Anthropic Model Sees Labor Share at 45.2% in Extreme AI Case

Anthropic's Economics team put out a scenario model this week that tries to bound what transformative AI could do to the US economy by 2030, and one figure carries the whole thing: in the extreme case, labor's share of income lands at 45.2%.

That is the tail scenario. The full model in Anthropic's post runs three of them. In the modest case, AI acts roughly like the internet: GDP grows 1.6% to $34.1 trillion by 2030 and labor's share holds near 59.4%. In the substantial case, "AI capability advances to handle half of knowledge work by 2030, primarily through autonomous systems"; GDP climbs 8.3% to $36.3 trillion, and knowledge-worker wages go flat while other workers see gains. In the extreme case — AI outperforms humans at most knowledge tasks, adoption is rapid — annual GDP growth hits 15%, the economy doubles every 4.5 years to $44.4 trillion, knowledge-worker wages decline over 10%, and "unemployment rises beyond typical recessionary levels."

The technical report is credited to Anton Korinek, Chad Jones, Szymon Sacher, Tess Cotter and Peter McCrory, with external review from Daron Acemoglu, David Autor, Ben Jones and Pete Klenow.

Anthropic paired the model with a Morning Consult survey of 10,980 US adults. "The typical respondent's predictions aligned with the 'substantial change' scenario," implying GDP roughly 10% higher by 2030 and unemployment around 5%. About 10% of respondents endorsed views consistent with the extreme case.

The company is upfront about what the framework leaves out: policy responses, business cycles, catastrophic risks and, notably, scenarios involving hyper-capable robots. Individual workers are not tracked. In the substantial scenario, Anthropic reports 59.7% of knowledge workers remain in their roles by 2030, 39.6% shift to other sectors and 0.7% remain displaced; how frictionless those transitions are is not specified. Our jobs tracker has logged 114 stories in the last 90 days on how that reshuffling is already playing out.

It lands during a busy stretch of Anthropic coverage, including the company's own alignment lead publicly putting AI extinction odds over 10% the same day.