Anthropic maps 3 ways AI could upend the economy by 2030

While most AI labs pitch a flawless utopia, Anthropic just ran the economic math on what happens if their technology works a little too well.

Anthropic's economics team launched an interactive scenario explorer based on their latest research. It maps out how AI could radically reshape US jobs, wages, and GDP growth between now and 2030.

Three roads to 2030

Anthropic does not look at jobs as static titles. Instead, borrowing from the US Department of Labor's O*NET database, the model breaks work down into bundles of specific tasks.

Take a nurse. AI cannot bathe a patient. But it can automate vitals logging, augment shift scheduling, and create new responsibilities like auditing AI-generated care plans. Multiply those shifts across every sector, and you get the whole macro economy.

From there, Anthropic modeled three distinct futures.

In the modest scenario, AI looks like the early internet. It fuels real productivity gains, but the changes arrive gradually, leaving unemployment and wages well within historical norms.

In the substantial scenario, AI can autonomously handle half of all knowledge work by 2030. GDP growth doubles its historical rate. Knowledge worker pay stalls while blue-collar wages climb. Interestingly, an August survey of over 10,000 Americans showed that typical public expectations align right here.

Then comes the extreme scenario. Driven by recursively self-improving AI, the technology outpaces humans at virtually all knowledge work without creating new desk jobs. US GDP growth rockets to 15% a year, doubling the entire economy every 4.5 years.

More wealth, less for workers

That breakneck growth comes with a brutal hangover for white-collar staff.

In the extreme case, knowledge worker unemployment spikes past normal recession territory. Wages for knowledge workers drop by more than 10%. Coders and customer service agents are forced to retrain as electricians or nurses, facing long and painful career transitions.

Even more telling is who pockets the money.

Right now, for every dollar the US economy generates, roughly 60 cents goes to workers and 40 cents goes to capital. Under heavy automation, capital becomes vastly more productive and valuable. Capital owners capture almost all the newly created upside.

Basically, the country ends up wealthier than ever, but total labor income across the US barely budges by 2030.

Why it matters

This explorer pops the feel-good bubble that technology always creates just as many good jobs as it destroys.

For knowledge workers, the takeaway is stark. If recursive AI takes off, relying purely on digital output will leave you vulnerable to severe wage stagnation. The safe ground shifts toward roles tethered to physical reality and human coordination. Waiting around to see whether your field gets hit is a losing bet.

On the industry side, Anthropic is playing smart politics. It is rare for a leading model builder to explicitly warn that its core product could trigger historic white-collar unemployment and supercharge wealth inequality. By publishing the simulator, Anthropic cements its posture as the responsible adult in the room while preparing regulators for aggressive safety nets.

That said, we think the extreme scenario leans heavily on perfect execution. It assumes computing power, energy grids, and recursive model improvements will face zero hardware bottlenecks before 2030. Still, even if reality lands closer to the middle path, the warning stands: when growth decouples from human labor, our current economic rulebook simply breaks down.