prospectus · p26 capitalization reconciliation
Preferred conversion A4,408.097conditionalConversion immediately before offering closes.November 17 amended prospectus · p26 capitalization reconciliation
Preferred conversion N4,153.018conditionalConversion immediately before offering closes.November 17 amended prospectus · p26 capitalization reconciliation
Initial cap-limited note conversion276.828conditionalAlready included in pre-offer shares; subtract before adding full conversion.November 17 amended prospectus · p26 capitalization reconciliation
Net initial RSU settlement49.755conditional101.333940m gross awards less51.578975m withholding.November 17 amended prospectus · p26 capitalization reconciliation
Initial public benefit shares846.841conditionalIncluded once as conservative future economic claims; subsidiary-held shares differ from accounting outstanding shares.November 17 amended prospectus · p26 additional adjustments
Acquisition holdback1.59disclosedMissing from prior economic denominator; now included conservatively.November 17 amended prospectus · p26 additional adjustments; Note15
September liquidity131,845disclosedUS$m unrestricted cash plus short-term investments; excludes restricted cash.November 17 amended prospectus · Consolidated Balance Sheets F6
Initial withholding51.579derived101.333940×50.90%=51.57897546m shares; issuer rounded amounts; multiply actual83offer, not preliminary74.November 17 amended prospectus · p24 RSU Settlement,50.90% assumption
Issuance fee0.01estimate1% of primary gross is an analyst estimate pending final prospectus; not verified final fee.November 17 amended prospectus · Use of Proceeds; preliminary terms
Private placement26,000conditionalUS$m; conditional closing, not existing cash.November 17 amended prospectus · Concurrent Private Placement
Post-September cash movementUnavailable / see scheduleunavailableA zero modeling adjustment is not evidence of zero cash movement. Illustrative downside bridge subtracts US$25B.November 17 amended prospectus · Latest statements September30; no current bank statement
Activity / price / utilization hypothesesUnavailable / see scheduleestimateIndices and utilization assumptions are not disclosed measurements or company guidance. Range and reverse-DCF tests expose the burden of proof.November 17 amended prospectus · Business; MD&A; Risk Factors
Future award cost0.03estimate3% of revenue cash-equivalent economic charge covers future service awards and matching; no new share count added for the same charge. Noncash in liquidity budget.November 17 amended prospectus · Equity Plans; Donation Matching Plan
Maintenance capex0.05estimate5% of revenue in mature state, separate from incremental growth investment. Not an issuer estimate.November 17 amended prospectus · MD&A Infrastructure; Peralta model assumption
Required return / terminal growthUnavailable / see scheduleestimateCommon11%USDrequiredreturn and3%terminalgrowth isolate operations; separate9–13% and2–4% sensitivity. No traded beta inference.November 17 amended prospectus · Peralta scenario design

SEC filing-index check · Targeted November19 inspection lists November17 S-1/A, November18 S-1MEF, registration and Form3 filings; no424B4orcompleted closing disclosure. Source check occurred in this review; no claim of a completed November20 closing.

Model: anthropic-ipo-decision-v2. The complete published inputs, scenarios, formulas’ outputs and reconciliation schedules are available in the downloadable model packet. This is a read-only snapshot of this publication, not an editable scenario tool.

Packet SHA-256: 5f89c9e7e0a8ae5d77ed3dc3b515f6cfd5d3f3224629d0c416696c0d2e662b38

Methodology informed by the actual 2026 CFA refresher readings inspected on the Mac mini: Equity Valuation: Applications and Processes, pp. 4–9; Introduction to Financial Statement Modeling, pp. 4–7. The independent Peralta analysis is not CFA Institute approval. No claim is made to a completed review of the 2026 FCFF chapter.

  • Operating indices are analyst hypotheses; no absolute useful-work unit, channel revenue split or measured utilization is claimed.
  • Quarterly timing is modeled, not a disclosed contractual calendar; no intra-quarter solvency or covenant certification.
  • No final priced prospectus or completed closing was located in the November19 SEC index check.
  • Post-September cash and net issuance costs are unverified; US$25B cash uncertainty is an illustrative stress, not an estimated actual burn.
  • Future equity financing is not assumed; funding shortfalls remain visible. Existing awards are economic shares; future service awards are charged once as economic cost, not cash in liquidity.
  • Reverse DCF is a conditional hurdle at fixed 11% / 3%, not uniquely inferred market expectations.
What changed in this edition

Capital claims corrected

Before: 18.198080B economic shares

Now: 18.199670B including acquisition holdback

Corrects omission; legal outstanding and economic claims remain distinct.

Operating assumptions separated from discount inputs

Before: Bull intrinsic US$194.19 at 9% required return/3.5% growth, with operating assumptions also changing

Now: Common 11%/3% operating cases; base US$55.64, upside US$162.16; adverse case not valued

Isolates operating outcomes; new indexed hypotheses also change cash forecasts, so this is not solely a rate effect.

Fundamental value relabeled and recalculated

Before: US$52.20 intrinsic; US$56.51 twelve-month target

Now: US$55.64 intrinsic; US$60.33 twelve-month fundamental value

Not a prediction of next year’s trading price. Recommendation remains Wait.

Financing risk made visible

Before: Annual economic-cash budget; all published scenarios required usable

Now: Quarterly cash timing, noncash-award separation, unavailable proceeds and explicit severe funding gap

No automatic rescue financing or hidden funding failure.

Publication proofreading

Before: Initial v2 prose contained spacing artifacts in several words and citations

Now: Corrected wording and source-locator typography; financial figures unchanged

Improves readability without changing the investment assessment or any numerical assumption.

Prior packet 68595c525f65cf5944ffa942fc6b365518b3a1703fe154f6564155ea11ce54df. The full prior edition is preserved in the trusted publication receipt; it has not been silently discarded. These are manual review checkpoints, not live monitoring or alerts.

Wait at the US$83 offer reference. Provisional central intrinsic value is US$55.64; US$60.33 is twelve-month fundamental value, not a market-price forecast. Final claims and the cash bridge remain unverified.

