Anthropic's extreme 2030 scenario sends the whole growth increment to capital

Two uplift percentages give you the counterfactual for free. Divide 36.3 by 1.083 and you get 33.52. Divide 44.4 by 1.324 and you get 33.53 ^[20]. Both scenarios sit on one no-AI 2030 economy of about USD 33.5 trillion, which means the two GDP figures can be compared against each other and against today's division of output.

That division is where the report gets interesting. Anthropic puts the current split at roughly 60 per cent of output to workers and 40 per cent to capital ^[12]. Sixty per cent of 33.5 trillion is 20.1 trillion. In the extreme scenario labour takes 45.2 per cent of 44.4 trillion, which is also 20.1 trillion ^[21]. Capital goes from 13.4 trillion to 24.3 trillion, and all 10.9 trillion of the extra output lands on that side ^[22]. The condition is that the no-AI 2030 baseline still splits 60/40; Anthropic gives the 60/40 as a current figure, not a projected one ^[12].

One number inside that scenario does not reconcile. Fifteen per cent compound growth doubles output in a shade under five years, and a 4.5-year doubling needs about 16.6 per cent ^[23]. So the 15 per cent ^[8] is a floor or a rounding, not the rate that produces the doubling time. A small discrepancy, and the kind you want closed before the figure reaches a board deck.

The unemployment result is a friction estimate rather than a capability estimate. Anthropic's stated mechanism is that displaced knowledge workers struggle to change occupation, because a switch means acquiring new skills and then finding suitable work ^[14]. The Decoder's summary makes the move concrete: programmers and call centre staff heading for electrician or nurse jobs ^[15]. Two conditions have to hold before 17.9 per cent ^[11] describes your headcount. Knowledge-work automation has to convert into job loss at close to the modelled rate, and no newly created occupation can absorb the displaced inside the window.

Neither report shows Anthropic attaching a probability to any of the three paths ^[25]. The labels modest, substantial and extreme carry the ranking instead, and "extreme" is doing quiet work in a document from a company whose chief executive told the public in May 2025 to expect unemployment of 10 to 20 per cent ^[18].

Anthropic's own summary is that the future is not predetermined, and that the outcome turns on capabilities, on adoption decisions by companies and workers, and on how the financial benefits get distributed ^[16]. It adds that in the most transformative case the hard part is sharing the gains rather than generating them ^[17]. The distribution arithmetic puts that in a currency you can check: output up by a third, workers' total take roughly where it started.