A•AI ECONOMY LABINDEPENDENT SCENARIO RESEARCH

ANTHROPIC · AN INTERACTIVE ECONOMIC THOUGHT EXPERIMENT

What Would Justify a ~$1.4 Trillion Valuation?

Explore the global AI economy, the profits Anthropic would need to generate by 2036, and whether alternative industry futures could support such an extraordinary valuation.

The ~$1.4 trillion figure is illustrative: $78 per share × approximately 18.013 billion diluted-equivalent shares ≈ $1.405 trillion. The share count includes conditional and future awards; this is not an observed market capitalisation. See methodology.

2026 → 2036USD · nominalAssumptions, not forecasts

THE CENTRAL DISTINCTION

“AI can create extraordinary economic value without any single model provider capturing extraordinary profits.”

Test the distribution of value, not just the growth of AI.

VERIFIED BASELINE
2026 world GDP$128T
IMF 2026 [S-03]
2025 labour income / GDP52.6%
ILO [S-05]
IPO indicative range$70–$78
SEC filing [S-01]
Compute commitments$900.9B
Multi-year [S-01]
LIVE ANALYSIS● Recalculates instantly

Frontier concentration

A few frontier labs retain pricing power

ILLUSTRATIVE
2036 world GDP [S-03]
$189.5T
2026 baseline × nominal growth
Annual end-customer AI spending MODEL
$2.39T
Knowledge-work task market
Anthropic 2036 net profit MODEL
$119.6B
Revenue × long-run net margin
Share of global corporate profits MODEL
0.63%
Depends on global profit-pool assumption

02 / FOLLOW THE MONEY

From global production to company profit

2036 · USD NOMINAL

Every arrow represents a different modelling assumption. Large GDP or labour exposure does not automatically become Anthropic revenue.

Reported anchor User-adjustable assumptionUS dollars, annual flows—not stock market values

03 / REVENUE SPLIT

Who captures AI spending?

$478BAnthropic external revenue
$1.91TOther AI providers / channels

This is a division of end-customer AI service spending. Reseller pass-throughs and internal purchases are not counted twice.

04 / MACRO CONSISTENCY

How large is too large?

2036 global profit pool · illustrative$18.95T
Bar scale: 0–5% of the hypothetical global corporate-profit pool.
Anthropic share0.63%
Net profit / world GDP0.063%

A plausibility lens, not a hard cap. Company GAAP profit and national-accounts profits are not perfectly comparable. Methodology ↗

SCENARIO CONTRAST

Three ways the future could unfold.

Illustrative operating scenarios, calculated using the same current entry-price, terminal-multiple and dilution settings for a fairer comparison. No probabilities are assigned.

Economic scenarioAI service spend / yearAnthropic revenueNet profit2036 modelled price10-year return / year

Numbers are conditional arithmetic, not price targets. “Distributed” and “ecosystem” scenarios use different explicit assumptions about task coverage, revenue capture and divisional margins. See the complete methods and source register.

05 / INVESTMENT HURDLES

What must shareholders earn?

A separate investor-return calculation connects your operating scenario to a hypothetical 2036 stock price. It is a terminal multiple exercise, not a discounted cash-flow valuation.

2036 / 10 YEARS

Capital-market assumptions

ALL EDITABLE
$74
$70–$78 was the preliminary range, not a guaranteed final price. [S-01]
Share-count basis

18.013bn indicative diluted-equivalent shares, reconstructed from S-1 p. 25; includes contingent awards.

10%
25×
1.0%
Additional post-IPO share growth, applied for 10 years. Not company guidance.
Alternative cash-flow test +
80%
25×

FCF is a *proxy* applied to scenario profit; not a model of capex, lease obligations or working capital. See limitations.

REQUIRED 2036 ANNUAL PROFIT

$152.8Bto meet your return target
Below hurdle
Entry market capitalisation$1.33T
Modelled 2036 equity value$2.99T
Modelled 2036 share price$150
Modelled 10-year annual return7.3%
Implied annual profit vs required profitModel: 78% of hurdle
Scenario
$120B
Required
$153B
Cash-flow sensitivity (P/FCF framework)
$95.7BFCF proxy5.0%Annual return

No dividends, FX movements, taxes or interim cash flows. IPO count is indicative. Changes in earnings, P/E, dilution, and actual cash investment can materially alter returns.

06 / ECONOMIC SCALE

Put $100 billion in perspective.

Compare hypothetical Anthropic profit in 2036 with actual reported net income of major global companies. These are different calendar years, displayed deliberately as a scale comparison—not a valuation ranking.

PRIMARY-FILING DATA

Annual net income · US$B

Anthropic scenario Reported historical
Alphabet's 2025 net income includes substantial non-operating gains. Nvidia's fiscal 2026 ended January 2026. See company source links and methodology.

07 / THE GROWTH TRAP

What if Big Tech keeps compounding?

A historical-earnings stress test, not a forecast. Apply the minimum, median or arithmetic mean of six companies’ FY2015–FY2025 annualised net-profit growth rates to their FY2025 earnings base, then compare their combined earnings with the global economy. Apple’s original filing [S-33] · Nvidia’s [S-37]

HISTORICAL GAAP DATA
Projection year
Minimum peer CAGR—Slowest of selected peers
Median peer CAGR—Midpoint of selected peers
Mean peer CAGR—Highly influenced by outliers

Rates from all six peers. These statistics refer to six company-level 10-year CAGRs—not the minimum, average or median of annual year-on-year growth rates. Amazon and Nvidia began with very small profits. A high historical CAGR is not a sustainable industry-wide forecast.