Company and research basis

ANT · Nasdaq · USD

Anthropic, PBC is a Delaware public benefit corporation founded in 2021 and headquartered in San Francisco. The issuer announced an IPO price of US$83 on November 18, 2026, with Nasdaq trading expected November 19 under ANT and closing expected November 20, subject to conditions. US$83 is the offering reference used here; it is not a verified opening trade, current quotation or attainable execution price.

Evidence: November 18 final pricing announcement · November 18 pricing announcement

What this report covers

Valuation cutoff is November 18, 2026; underwriting reviewed November 19 using the latest amended prospectus, issuer pricing and a targeted SEC filing-index check. Financial statements end September 30 and interim figures are unaudited. Closing and final priced capitalization remain unverified. This edition adds indexed operating assumptions, separate rate sensitivities, reverse valuation and cash-timing stresses. It supplies no observed trading price, measured channel split or market-price forecast.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Prospectus cover; Financial Statements F-6–F-14; pricing announcement

Investment summary

A strong product does not settle the investment case

Wait at the US$83 offer reference. Our provisional central intrinsic value is US$55.64; US$60.33 is twelve-month fundamental value, not a prediction of the trading price. The offer requires materially stronger operating economics, while closing capitalization and the cash bridge remain unverified. Revenue growth can create value only if retained pricing and productive capacity cover compute, staff, distribution, compensation and reinvestment. The recommendation is evidence-gated rather than a claim that the stock must fall to our value estimate.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Business; MD&A; Peralta FCFF model, base case

The disagreement is cash conversion

At 11% discounting and 3% terminal growth, the offer requires approximately 1.55× the base activity path with linked capacity and inference spending, or a 42.9% mature cash operating margin at unchanged revenue. The latter compares with the base-derived 23.0%. These are conditional alternative hurdles, not jointly required assumptions or uniquely inferred market beliefs.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Business; Risk Factors; Management’s Discussion and Analysis

Pricing changed the starting valuation

The final announcement increased the public offering from the roadshow’s 825 million shares to 980 million and set the price at US$83, above the initial US$70–78 range. Of the final shares, 956, 351, 354 are newly issued and 23, 648, 646 are sold by existing holders. Secondary proceeds do not fund Anthropic. The US$26 billion concurrent private placement is part of the provisional financing bridge, subject to its closing terms.

Evidence: November 18 final pricing announcement · November 18 pricing announcement

Operating leverage must survive capacity expansion

The base first annual forecast is US$110B revenue, ending at approximately US$551.9B in year 10. Activity rises 6.6×from its first-year index while realized yield declines 3% annually. Assumed utilization rises 65% to 85%; capacity-unit cost falls 6% and inference-unit cost 10% annually. These estimates require large adoption and efficiency gains; the company has not disclosed equivalent physical operating units that independently validate them.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Consolidated Statements of Operations F-7; Peralta base forecast

Our own target, with a wide range

Central intrinsic value is US$55.64; twelve-month fundamental value is US$60.33. The upside operating case gives US$162.16 current intrinsic value at the same 11%/3% discount assumptions. The adverse case has no positive sustainable terminal cash flow and no assigned price. This is an asymmetric set of conditional outcomes, not a probability range or market forecast.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model; Capitalization; Description of Capital Stock

Wait before adding

Wait before adding at the offer reference. Reassess when final claims and cash reconcile or new operating evidence changes retained pricing, compute cost, utilization or collections. Consider investing only after funding and concentration risks are acceptable and a revised valuation supports an attractive actual entry. There is no automatic buy price, promised margin of safety or trading instruction.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta investment assessment; Risk Factors

Evidence that can change the thesis

Reassess on the final priced and closing share/cash reconciliation, evidence of cash retained after compute and reinvestment, and operating disclosures that test activity, pricing, utilization and efficiency. These are manual evidence checkpoints; no live monitoring or alerts have been activated. A stronger product does not necessarily create a near-term stock-price catalyst.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Shares Eligible for Future Sale; MD&A; Peralta thesis tests

The strongest countercase

The adverse operating case does not earn a positive sustainable terminal cash flow under its stated pricing, capacity and reinvestment assumptions; we therefore withhold a numerical bear valuation. This is more informative than inserting a positive terminal margin to rescue the result. A combined severe funding stress also exposes a cash shortfall. Both could require a financing or restructuring analysis outside this positive-perpetuity model. The cautious view could still be wrong if durable enterprise integration and efficiency support much stronger retained cash margins.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Business; Risk Factors; Peralta scenario analysis

Business description

A frontier model developer

Anthropic develops advanced AI models and provides access through its platform and applications. Its public benefit purpose gives responsible development a formal place in corporate governance. The economic question for shareholders is whether model capability, reliability and enterprise integration can create a durable stream of customer cash exceeding the recurring cost of training, inference, talent and infrastructure. A public benefit structure does not itself establish commercial advantage.

Evidence: November 17 amended prospectus · Business; Note 1, Description of Business

Subscriptions and usage monetize intelligence

Customers access Claude through subscriptions, enterprise arrangements and developer usage. The model combines recurring access with consumption-sensitive activity, so revenue can grow through more customers, wider deployment and more intensive use. Revenue quality depends on retention, realized prices and collections. It should not be treated as pure software subscription revenue with negligible marginal costs: each increment of inference can consume expensive compute, and channel distribution can carry significant fees.

Evidence: November 17 amended prospectus · Business; MD&A, Revenue and Sales, Marketing, and Partnerships

Claude across applications and development

Claude is the core model family underpinning customer-facing applications and developer access. Coding and workflow use broaden the opportunity beyond a general chat interface. The investment case depends on customers moving from experimentation to repeat production use, where reliability, security, latency and task completion matter. No standalone product segment profit is supplied here; consolidated results cannot be attributed to an individual application without a disclosed reconciliation.

Evidence: November 17 amended prospectus · Business, Products and Platform; MD&A

Adoption, usage and realized yield

Revenue expansion can come from new enterprise customers, broader deployment within existing accounts and greater usage per workflow. These drivers interact: lower unit prices may stimulate volume while reducing revenue per unit of computation. We therefore assess revenue together with compute cash and collections. A headline annualized run rate is not a substitute for recognized revenue, and a contract or capacity commitment is not customer cash already earned.

