Anthropic opens an economic scenario explorer: plug in AI capability forecasts, see 2030 GDP and unemployment
Synopsis
Anthropic's economics team released an economic scenario explorer that represents the US economy as bundles of tasks and lets users set expectations for AI capability, adoption and autonomy to see possible 2030 outcomes for GDP, occupational switching, wages and the labor-versus-capital split. The three scenarios run from internet-scale modest effects to an extreme case with GDP 32.4% above a no-AI baseline at $44.4 trillion; there, knowledge-worker wages fall more than 10% and unemployment rises beyond typical recessionary levels. The model excludes policy responses, business cycles and hyper-capable robots.
Interpretation
Anthropic's economics team launched an economic scenario explorer that turns guesses about AI's future into an adjustable picture of the 2030 US economy, covering GDP, employment, wages and the split between labor and capital. The earlier Economic Index measured how AI is used today; this tool looks ahead and lets users set AI capability and adoption speed themselves. The three scenarios differ in scale: the modest case puts 2030 GDP at $34.1 trillion, 1.6% above no AI; the substantial case at $36.3 trillion, 8.3% higher; the extreme case at $44.4 trillion, 32.4% higher, all at 2025 price levels.
The model represents all work in the country as bundles of tasks, and AI can leave a task unchanged, augment it, automate it or create a new one; each task happens millions of times a day nationwide, and their sum makes up the economy. Task lists come from the US Department of Labor's O*NET occupation taxonomy, illustrated with a nurse's day showing which tasks only humans can do, which are augmented and which may be automated. The path from tasks to macroeconomic outcomes depends on three adjustable quantities: how much augmentation and automation AI takes on, how much more productive it makes people, and how fast workers and firms adopt it.
In the extreme scenario growth decouples from distribution: the economy is larger than ever, yet knowledge-worker wages fall more than 10% by 2030, unemployment rises beyond typical recessionary levels, and the labor share drops from about 60 cents to 45.2%. In the modest and substantial scenarios unemployment stays within historical ranges and wages are flat or rise by industry; adverse outcomes for knowledge workers appear only where growth outpaces anything in economic history. The labor share is 59.4%, 56.1% and 45.2% across the three scenarios, with the capital share rising correspondingly to 40.6%, 43.9% and 54.8%.
Perspective
The tool is aimed at researchers, policymakers and the public who want to discuss AI's macroeconomic consequences, and it outputs scenarios for the US in 2030 under given capability and adoption assumptions rather than predictions. The model excludes hyper-capable robots as well as policy responses, business cycles, aggregate demand and financial-market disruptions.
The authors state that the model does not follow individual workers, so it can only paint a coarse picture of displacement costs; some reviewers questioned whether AI-exposed occupations will shrink rather than grow, some read the most extreme scenario as a thought experiment, and others felt the most modest one understates what is already visible in the data. The model also omits aggregate-demand effects from the data center buildout, and the authors acknowledge it may underestimate how much AI accelerates technological progress itself.