CompanyFY2015
profit
FY2025
profit
2015–25
CAGR
At minimum
peer CAGR
At median
peer CAGR
At mean
peer CAGR
Own CAGR
capped at 10%
Own past
CAGR repeated

All projections are hypothetical annual net income for 2036, compounding FY2025 reported profits over 11 years. “Own CAGR capped at 10%” uses the lesser of each firm’s historical CAGR and 10%, an illustrative downside-to-growth sensitivity; “own past CAGR repeated” intentionally illustrates unreasonable extrapolation. Reported input sources are linked in the source register [S-08–S-13, S-33–S-40]. Tesla is excluded because its 2015 profit was negative; Aramco because a directly comparable public 2015 earnings base was not available.

Macro plausibility of the median case

—
Share of projected world GDP—
Share of projected global corporate profit pool—

World GDP extrapolates the 2026 IMF baseline using the current GDP-growth slider. The corporate-profit share is another illustration, not an observed worldwide statistic. These are scale indicators, not accounting identities.

Compare with Anthropic's required profit

Interpretation: today's large tech firms could grow rapidly while Anthropic prospers too—but the combined claim on future profits must be consistent with the economic structure. Double-counted investment gains and supplier-customer relationships are a material caveat.

Why these three futures are plausible Explore the academic, technological and industry evidence for each scenario, including counterarguments.

Research supports the plausibility of these different industry structures—not the numerical assumptions or probabilities of the scenarios.

01 / DISTRIBUTED INTELLIGENCE

Competition disperses the value of AI

Open-weight models, improving smaller models and multi-provider access may lower switching costs and limit the economic rents earned by any single frontier lab.

CounterargumentOpen-weight models do not always match the best closed models: the 2026 Stanford AI Index reports that the gap reopened in 2025. [S-44]
02 / FRONTIER CONCENTRATION

Scale and scarce inputs protect a few labs

High training costs, compute requirements, talent and economies of scale may make sustained frontier development accessible mainly to well-capitalised organisations.

CounterargumentConcentration need not mean monopoly profits: leading labs can still compete intensely on quality and price, as documented by Demirer et al. (2026). [S-41]
03 / AI ECOSYSTEM

Agents and workflows capture downstream value

Model developers may earn more by owning applications, integrations and enterprise workflows—but cloud and software incumbents can compete for the same customer relationships.

CounterargumentAnthropic may not control the downstream layer. Salesforce's Agentforce Operations shows that established software providers are pursuing similar workflow positions. [S-50]

Interpretation: These references support the plausibility of each economic mechanism, not the preset's exact 2036 adoption, spending, Anthropic market share, net margin or valuation. Company announcements are evidence of commercial strategies, not independent proof of adoption, financial returns or sustained competitive advantage. No scenario probabilities are assigned.

08 / THE AUDIT TRAIL

Every assumption is visible.

Reported observations are linked to original filings or official institutions. All forward-looking behavioural assumptions are explicitly labelled as author-defined thought experiments.

REPRODUCIBLE MODEL
01

Economic value-capture framework

World GDP × labour share × knowledge-work share × AI-addressed task share × AI spending rate × Anthropic retained share × blended net margin (model and application businesses).

Annual profit = annual flow × six assumptions
02

Shareholder hurdle

Entry share price × 2036 share count × (1 + required return)10 ÷ terminal P/E. The 2036 share count grows with assumed annual dilution.

Terminal P/E analysis ≠ DCF
03

What it does not model

Competition dynamics, capital availability, detailed compute cash outflows, net debt, cross-border tax, contract timing, risk-adjusted discount rates or interim distributions.

Use for reasoning, not forecasting

How assumptions are classified

Reported: IPO disclosures, IMF GDP, ILO labour share, corporate annual net income. Derived: arithmetic from reported numbers. Illustrative: 2036 GDP growth, task share, monetisation, market capture, margins, P/E, dilution and global corporate profit share.

Sources and data notes

50 documented references
Important limitations and correct interpretation Read before drawing conclusions +

Not a probability model. The scenarios have no assigned likelihoods. Outputs are conditional numerical consequences of stated assumptions. A target return hurdle is not an estimate of fair value.

Wage bill is not AI revenue. The model does not assume that automated tasks disappear or that firms will pay model providers the wages they save. Revenue arises only after an independent spending-rate assumption.

Economic production vs corporate earnings. GDP is annual value added. Global profit share is a user-set thought-experiment parameter; US BEA national-accounts corporate profits are not the global equivalent of listed firms' GAAP net income.

Share base and dilution. Basic post-IPO economic shares approximate 13.087bn. The diluted-equivalent share calculation adds the filing's options (treasury method), RSUs, rollover-note shares, acquisitions, donation-matching obligation and public-benefit issuance: approximately 18.013bn at the midpoint $74 assumption. The count includes conditional or future awards and can change. Share price sliders do not recalculate the contractual financing/treasury-method share count. A separate annual dilution slider approximates subsequent additional issuance. Both are simplified.

Capital commitments matter. Anthropic's ~$900.9bn aggregate future commitments span multiple years; they are not all debt, nor should they be subtracted as an immediate lump sum from equity value. Net margin implicitly assumes the eventual costs of running the business, but does not model each lease or purchase obligation.

Fiscal-year differences. The comparator chart uses FY2025 GAAP/IFRS net income except Nvidia FY2026 (ended January 2026). Some net income includes non-operating investment gains, impairments or one-off taxes. Do not read the bars as sustainable annual operating cash flows. The chart is a scale benchmark, not proof of attainability.

Public documentation. Sources were checked 17 November 2026. Market prices, offering details and company filings can change. Edit the source data as new filings appear. No personal financial information is requested or transmitted.