Evidence: November 17 amended prospectus · MD&A, Key Factors Affecting Performance and Revenue

Compute is central to the economics

Nine-month 2026 compute and infrastructure expense was US$18.808 billion; personnel expense was US$4.863 billion and sales, marketing and partnerships expense was US$5.529 billion. These categories do not form a conventional cost-of-revenue statement, so a GAAP gross margin cannot be reconstructed simply by subtracting compute expense. Training intensity, inference efficiency, hardware availability and partner charges all affect the cost of delivering growth.

Evidence: November 17 amended prospectus · Consolidated Statements of Operations F-7; MD&A expense definitions

A US issuer with a global supply chain

Anthropic is headquartered in the United States, while its supply chain depends on internationally concentrated semiconductor manufacturing and data-center inputs. The filing highlights Taiwan-related supply exposure and cross-border restrictions. This report does not assign an unsupported revenue-by-country mix. Customers’ geographic location, the location of compute and the domicile of the issuer create different regulatory, currency and continuity risks.

Evidence: November 17 amended prospectus · Note 1; Risk Factors, Supply Chain and International Operations

Measure useful work against its full cash cost

The most useful operating tests are customer retention and collections, useful tasks completed per unit of compute, utilization of committed capacity and the cost of acquiring and supporting enterprise usage. Better models can improve demand but also require new training expenditure. Efficiency creates value only to the extent it is retained after price competition and additional usage. The report therefore models compute cash explicitly rather than assuming a standard software margin.

Evidence: November 17 amended prospectus · Business; MD&A; Peralta model design

Partners widen reach and retain bargaining power

Direct enterprise and developer access coexist with distribution through major cloud partners. That reach can reduce friction for customers already buying through a cloud platform, but partner economics can reduce Anthropic’s retained revenue and strategic freedom. Some partners are also investors, infrastructure suppliers or competitors. Their combined role should be evaluated through contract terms and net cash retained, rather than treating every distribution relationship as an unqualified moat.

Evidence: November 17 amended prospectus · Business; MD&A, Sales, Marketing, and Partnerships; Related Party Transactions

Trust and legal rights are operating requirements

AI development faces privacy, copyright, safety, export-control and sector-specific restrictions. Customer confidence requires reliable handling of data and credible controls over model behavior. Compliance can constrain available data, deployment and capacity while raising cost. The filing’s litigation and risk disclosures provide evidence of exposure; an issuer’s safety commitments are not proof that all technical or legal risks have been resolved.

Evidence: November 17 amended prospectus · Risk Factors; Legal Proceedings; Note 13

Industry and competitive positioning

An industry joining research and infrastructure

Frontier AI combines model research, large-scale computing, application distribution and enterprise integration. It sits between capital-intensive infrastructure and software services. That mix makes conventional software comparisons incomplete: leading capability can attract customers, yet retaining it may require repeated training investment and access to scarce chips and power. Industry growth should be assessed alongside capital requirements and the distribution of profits between model vendors and suppliers.

Evidence: November 17 amended prospectus · Business; Risk Factors; Management’s Discussion and Analysis

Rapid adoption does not remove the cycle

Enterprise use is expanding quickly, but budgets, deployment approvals and returns on customer projects can constrain adoption. Capacity lead times can amplify the cycle when commitments are made before demand is certain. A slower customer rollout can leave fixed commitments underutilized; shortages can limit growth even when demand is strong. No third-party total-addressable-market forecast is used as a valuation input in this report.

Evidence: November 17 amended prospectus · Business; Risk Factors; Management’s Discussion and Analysis

Competition comes from several directions

Anthropic competes with other frontier model developers, large technology companies with their own models and distribution, and open or lower-cost alternatives. Hyperscalers can combine infrastructure, cash generation and existing customer channels. Smaller vendors can compete on specialization or price. The relevant test is whether Claude provides enough incremental customer value to retain demand and pricing after rivals improve, not whether it leads one benchmark at one moment.

Evidence: November 17 amended prospectus · Business; Risk Factors; Management’s Discussion and Analysis

Revenue establishes relevance, not measured share

The scale and growth of disclosed revenue demonstrate commercial traction. They do not establish a precise share of a consistently defined AI market: reported revenue across competitors can differ in product scope, gross-versus-net presentation and recognition. This report does not supply a fabricated market-share percentage. Positioning must be tested through retention, expansion and cash economics as comparable public disclosures develop.

Evidence: November 17 amended prospectus · Business; Risk Factors; Management’s Discussion and Analysis

A plausible moat still needs economic evidence

Research capability, a recognizable product, enterprise trust and integration into customer workflows may support differentiation. Switching can become harder when customers build processes around a model, but model access and applications are also changing rapidly. We view these as possible sources of advantage, not proven perpetual pricing power. Durable evidence would be strong retention and useful-work efficiency alongside a sustained margin after training and distribution costs.

Evidence: November 17 amended prospectus · Business; Risk Factors; Management’s Discussion and Analysis

Efficiency gains can be competed away

Improved inference efficiency can lower the cost of serving a task. The benefit to shareholders depends on whether prices fall at the same speed and whether increased usage requires more total capacity. A technically cheaper token is not automatically a higher cash margin. The base case assumes significant improvement over time but still includes large compute outlays; the bear case represents weaker retention of efficiency gains.

Evidence: November 17 amended prospectus · Business; Risk Factors; Management’s Discussion and Analysis

Scale creates barriers and replacement risk

Training frontier models requires talent, computing access and financial resources. Those barriers protect current scale but do not exclude substitutes, including specialized models, open alternatives or customer-built systems. Customers may route different tasks to different vendors. A successful model can therefore face pressure from both well-funded frontier rivals and cheaper tools that are sufficient for a narrower task.

Evidence: November 17 amended prospectus · Business; Risk Factors; Management’s Discussion and Analysis

Strategic dependence runs both ways

Large compute and hardware partners supply essential capacity and can finance expansion, while also distributing or competing with AI models. This creates access advantages and concentration risk together. The disclosed commitments are much larger than a routine variable software expense. Enterprise buyers can negotiate on price and diversify vendors; the strongest evidence of bargaining power is cash retained after both sides have been paid.

