What will the AI economy look like? Anthropic sees US GDP soar 32% by 2030, but jobs crisis looms

Artificial intelligence could make the US economy dramatically richer by 2030, but the biggest gains may not necessarily translate into better pay or greater job security for the workers whose tasks AI replaces.

That is the central finding of a new economic scenario model developed by Anthropic’s Economics team. The model explores how different levels of AI capability, adoption, autonomy and productivity could reshape GDP, employment, wages and the division of economic gains between workers and capital.

Anthropic stresses that the scenarios are not forecasts. Instead, they are simulations designed to show how different assumptions about AI development could translate into different economic outcomes.

Three possible AI futures 

Anthropic's model divides the future into three broad scenarios: modest, substantial and extreme.

In the modest scenario, AI has an economic impact broadly comparable to the internet. Productivity improves and GDP rises, but the gains arrive gradually and remain within the historical range associated with major technological advances.

The substantial scenario represents a much bigger disruption. By 2030, AI is capable of performing around half of all knowledge-work tasks, with most of that work technically possible to perform autonomously. However, adoption remains incomplete.

Under this scenario, economic growth roughly doubles its normal rate. Knowledge workers see little wage growth, while workers in other occupations benefit from rising demand.

The extreme scenario is where the economic picture changes radically. AI becomes more productive than humans across most knowledge-work tasks, performs nearly all of them autonomously and creates few new knowledge-work tasks for people.

Anthropic says this outcome would likely require recursively self-improving AI and very rapid adoption.

Annual GDP growth could eventually reach 15%, causing the economy to double in size roughly every 4.5 years. But that extraordinary growth would come alongside unemployment rising beyond typical recessionary levels.

GDP could rise sharply 

The model estimates that AI increases US GDP under all three scenarios.

By 2030, compared with an economy without AI:

• Modest: GDP reaches about $34.1 trillion, 1.6% higher. 
• Substantial: GDP reaches about $36.3 trillion, 8.3% higher. 
• Extreme: GDP reaches about $44.4 trillion, 32.4% higher. 

The figures are calculated using 2025 price levels.

But Anthropic's model makes an important distinction: a larger economy does not automatically mean proportionally higher incomes for workers.

Knowledge workers face the biggest disruption 

The model treats jobs as collections of individual tasks rather than assuming that an entire occupation will simply disappear.

AI can leave a task unchanged, augment it, automate it or create new tasks around it. A nurse, for example, could use AI to monitor patients, prepare documentation or plan schedules while continuing to perform tasks requiring human involvement. At the same time, AI could create new responsibilities, such as reviewing AI-generated care plans.

This distinction is crucial because automation does not necessarily eliminate an entire job. It can instead change what workers spend their time doing.

However, in the more transformative scenarios, some occupations could lose large numbers of tasks to AI. Coders and customer-service workers, for instance, could face pressure to move into occupations less exposed to automation.

That transition could be difficult because changing careers requires retraining and finding an entirely new job.

Wages may not rise equally 

Anthropic finds that average wages increase across the scenarios, but the gains are uneven. In the substantial scenario, wages for knowledge workers are essentially flat. In the extreme scenario, their wages fall by more than 10% by 2030.

Meanwhile, wages in occupations less exposed to AI could rise as productivity improvements generate greater demand for physical and other human-intensive work. For example, faster AI-assisted design and permitting could increase construction activity, boosting demand for construction workers.

This creates an unusual possibility: AI could simultaneously make the economy much richer while making some highly skilled workers economically worse off.

Bigger issue: who owns the gains? 

Perhaps the most consequential finding concerns the split between labour and capital.

Anthropic's model starts from an economy where roughly 60% of output goes to labour and 40% to capital. As AI automates more tasks, the share going to capital increases.

1. In the modest scenario, labour receives 59.4% of GDP and capital 40.6%. 
2. In the substantial scenario, the labour share falls to 56.1%, while capital rises to 43.9%. 
3. In the extreme scenario, the split changes dramatically: only 45.2% goes to labour, while capital captures 54.8%. 

That means companies and individuals owning the AI systems, computing infrastructure and other productive capital could capture a growing portion of the economic gains.

What do Americans expect? 

Anthropic also surveyed more than 10,000 Americans about AI's future capabilities, adoption, productivity and the time workers might need to find new jobs.

The typical respondent's answers were broadly consistent with the substantial-change scenario. Their expectations imply an economy roughly 10% larger by 2030 than it would otherwise have been, with unemployment rising to around 5%.

Around 10% of respondents gave answers consistent with the extreme scenario.

Anthropic emphasises that its model is deliberately simplified. It does not incorporate several potentially important factors, including policy responses, business cycles, aggregate-demand effects, financial-market disruptions, catastrophic risks or a future involving highly capable robots.