Anthropic economists sketch three AI futures for 2030

Anthropic's Economics team has built a model of the US economy in 2030, and you get to turn the dials yourself: the announcement on Threads links to an interactive page with three possible futures and room to map your own predictions. The model covers how AI might change jobs, growth and unemployment by the end of the decade.

At a glance

• The release is an interactive explorer at anthropic.com/institute/econ-scenarios, where three AI scenarios for the US economy sit side by side and readers can plot their own forecast against them.
• According to Economic Insider, 2030 output ranges from $34.1 trillion to $44.4 trillion in 2025 dollars, while labor's share of GDP falls from 59.4% in the modest scenario to 45.2% in the extreme one.
• The Threads announcement itself gives no figures, and according to Economic Insider, Anthropic frames the work as scenario planning rather than a forecast, so every outcome shifts with the assumptions behind it.

If you have not been following, the model arrives after a warning from Anthropic's own CEO. According to Fortune, in May 2025 Dario Amodei told Axios that AI could eliminate half of all entry-level white-collar jobs within five years, pushing unemployment to between 10% and 20%. Fortune also reported that he saw time to mitigate the worst case through public awareness, helping workers use AI, and policy, quoting him:

You can't just step in front of the train and stop it... The only move that's going to work is steering the train.

Three futures for jobs, growth and unemployment, plus a page for your own forecast

Anthropic's Economics team modelled three scenarios for how AI could reshape the US economy by 2030, and according to qz.com the interactive version was released on a Wednesday. The Threads post that announced it carries no numbers for any scenario. The figures live inside the tool and in coverage of it.

According to Economic Insider, Anthropic presents the explorer as scenario planning, not a forecast. Outcomes move with assumptions about AI capability, adoption, autonomy, productivity, and how quickly workers can move between occupations. The same outlet reports that Anthropic surveyed 10,980 Americans in August, and the typical respondent's answers implied an economy about 10% larger by 2030 than without AI, with unemployment around 5%.

About 10% of those respondents, per Economic Insider, gave answers broadly aligned with the extreme scenario. The outlet also cites Stanford payroll research alongside the model, which finds no economy-wide AI job collapse but a widening employment gap for workers aged 22 to 25 in highly exposed occupations.

How much bigger does the US economy get by 2030?

That depends on the scenario, and the spread is wide. According to Economic Insider, Anthropic puts 2030 output at $34.1 trillion in 2025 dollars in the modest scenario, $36.3 trillion in the substantial scenario and $44.4 trillion in the extreme scenario. Against comparable economies without AI, those figures are 1.6%, 8.3% and 32.4% higher respectively.

A second line in the model moves the other way. The same outlet reports that labor receives 59.4% of GDP in the modest scenario, 56.1% in the substantial scenario and 45.2% in the extreme scenario. Anthropic's explanation, as Economic Insider relays it, is that capital becomes more important as AI performs a larger share of tasks, so labor's slice can shrink even while total output grows.

Why do nurses and electricians show up in an AI model?

They show up because, according to Economic Insider, the model does not treat a job as one solid block. It treats occupations as collections of tasks. AI can assist with some of them and automate others, some stay mainly human, and new tasks can emerge as technology changes how work is organized.

Think of a job as a toolbox rather than a single tool. If AI picks up a good share of the tools in a programmer's box, the question becomes what that person does next. That is why, per the same report, worker mobility becomes a major variable in the model.

Anthropic's own examples, according to Economic Insider, include programmers and customer-service workers moving into less AI-exposed occupations such as nursing and electrical work. Those moves may require training, relevant experience or longer job searches, especially if many workers change fields at the same time.

Every figure worth quoting comes from one outlet's reading of the tool. The Threads post publishes none, and the coverage does not spell out what values each scenario assumes for capability, adoption or autonomy. In our view, the labor-share line is the more telling design choice: an economy 32.4% larger in which labor's share drops to 45.2% looks like a very different 2030 from the one the growth figure alone appears to describe.

Where your 2030 guess lands

The real checkpoint is the horizon itself. Amodei's five-year warning from May 2025 runs out around 2030, the year the scenarios target, so his 10–20% unemployment range and the survey's roughly 5% will eventually face the same data. No schedule for updating the model with new numbers has been given. Until then, the explorer is a place to write down your own bet.