Evidence: November 17 amended prospectus · Business; Risk Factors; Management’s Discussion and Analysis

Physical and legal constraints can cap growth

Power connections, cooling water, permits and semiconductor availability can delay capacity. Copyright, privacy and AI-safety rules can alter development and deployment economics. These constraints need not move with customer demand and may affect competitors differently. Our required return and cash-cost scenarios reflect uncertainty, but they are not a quantified probability of a particular regulatory or infrastructure event.

Evidence: November 17 amended prospectus · Risk Factors, Infrastructure, Energy, Regulation and Litigation

Financial analysis

Revenue growth has outpaced operating expense

Revenue increased from US$2.545 billion to US$33.242 billion in the comparable nine-month periods, while operating expense rose from US$7.769 billion to US$34.502 billion. Operating loss narrowed to US$1.260 billion. The operating loss is distinct from the much larger US$104.039 billion net loss, which includes substantial financing remeasurement. Growth has improved the operating cost ratio, but profitability is not yet established across a full capacity cycle.

Evidence: November 17 amended prospectus · Statements of Operations F-7; MD&A

Financing explains most of the liquidity increase

September unrestricted cash and investments total US$131.845B. The pro-forma bridge adds conditional issuer proceeds and a US$26B private placement, deducts estimated fees and withholding, and shows option proceeds separately. The pre-IPO balance-sheet deficit includes financing measurement effects and redeemable preferred claims. It cannot be turned into a normal debt ratio or subtracted again after those claims are represented as converted shares.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Balance Sheets F-6–F-7; Cash Flows F-13

Positive operating cash needs a fuller bridge

Nine-month operating cash was US$6.569 billion. Deducting US$925 million of owned equipment purchases, US$1.955 billion of leased-asset and equipment prepayments and US$88 million of finance-lease principal leaves US$3.601 billion. This historical diagnostic is not normalized free cash flow: it includes working-capital timing and noncash compensation addbacks and does not capture every future contractual cash requirement.

Evidence: November 17 amended prospectus · Cash Flows F-12–F-14; Peralta arithmetic

The business has changed scale quickly

Revenue rose from US$41 million in 2023 to US$386 million in 2024 and US$4.585 billion in 2025, followed by US$33.242 billion in the first nine months of 2026. Comparisons begin from a very small base and span changing infrastructure and product scale. A mechanically extrapolated growth rate would overwhelm any realistic capacity or competitive analysis. We therefore model explicit annual revenue and costs rather than extending the historical percentage growth rate.

Evidence: November 17 amended prospectus · Statements of Operations F-7

Separate operating performance from financing marks

The nine-month US$94.324 billion fair-value financing charge is largely noncash and should not be projected as recurring compute cost. Removing it does not make all remaining earnings cash: US$3.255 billion of stock-based compensation, donation-matching liabilities and other financial-instrument adjustments also affect the reconciliation. We account for existing awards in diluted shares and prospective service awards in future economics, avoiding both omission and double counting.

Evidence: November 17 amended prospectus · Statements of Operations; Cash Flows; Notes on Financing and Stock Compensation

Cash is large; commitments are larger

Eight modeled quarters separate collections from compute and other cash uses. Future-service equity awards remain an economic charge in valuation but are noncash in this liquidity schedule. Equal-duration quarter timing is an analyst allocation from the research date, not an issuer payment calendar. The combined severe case overlays weaker prices, higher compute cash, front-loaded deployment, delayed collections, unavailable offering proceeds and a lower cash bridge; it exposes an unfinanced shortfall rather than assuming rescue capital.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Contractual Obligations; Liquidity; Peralta model liquidity review

Use transparent measures while earnings are negative

Nine-month operating margin is -3.8%, compared with -205.3% a year earlier; operating cash is 19.8% of revenue. Compute and infrastructure expense represents 56.6% of revenue, but this is an expense ratio, not a GAAP gross margin. Negative earnings make a P/E comparison uninformative. The separate ratios table shows the offer-based economic valuation and clearly labels annualized interim revenue as a simple scale reference.

Evidence: November 17 amended prospectus · Statements of Operations and Cash Flows; Peralta arithmetic

A demanding base case, not guidance

The central operating hypotheses produce year 10 revenue of US$551.9B and compute cash of US$219.1B, or 39.7% of revenue. Capacity, variable inference and training costs are derived separately. Mature cash margin of 23.0% uses 5% maintenance capex; incremental terminal growth is funded separately. These are Peralta assumptions requiring evidence, not guidance.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta forecast; Contractual Obligations

Do not normalize away continuing economic burdens

The US$1.5 billion Bartz settlement charge in 2025 is separately identified rather than repeated mechanically. Financing marks also require separate treatment. In contrast, research expense, continuing employee incentives and the recurring cost of compute are central economic demands, even when their accounting presentation changes. We do not turn one favorable period of operating cash into an assumed permanent cash margin.

Evidence: November 17 amended prospectus · MD&A, Bartz Settlement; Cash Flows; Risk Factors

Valuation analysis

Ten explicit annual forecasts

The base activity index rises from 100 to 660, while annual price erosion of 3% reduces realized yield to approximately 76% of first-year yield. The resulting revenue path is derived from both assumptions. Activity is a normalized analytical index, not a disclosed task/token/customer count. Revenue-by-channel and physical throughput cannot be independently reconstructed from the current disclosure set; we do not manufacture those segments.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model, revenue schedules

Cash margin must include the full service burden

Compute cash is the sum of capacity commitments scaled by activity/utilization and unit cost, variable inference scaled by activity and unit cost, and a separate training budget. Other operating cash is 20% of base revenue; future-service award economics are 3%; cash tax is 25% of positive modeled pretax economics. The base-derived mature margin is 23.0%, while adverse operating economics do not support a positive perpetuity. No independent observed utilization or task-cost dataset is claimed.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model; Statements of Operations; Stock Compensation

