Anthropic releases another article: What will the economy look like in the AI era?

Will AI ultimately bring unprecedented economic growth, or trigger a massive wave of unemployment? Or will it do neither, perhaps leading us into a future we have yet to imagine?

The Anthropic economics team recently released a set of economic scenario models, attempting to quantify AI’s potential impact on employment, growth, and wages, and has made them publicly accessible as interactive tools, allowing everyone to preview the corresponding economic landscape based on their own expectations of AI capabilities.

The model is structured around three scenarios: from a moderately gradual “moderate scenario,” to a “substantial transformation scenario” where economic growth doubles, and finally to an “extreme scenario” in which annual GDP growth reaches 15% but many knowledge jobs disappear.

Research shows that in most scenarios, unemployment and wage volatility remain within historic norms; but in extreme scenarios, knowledge workers face pay cuts or even long-term unemployment, and the share of capital in the economy rises significantly.

Anthropic states that this model and broader research will guide its sponsored labor market studies and inform policy recommendations, with the goal of ensuring that the economic gains brought by AI can be widely shared throughout society.

The current version of the model is v1.0, released in September 2026. The researchers acknowledge several limitations in the model and plan to iteratively update it.

To understand how AI will impact the economy, we first need to break our conventional notions of “work.” In Anthropic’s model, all jobs in the economy are not singular but are composed of a series of “task bundles.”

Take for example a day in the life of a nurse: her job includes doing ward rounds, drawing blood, triaging patients, recording vital signs, ordering supplies for the ward, and more. As AI is introduced, how might this “task bundle” change?

• Some tasks can only be done by humans: such as bathing a patient—AI is powerless here.
• Some tasks are AI-enhanced: AI can help nurses draft discharge instructions more efficiently, remotely monitor patients, and optimize shift scheduling.
• Some tasks will be fully automated: for example, recording vital signs or ordering supplies can be completely handled by AI.
• Entirely new tasks may emerge: history shows that new technologies always create new jobs, such as future nurses reviewing AI’s triage results or assessing AI-generated care plans.

What’s the result? As AI integrates with work, a nurse’s productivity increases dramatically, allowing her to spend more time communicating with patients. When such “task-level transformation” happens millions of times every day across all sectors, it creates a massive economy with a future value of over $30 trillion.

GDP, the labor market, and the share of workers’ wages will all depend on the speed and scope with which AI enhances or automates tasks.

The future direction of the economy will depend on the speed of AI capability evolution and adoption across industries. Anthropic’s report outlines three distinctly different macro scenarios.

The first is a moderate scenario, with gradual gains.

In this scenario, the economic impact of AI is similar to the spread of the internet. It does drive actual economic growth, but the growth rate remains within the historical norm for new technologies, and changes occur incrementally. In macroeconomic data, it’s hard to immediately perceive AI’s dramatic impact.

The second is a significant scenario—a revolution in mental labor.

By 2030, AI will be able to perform half of knowledge work (most of it autonomously), but will not be fully adopted everywhere. Here, the economic growth rate doubles the normal level. Notably: knowledge workers’ wages stagnate, while other types of workers benefit.

In an August survey of over 10,000 Americans, Anthropic found that most people’s expectations most closely match this scenario, which means that by 2030, GDP will be 10% higher than in a world without AI, and the overall unemployment rate will rise to about 5%.

The third is the extreme scenario featuring profound economic reshaping.

In this scenario, AI outperforms humans on the vast majority of knowledge work tasks, completing nearly all work autonomously, and essentially creates no new knowledgeable tasks for humans. This typically requires AI to have “recursive self-improvement” capabilities and to be quickly adopted.

The result is staggering: annual GDP growth reaches 15%, with the economy doubling every 4.5 years. As a society, we would be richer than ever before; but at the cost of a dramatic reduction in knowledge jobs, with unemployment surging to levels exceeding typical economic recessions.

Based on this model, Anthropic arrived at four key findings that should provoke deep reflection for every worker.

Finding One: The economic pie gets bigger, but may come with historic unemployment.

In all scenarios, AI drives GDP growth and society’s overall wealth rises sharply. But in the extreme scenario, recursive AI self-improvement and rapid adoption could push unemployment to historic highs.

Finding Two: Painful career reshuffling.

In both substantial and extreme scenarios, mental laborers (knowledge workers) face severe automation-driven replacement. At the individual level, programmers and call center operators may be forced to switch to AI-resilient jobs like electricians or nurses.

However, cross-industry transitions are extremely difficult: workers may be unwilling to change, need to learn new skills, and the new jobs may not be easy to find. The more transitions required, the more people get trapped in prolonged unemployment.

Finding Three: Wage polarization, with knowledge workers facing pay cuts.

According to the model, although average wages rise across society, this is mainly concentrated in non-knowledge jobs. As the demand for human mental labor declines, knowledge workers will face downward wage pressure.

Conversely, AI improves the efficiency of mental labor, for instance by speeding up the design and approval for physical infrastructure, boosting demand for blue-collar jobs in construction and therefore raising those wages.

In the substantive scenario, knowledge workers’ wages are flat; in the extreme scenario, knowledge worker pay falls more than 10% by 2030.

Finding Four: Capital wins, labor’s share shrinks.

Today, for every $1 of economic output, about 60 cents go to labor and 40 cents to capital. But as AI takes over more tasks, the technology and resources capital uses to create wealth becomes more valuable and in higher demand, driving prices up.

Research finds that labor’s share falls significantly in the substantial and extreme scenarios while capital’s share rises. In the extreme scenario, even as the economy balloons, the total income for labor changes little by 2030.

The vast majority of knowledge workers will face pay cuts or unemployment, with the working class receiving a smaller slice of a much bigger economic pie.

In the face of such a rather grim forecast, Anthropic notes that the economic picture of 2030 is not an unavoidable destiny. It depends on what AI is actually capable of, how businesses and workers leverage it, and most importantly, how the financial gains from this technology are distributed.

The Anthropic team admits, as a v1.0 model, it greatly simplifies complex realities: for instance, it does not consider policy responses, economic cycles, or the emergence of superintelligent robots.

Yet the core value of this explorer is as a wake-up call: under an extreme AI boom, our primary challenge will not be how to achieve economic growth, but how to ensure that the dividends are broadly shared, rather than the costs disproportionately borne by certain groups, especially knowledge workers.

The AI age is already upon us. It’s not just about how fast the train is running, but whether everyone on board can have a seat of their own.