AN INTERACTIVE ECONOMIC THOUGHT EXPERIMENT
Who captures the AI economy?
Frontier intelligence, personal AI or an enterprise ecosystem? Explore how alternative futures could change Anthropic's 2036 profits—and what those outcomes would imply for shareholders.
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.
Frontier concentration
A few frontier labs retain pricing power
02 / FOLLOW THE MONEY
From global production to company profit
Every arrow represents a different modelling assumption. Large GDP or labour exposure does not automatically become Anthropic revenue.
03 / REVENUE SPLIT
Who captures AI spending?
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?
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 scenario | AI service spend / year | Anthropic revenue | Net profit | 2036 modelled price | 10-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.
Capital-market assumptions
ALL EDITABLEAlternative cash-flow test +
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 targetNo 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 $100 billion in perspective.
Compare hypothetical Anthropic profit in 2036 with actual reported net income of major global companies. These are different calendar years, displayed deliberately as a scale comparison—not a valuation ranking.
Annual net income · US$B
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]
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.
| Company | FY2015 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
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.
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.
Model equation
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 assumptionsShareholder 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 ≠ DCFWhat 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 forecastingHow 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
20 documented referencesImportant 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.