Growth requires funded capital

Explicit base owned capex is 7% of revenue, with 5% of incremental revenue required for working capital. The mature state assumes 5% maintenance capex, followed by incremental growth investment using 2.5×sales-to-incremental-capital. The terminal bridge shows the release of explicit growth spending and working capital. Lease and hosting cash remain in compute and are not deducted again as the same financial debt.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model; Liquidity; Commitments

A USD required return, not a measured IPO beta

All operating cases use 11% USDrequired return and 3% terminal growth to isolate operating outcomes. A separate grid varies required return 9–13% andgrowth 2–4%. These are analyst assumptions, not an IPO beta or a measured market risk premium. Differences between the original combined bull outcome and the new common-rate operating case are explained in the revision history; the discount rate is not adjusted to make the offer appear reasonable.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model, discount assumptions

Most value still depends on the distant economics

Base year 10 economic free cash flow is US$114.9B. Terminal cash before incremental growth investment is US$130.6B; growth investment and the resulting terminal FCFF are shown in the model bridge. The mature margin is derived from year 10 costs with maintenance capex replacing total capex. The adverse terminal outcome remains nonpositive and is not repaired with an arbitrary positive margin.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model, terminal bridge

Use peers as economic checks

No matched-date public peer quotation set or Street target has been verified. The offer-based ratios remain transparent context; diversified technology firms are not interchangeable with a standalone frontier model developer. Reverse valuation is more useful here than importing an unsupported peer multiple: it shows what activity or mature cash economics would be needed to support US$83.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Business; Risk Factors; Peralta valuation methodology

Provisional value remains below the offer

Wait at the US$83 offer reference. Our provisional central intrinsic value is US$55.64; US$60.33 is twelve-month fundamental value, not a prediction of the trading price. The offer requires materially stronger operating economics, while closing capitalization and the cash bridge remain unverified. The updated denominator adds the acquisition holdback, and all operating cases now share discount assumptions. The resulting change is an underwriting revision, not evidence that the new point estimate is a precise market forecast.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model; Capitalization; pricing announcement

No numeric probability assigned

The report does not assign a calibrated probability of convergence to the target. A newly listed company has limited public evidence on execution, price discovery and shareholder behavior. Turning scenario labels into precise probabilities would suggest evidence we do not possess. The appropriate output is a conditional value range with explicit operating tests and uncertainty about both outcomes and timing.

Evidence: November 17 amended prospectus · Peralta scenario methodology

Evidence needed for greater confidence

Confidence should improve only when audited and interim public disclosures establish customer cash conversion, a reconciled post-closing share count, capacity deployment and the cost of sustaining model performance. The offer price is an issuer-announced transaction reference, not proof that the market agrees with a particular long-term forecast. We will reassess the model when operating evidence or final financing terms change, rather than treating elapsed time as convergence.

Evidence: November 17 amended prospectus · Peralta investment assessment; prospectus and pricing limitations

Scenario analysis

Bear: growth fails to absorb committed costs

Adoption grows, but pricing declines 5% annually, utilization reaches only 70%, and capacity-unit cost falls 6% annually. These are analyst stress assumptions, not disclosed operating metrics. The resulting mature cash economics are nonpositive after maintenance and growth investment. A positive-perpetuity DCF does not apply, so no intrinsic value or twelve-month fundamental value is assigned. This is not a zero-price prediction; financing, restructuring or a different recovery framework would need separate analysis.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model, bear case

Base: strong adoption, demanding cash burden

Activity rises 6.6× from first-year index while realized yield declines 3% annually; utilization improves from 65% to 85%. Capacity-unit cost declines 6%, variable inference-unit cost 10%, and training grows 8% annually. All are unverified analyst hypotheses requiring operating evidence. At the common 11% discount rate and 3% terminal growth, current intrinsic value is US$55.64 and twelve-month fundamental value is US$60.33. The five-year conditional annualized USD return from the offer reference is 0.4% after illustrative costs. No probability or forecast trading price is assigned.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model, base case

Bull: efficiency and enterprise adoption reinforce each other

Activity rises 11× and utilization reaches 90%, with stronger capacity and inference efficiency. First-year revenue is US$150B. Exceptional execution is required; no probability is assigned. At the common 11% discount rate and 3% terminal growth, current intrinsic value is US$162.16 and twelve-month fundamental value is US$178.56. The five-year conditional annualized USD return from the offer reference is 25.8% after illustrative costs. No probability or forecast trading price is assigned.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta FCFF model, bull case

Market, Street and our central assessment

Operating outcomes now share 11% USDdiscounting and 3% terminal growth, with rates varied separately. There is no verified market-implied case, Street distribution or calibrated scenario probability. The base is a working central hypothesis, not a measured most-likely outcome. USD returns cannot be compared directly with MSCI ACWI IMI CAD without investor-specific FX and costs; XIC remains only the Canadian-sleeve comparator.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta scenario and benchmark methodology

Investment risks

Capacity and model execution

Model training, inference reliability and capacity commissioning must work together. Chip delivery, data-center readiness and access to power and cooling can constrain service or leave assets underused. Technical leadership can decay quickly. Monitor productive capacity and customer outcomes rather than treating contracted capacity as delivered revenue; a missed deployment can weaken both the growth path and the cost position.

Evidence: November 17 amended prospectus · Risk Factors, Infrastructure, Model Development and Suppliers

Competition can transfer value to customers

Rival frontier models, hyperscaler bundling and lower-cost substitutes can reduce realized prices or retention. Suppliers may keep scarcity rents even when end-customer prices fall. A superior product today may require large recurring investment to remain competitive. Our cash-margin assumptions are therefore a material source of downside risk, particularly if efficiency gains benefit customers more than shareholders.

