A New Model Puts American GDP Up 32 Per Cent by 2030 and One in Five Cognitive Workers Out of Work

Researchers at the Anthropic Institute have modelled three scenarios for the American economy through 2030, running from modest AI adoption to an extreme case in which the technology becomes more productive than people at most knowledge work.

They are emphatic that these are not predictions, and have assigned no probabilities to any of them.

The outcomes diverge sharply. Under the modest scenario, GDP in 2030 is 1.6 per cent higher than in an economy without AI, at 34.1 trillion dollars in 2025 money. The substantial scenario puts it 8.3 per cent higher, at 36.3 trillion. The extreme scenario lifts it 32.4 per cent, to 44.4 trillion.

What the extreme case assumes

The top-end number rests on three conditions holding at once.

AI performs nearly all knowledge-work tasks autonomously. Adoption is rapid. And essentially no new knowledge-work tasks are created for humans to move into.

Under those assumptions annual GDP growth reaches 15 per cent, which doubles the size of the economy roughly every four and a half years.

That third condition is the one carrying most of the weight, and it is the historically unusual one. Previous automation waves destroyed categories of work and generated new ones. This scenario assumes the second half does not happen.

Nearly one in five cognitive workers unemployed by 2030

Who pays for the growth

In the extreme scenario, wages for knowledge workers fall more than 10 per cent below where they would have been without AI by 2030. Pay rises substantially for workers in occupations less directly exposed.

The employment effect is larger. The underlying working paper, by Anton Korinek, Charles I. Jones, Szymon Sacher, Tess Cotter and Peter McCrory, estimates that nearly one in five cognitive workers would be unemployed by 2030 under those assumptions.

Cognitive occupations here mean management, professional, sales and office jobs. Displaced workers would have to move into fields less exposed to the technology, such as construction or electrical work, and the model accounts for the time and difficulty of making that switch rather than assuming it happens instantly.

The middle scenario is much calmer

Under the substantial case, AI becomes capable of doing half of knowledge work by 2030, but most of those tasks continue to be done without it because adoption remains incomplete.

Growth runs at about twice the normal rate. Knowledge workers' wages are essentially flat. Workers in other occupations gain.

That gap between capability and adoption is doing the protective work, and it is a reminder that the constraint on economic disruption has rarely been what a technology can do.

The share going to labour

The clearest single indicator in the model is how income splits.

Workers receive 59.4 per cent of income in the modest scenario, 56.1 per cent in the substantial one, and 45.2 per cent in the extreme case, where capital's share reaches 54.8 per cent.

Average wages still rise in all three, because the gains concentrate among workers outside knowledge-intensive occupations. In the extreme case, total labour income is barely changed by 2030 even as the economy becomes substantially larger.

That is the finding worth holding onto. A third more economy, and roughly the same amount of money reaching people who work for a living.

What the public expects

Anthropic also surveyed 10,980 American adults in August about AI capability, adoption, productivity, automation and how long displaced workers would take to find new jobs.

The median responses produced something close to the substantial scenario: GDP about 10 per cent higher by 2030 than it would otherwise be, and unemployment around 5 per cent.

About 10 per cent of respondents gave answers consistent with the extreme scenario. One in ten Americans, in other words, already expects the version where a fifth of office workers lose their jobs.