Evidence: November 17 amended prospectus · Risk Factors, Competition; Business

Annual cash tests can miss timing stress

The adverse operating case does not earn a positive sustainable terminal cash flow under its stated pricing, capacity and reinvestment assumptions; we therefore withhold a numerical bear valuation. This is more informative than inserting a positive terminal margin to rescue the result. A combined severe funding stress also exposes a cash shortfall. Both could require a financing or restructuring analysis outside this positive-perpetuity model. The quarterly cash schedule excludes option proceeds and does not certify intra-quarter solvency or legal covenant compliance. The US$25B cash-bridge stress is illustrative, not an estimate of actual cash burn.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Liquidity; Contractual Obligations; Capitalization

Copyright, privacy and safety exposure

Legal claims can affect access to training data, generate damages or require operational changes. Privacy, export-control and AI rules can limit deployment and raise compliance costs. The 2025 Bartz charge demonstrates that legal exposure can be financially material; it does not quantify all future claims. The report carries no assumption that safety policies eliminate model misuse, output errors or liability.

Evidence: November 17 amended prospectus · Risk Factors; Legal Proceedings; Note 13

Both our caution and our target can be wrong

The base requires strong adoption and sustained declines in compute cost per unit of activity. Efficiency may be passed to customers in lower prices, preventing those gains from reaching shareholders. Reverse valuation quantifies how much stronger activity or mature margin must be to justify the offer. Final share claims, unverified cash movements and operating proxies can still materially change the result; the point estimate should not conceal that uncertainty.

Evidence: November 17 amended prospectus · Peralta ANT decision model v 2; Peralta sensitivity review; Capitalization; Risk Factors

Environmental, social and governance

Power and water are material operating inputs

The filing identifies electricity availability, utility rate allocation, cooling-water scarcity and permitting as constraints on data-center economics. Higher temperatures and physical disruption can increase cooling requirements or interrupt service. Exposure exists through partners as well as owned infrastructure. We have not verified a complete comparable emissions, water-use or renewable-energy baseline for this report, so no ESG score, carbon-neutral claim or unsupported emissions estimate is assigned.

Evidence: November 17 amended prospectus · Risk Factors, Energy and Water; Environmental and Climate Disclosures

Trust depends on product behavior and data rights

AI output errors, misuse, security failures and privacy breaches can harm customers and weaken adoption. Training-data rights and copyright claims create obligations to creators as well as legal exposure. Skilled researchers and engineers are essential, with retention costs reflected in compensation and equity claims. The filing’s stated public-benefit and safety goals are relevant governance commitments, but they are not independent proof of safe outcomes or an absence of social harm.

Evidence: November 17 amended prospectus · Risk Factors, AI Safety, Privacy, Intellectual Property and Personnel; Legal Proceedings

Control differs from economic ownership

The Long-Term Benefit Trust’s Class T rights include authority over a board majority. Founder LLC’s Class F is designed to hold 50.1% of general voting power after the offering; these control shares do not carry ordinary economic participation. Class A and non-voting Class N provide different voting rights. The structure can protect a long-term mission while limiting outside shareholders’ ability to change direction or discipline management. It should be assessed as a tradeoff, not awarded an automatic governance premium.

Evidence: November 17 amended prospectus · Principal and Selling Stockholders, p.300–301; Description of Capital Stock; Governance

Material ESG risks transmit through cash and control

Power and water restrictions can delay capacity or raise compute costs. Data and safety failures can produce legal expense, customer loss and deployment limits. Mission-oriented control can permit decisions that favor the public benefit purpose over near-term shareholder returns. Strategic partners’ simultaneous roles as investors, suppliers and channels add conflict-management demands. No separate arbitrary ESG valuation haircut is layered on top of cash costs already modeled.

Evidence: November 17 amended prospectus · Risk Factors; Related Person Transactions; Peralta accounting of economic risk

Responsible execution can support enterprise adoption

Reliable systems, careful data handling and transparent oversight can help customers deploy AI in sensitive workflows. Better useful-work efficiency can also reduce resource use per task, although total energy and water demand may still rise with adoption. Evidence of these benefits should come from measurable outcomes, customer retention and resource disclosure. We treat them as opportunities requiring validation rather than assuming a sustainability advantage translates automatically into pricing power.

Evidence: November 17 amended prospectus · Business; Risk Factors; Peralta qualitative assessment

Valuation and profitability ratios

Offer-based reference, not a peer consensus

The following ratios use the US$83 offer reference, the fully converted economic share estimate and modeled opening cash from the same valuation packet. This differs from a basic legal-share market capitalization. Annualized nine-month revenue is a simple four-thirds scaling for context, not a forecast or trailing-twelve-month figure. No matched-date peer quotation set or Street consensus has been verified, so those numerical comparisons are unavailable.

Evidence: November 17 amended prospectus · Peralta arithmetic; Financial Statements; Capitalization

Anthropic reference ratios
MetricValue / basis
Offer-based economic equity valueUS$1,510.6B
Operating enterprise value proxyUS$1,273.3B
EV / FY2025 revenue277.7×
EV / annualized 9M2026 revenue28.7×; simple scaling only
9M2026 operating margin-3.8%
9M2026 CFO / revenue19.8%
9M2026 cash diagnostic / revenue10.8%
P/E / GAAP gross marginNot meaningful / not separately presented
Numerical peer multiples / Street targetsUnavailable: no verified matched-date data

Equity = offer price × economic diluted shares. EV proxy = equity less modeled opening cash; converted claims are included in shares and operating leases in compute cash. This is a model-based proxy, not a quoted provider EV. Swipe the table to view all columns.

Source: November 17 amended prospectus · Peralta capitalization and financial arithmetic

Income statements

Reported historical income statements

Selected consolidated US GAAP line items are shown in US$ millions. FY columns end December 31; the nine-month columns end September 30 and are unaudited. Interim periods are not full-year results. Financing fair-value charges are shown separately from operating expense. The full filing and accompanying notes remain the authoritative statements; no artificial quarterly or five-year listed history is supplied.

Evidence: November 17 amended prospectus · Consolidated Statements of Operations F-7

Consolidated income statement · selected lines
US$ millionsFY2023FY2024FY20259M2025*9M2026*
Revenue413864,5852,54533,242
Compute and infrastructure4332,5227,3264,38118,808
Personnel954541,8391,0274,863
Bartz settlement001,5001,5000
Sales, marketing and partnerships2435683225,529
Other operating and administrative623471,4125395,302
Total operating expense5923,36612,6457,76934,502
Operating loss-551-2,980-8,060-5,224-1,260
Interest income691064312392,359
Financing fair-value changes-1,045-5,434-34,308-20,107-94,324
Other income / expense-20-8-3-10,397
Total other expense-978-5,328-33,885-19,871-102,362
Pretax loss-1,529-8,308-41,945-25,095-103,622
Income tax expense212012417
Net loss-1,531-8,309-41,965-25,107-104,039

* Unaudited interim periods. Expenses are positive; losses are negative. No conventional cost-of-revenue or gross-profit line is manufactured. Swipe the table to view all columns.

Source: November 17 amended prospectus · Consolidated Statements of Operations F-7

Balance sheets

Pre-offering balance sheets

These are historical pre-offering balance sheets in US$ millions, not post-IPO capitalization. September 30, 2026 is unaudited. Convertible notes include fair-value measurement effects and redeemable preferred stock is presented outside stockholders’ equity. Applying their book carrying values as ordinary debt deductions on top of fully converted economic shares would mix incompatible valuation treatments.

Evidence: November 17 amended prospectus · Consolidated Balance Sheets F-6–F-7

Historical balance sheet · selected lines
US$ millionsDec2024Dec2025Sep2026*
Cash and equivalents3,14117,38895,562
Short-term investments02,88936,283
Accounts receivable, net959097,892
Prepaid / other current assets2,0762,5601,283
Total current assets5,31223,746141,020
Restricted cash13132,540
Property and equipment, net191463,061
Operating lease ROU assets351262,839
Goodwill0154595
Other noncurrent assets2,3734764,265
Total assets7,75224,661154,320
AP, accruals and other current liabilities3332,5119,426
Operating lease liabilities, current1337274
Deferred revenue282691,971
Total current liabilities3742,81711,671
Operating lease liabilities, noncurrent39802,536
Convertible notes, carrying value14,11641,677130,502
Other noncurrent liabilities1508231,974
Total liabilities14,67945,397146,683
Redeemable convertible preferred stock2,87429,769157,510
Stockholders’ deficit-9,801-50,505-149,873

* Unaudited. Restricted cash is separate from cash and equivalents. Totals include all statement categories; the identity assets = liabilities + redeemable preferred + stockholders’ deficit reconciles. Swipe the table to view all columns.

Source: November 17 amended prospectus · Consolidated Balance Sheets F-6–F-7

Cash-flow statements

Cash movements and the financing bridge

Selected US GAAP cash-flow lines are in US$ millions. Negative amounts are uses. Investment purchases are cash movements between liquidity categories and should not be mistaken for operating capex. Beginning and ending balances include restricted cash. The capital-spending diagnostic below adds leased-asset prepayments and finance-lease principal to owned equipment purchases, preserving the distinction between operating cash and cash retained after those uses.

Evidence: November 17 amended prospectus · Consolidated Statements of Cash Flows F-12–F-14

Consolidated cash-flow statement · selected lines
US$ millionsFY2023FY2024FY20259M2025*9M2026*
Net loss-1,531-8,309-41,965-25,107-104,039
Stock-based compensation addback291921,1585843,255
Depreciation and amortization3214024250
Noncash financing fair-value adjustment1,0455,43434,30820,10794,324
Donation matching liability0054102,086
Other financial instruments000010,277
Other reconciliation07114-11
Receivables movement-10-90-834-500-7,076
Prepaids / other assets movement-1,253-3,1901,7717791,435
AP / accruals / other liabilities movement842222,0791,8674,377
Lease liability movement-101324-11
Deferred revenue movement7212411941,702
Operating cash flow-1,627-5,692-2,637-2,0246,569
Owned equipment purchases-2-15-131-111-925
Leased-asset / equipment prepayments00-54-51-1,955
Short-term investment purchases00-3,082-1,831-40,318
Investment sales / maturities0021396,900
Other investing0-7-23-23-248
Investing cash flow-2-22-3,077-2,007-36,546
Preferred stock proceeds, net4451,18816,54816,548110,432
Finance-lease principal0000-88
Convertible notes cash proceeds1,7506,0503,4502,0500
Common stock cash proceeds31011159407
Common stock repurchases00-149-1490
SAFE proceeds1220000
Offering costs paid0000-16
Financing cash flow2,3207,24819,96018,508110,735
FX effect0011-7
Net cash increase6911,53414,24714,47880,751
Beginning cash including restricted9291,6203,1543,15417,401
Ending cash including restricted1,6203,15417,40117,63298,152

* Unaudited. All periods reconcile operating + investing + financing + FX to the change in cash, including restricted cash. Swipe the table to view all columns.

Source: November 17 amended prospectus · Cash Flows F-12–F-14

Major shareholders and control

Economic ownership and voting power are different

The table below preserves the November 17 prospectus’s preliminary ownership basis, including its US$74 assumed price and smaller offering. It is not a final US$83 closing ownership register. Class N percentages are within that class and are not total-company percentages. The Founder LLC and Trust control rights must be read separately from economic shares. No preliminary percentage is silently restated as a final diluted ownership figure.

Evidence: November 17 amended prospectus · Principal and Selling Stockholders, p.300–301; Offering Summary

Disclosed ownership and control reference
Holder / classDisclosed referenceInterpretation
Dario Amodei · Class A245,259,100 before offer, including proxiesPreliminary table; 3.7% of Class A, not total economic ownership
Daniela Amodei · Class A241,408,940 before offerPreliminary table; 3.6% of Class A
Officers / directors group · Class A496,360,810 before offerIncludes individual rows; do not add again
Amazon affiliates · Class N1,920,511,218 preliminary post-offer33.9% of Class N / 14.7% of then-capital; conversion cap affects timing
Google · Class N1,789,598,269 preliminary post-offer31.6% of Class N / 13.7% of then-capital
Founder LLC · Class FOne control share; 50.1% general voting powerNon-economic control; preliminary post-offer structure
Long-Term Benefit Trust · Class TOne control share; board-majority authorityNon-economic rights; not a public economic stake

Share amounts are actual shares, not millions. Preliminary beneficial ownership uses SEC-specific denominators and 60-day rules; final priced/closing reconciliation is pending. Holdings and control rights are not additive. Swipe the table to view all columns.

Source: November 17 amended prospectus · Principal and Selling Stockholders, p.300–301; Offering Summary, Class N holdings

Independent valuation model

Ten annual nominal USD forecasts derive revenue from indexed activity and realized pricing. Capacity cash varies with activity, utilization and capacity-unit cost; inference and training are separate. These are analyst hypotheses, not measured tasks, tokens, channels or utilization. Other operating cash, future-service award economics, taxes, owned capex and working capital are charged explicitly. Mature cash margin reconciles year-ten costs to maintenance capex; terminal growth receives a separate capital charge.

bear

Conditional case · no probability assigned

Current intrinsic
Not valued
12-month fundamental value
Not valued
Return from offer
Unavailable
USD required return
11.0%
Terminal growth
3.0%
Terminal share of operating value
Unavailable
Five-year annualized return
Unavailable

Adoption grows, but pricing declines 5% annually, utilization reaches only 70%, and capacity-unit cost falls 6% annually. These are analyst stress assumptions, not disclosed operating metrics.

No positive sustainable terminal cash flow. This model does not assign a price; it does not imply a zero trading price.

base

Conditional case · no probability assigned

Current intrinsic
US$55.64
12-month fundamental value
US$60.33
Return from offer
-27.5%
USD required return
11.0%
Terminal growth
3.0%
Terminal share of operating value
70.4%
Five-year annualized return
0.4%

Activity rises 6.6× from first-year index while realized yield declines 3% annually; utilization improves from 65% to 85%. Capacity-unit cost declines 6%, variable inference-unit cost 10%, and training grows 8% annually. All are unverified analyst hypotheses requiring operating evidence.

bull

Conditional case · no probability assigned

Current intrinsic
US$162.16
12-month fundamental value
US$178.56
Return from offer
114.7%
USD required return
11.0%
Terminal growth
3.0%
Terminal share of operating value
68.6%
Five-year annualized return
25.8%

Activity rises 11× and utilization reaches 90%, with stronger capacity and inference efficiency. First-year revenue is US$150B. Exceptional execution is required; no probability is assigned.

A terminal contribution above 100% means explicit forecast losses offset part of the discounted terminal value. These wide cases reflect substantial uncertainty, not calibrated probabilities.

All operating cases share an 11% USD required return and 3% terminal growth. A separate grid varies discounting from 9% to 13% and growth from 2% to 4%. These are analyst assumptions, not a measured IPO beta or uniquely inferred market expectations.

Cash and dilution bridge

Gross new primary proceeds
US$79.4B
Secondary proceeds to selling holders
US$2B
Modeled opening cash and investments
US$237.3B
Fully converted economic share estimate
18.20B
Nine-month cash after owned capex, prepayments and finance-lease principal
US$3.6B

The economic denominator includes gross existing options, unvested awards, matching obligations, acquisition issuances and holdback, initial benefit shares and full existing-note conversion less the conversion already included in pre-offer shares. Legal pro-forma shares and voting rights are distinct. Option proceeds are a valuation convention and are excluded from liquidity. Existing awards are represented in shares; prospective service awards are charged once as economic cost.

The US$70.7 billion contractual amount covers October 2026 through September 2027. It is not an exact minimum for a forecast year beginning November 18. The obligation register separates dated scheduled payments, excluded and subsequent contracts, and overlapping conditional subsets. Forecast compute cash includes hosting and lease service; the same payments are not deducted again as debt.

Base annual cash-flow forecast · USD billions
YearRevenueCompute cashOwned capexFree cash flowEconomic retained cash
2027110.080.07.7-7.5229.9
2028165.4108.311.6-0.1229.8
2029227.7136.315.910.2240.0
2030291.1160.720.424.1264.1
2031360.3184.225.241.3305.3
2032415.6200.129.158.1363.4
2033467.3212.432.775.3438.7
2034506.6218.135.591.6530.3
2035534.5220.737.4104.4634.7
2036551.9219.138.6114.9749.6

Free cash flow also deducts staff, research and distribution cash, future-service award costs, taxes and incremental working capital. This retained-cash path includes the economic award charge. Actual cash funding is analyzed separately above, with noncash awards excluded. Zero future partner-facility proceeds, buybacks or dividends are assumed.

Five-year investment assessment

Base conditional annualized return: 0.4% through November 18, 2031.

Conditional USD five-year stock returns use the US$83 offer as an entry reference, zero dividends and illustrative 10 basis points each-way trading costs. The IPO price is not a guaranteed retail fill or opening trade. Canadian investor FX and account-specific fees are unverified. The portfolio comparator remains MSCI ACWI IMI CAD net total return over rolling five-year periods; no benchmark forecast or alpha probability is supplied.

Required-return and terminal-growth sensitivity
Required returnTerminal growth12-month fundamental value
9.0%2.0%US$73.97
9.0%3.0%US$81.10
9.0%4.0%US$91.08
11.0%2.0%US$56.85
11.0%3.0%US$60.33
11.0%4.0%US$64.80
13.0%2.0%US$46.27
13.0%3.0%US$48.17
13.0%4.0%US$50.50

Assumptions and evidence limits

Final priced and closing capitalization remain unverified. September liquidity is not a November cash balance; zero intervening cash adjustment is an unresolved placeholder, tested with an illustrative US$25 billion adverse bridge. Fees and withholding remain estimates. Quarterly allocations are timing hypotheses, not disclosed calendar quarters or covenant compliance. The adverse operating case has no positive sustainable terminal cash and receives no numerical valuation. No current trading quote, calibrated probability, CAD FX return or benchmark forecast is provided.

Financial facts through 2026-09-30. Reviewed by an AI analyst; forecasts and valuation assumptions are Peralta estimates.

Primary sources

  • November 17 amended prospectus

    Capitalization pp. 130–136; notes pp. 173–179; financials F-5–F-13; awards and matching pp. 23–24. All share amounts are post-split.

  • November 18 final pricing announcement

    US$83; 980 million public shares, including 956.351354 million primary and 23.648646 million secondary; 147 million optional overallotment; expected November 19 trading and November 20 closing.