Artificial intelligence could add trillions of dollars to the US economy by 2030. But Anthropic’s latest economic model shows that the biggest gains could come with a serious cost for workers.
Anthropic’s Economics team has released a new technical report called “Economic Scenarios for Transformative AI” alongside an interactive Econ Scenario Explorer. The model examines how different levels of AI capability and adoption could affect US GDP, employment, wages and the distribution of income by 2030.
The most dramatic scenario puts US GDP at $44.4 trillion in 2030, 32.4% above the model’s no-AI baseline.
That figure sounds like a prediction of an enormous AI boom. Anthropic says that would be the wrong way to read the result.
The company presents three possible scenarios based on different assumptions about how capable AI becomes, how quickly businesses adopt it and how much work AI performs without human involvement. The extreme case requires conditions that would represent a major shift from today’s economy.
The catch is clear. The larger the economic gains become, the greater the pressure on some knowledge workers.
Anthropic’s AI Economic Model Looks at 2030
Anthropic released the Economic Scenario Explorer in September 2026 as a tool for examining possible economic outcomes from advances in AI.
The model treats the economy as a collection of tasks rather than assuming that entire occupations disappear overnight. A job contains many different tasks, and AI could affect each task in different ways.
AI could help a worker complete a task faster. It could automate a task completely. Some tasks could remain unchanged. New tasks could also appear because of AI.
Anthropic says its model uses task information based on the US Department of Labor’s O*NET occupational taxonomy. The approach is designed to examine how changes at the task level could affect the wider economy.
That distinction matters because a job title does not tell the whole story.
A software developer, accountant, nurse or manager performs many tasks. AI could handle some of them while leaving others to humans. In some cases, AI could also create new responsibilities for workers.
The model then connects those changes to GDP, wages, employment and unemployment.
Anthropic Says These Are Scenarios, Not Predictions
This is one of the most important points in the report.
Anthropic does not assign probabilities to the three scenarios. The company says the model is designed to show what the economy might look like if particular assumptions about AI development and adoption become true.
The scenario explorer itself describes the model as a simplification of a much more complicated economy. It leaves out several factors, including policy responses, business cycles, financial market disruptions and some potential risks from advanced AI.
That means the $44.4 trillion figure should not be read as Anthropic saying the US economy will reach that size.
Instead, the figure answers a conditional question.
What happens if AI becomes highly capable, is adopted quickly and performs most knowledge-work tasks with little need for human involvement?
Under those assumptions, the economic output becomes much larger.
The Modest Scenario Adds 1.6% to GDP
The first scenario is called “modest.”
Here, AI has an economic effect roughly comparable to the internet. AI creates real productivity gains, but the change stays within the historical range associated with major technologies.
Under this scenario, US GDP reaches $34.1 trillion in 2030, measured at 2025 price levels.
That is 1.6% above the model’s no-AI baseline.
Labor also keeps most of its current share of national income. The model puts labor’s share at 59.4%, compared with 40.6% going to capital.
The employment effects are also limited compared with the more aggressive scenarios.
This is the least disruptive path in Anthropic’s model.
It does not mean AI has little value. Rather, the economic effect looks more like a gradual technology transition than a sudden restructuring of the labor market.
The Substantial Scenario Takes GDP to $36.3 Trillion
The second scenario is called “substantial.”
Here, AI has a much larger effect on knowledge work. Anthropic assumes AI can perform about half of all knowledge work by 2030, with much of that work performed autonomously.
Under this scenario, US GDP reaches $36.3 trillion.
That is 8.3% above the no-AI baseline. The economy also grows at roughly twice its normal rate within the model.
The labor market changes more noticeably.
The model puts unemployment among cognitive workers at about 4.5%, while overall unemployment reaches about 4.6%.
Knowledge-worker wages remain roughly flat compared with the no-AI path, while wages for other workers rise more strongly.
This creates an important distinction.
AI does not need to eliminate huge numbers of jobs for income distribution to change. A technology that makes knowledge workers more productive could still reduce demand for some human tasks and increase demand for workers in other areas.
The Extreme Scenario Reaches $44.4 Trillion
The third scenario is where the headline figure comes from.
Anthropic calls this the “extreme” scenario.
It assumes AI becomes more productive than humans at most knowledge-work tasks. AI performs nearly all of those tasks autonomously, while almost no new knowledge-work tasks are created for humans.
The scenario would likely require recursively self-improving AI and rapid adoption across knowledge work.
Under those conditions, US GDP reaches $44.4 trillion in 2030.
That is 32.4% above the no-AI baseline.
Annual GDP growth reaches about 15.4% in the model. At that rate, the economy would roughly double every 4.5 years.
This is the point where the economic story changes from a normal technology transition to something much more disruptive.
The economy becomes substantially richer.
But workers do not share those gains evenly.
The Biggest Cost Falls on Knowledge Workers
The extreme scenario produces a sharp rise in unemployment among cognitive workers.
Anthropic’s model puts cognitive-worker unemployment at 17.9% in this scenario.
Overall unemployment reaches 11.9%.
Knowledge-worker wages also fall 11.5% below the no-AI path, according to the model.
At the same time, wages for workers outside knowledge occupations rise significantly. The model estimates those wages at 33.6% above the no-AI path in the extreme scenario.
That difference reflects how AI changes demand across the economy.
If AI handles a large share of cognitive work, companies need fewer people to perform some of those tasks. At the same time, higher productivity could increase demand for physical infrastructure and other services.
Anthropic gives a similar logic in its explanation of the model. If AI makes infrastructure design and permitting faster, for example, more construction projects could follow, increasing demand for construction workers.
The result is not simply “AI takes all the jobs.”
Instead, the model shows a major shift in which types of work are in demand.
Labor Could Receive a Smaller Share of a Bigger Economy
Another major finding concerns the split between labor and capital.
In the modest scenario, labor receives 59.4% of GDP while capital receives 40.6%.
In the substantial scenario, labor’s share falls to 56.1%, while capital receives 43.9%.
In the extreme scenario, the difference becomes much larger.
Labor receives 45.2% of GDP while capital receives 54.8%.
This creates one of the most important questions raised by the report.
What happens when the economy becomes much larger but workers receive a smaller portion of the total income?
Anthropic’s model suggests that AI could make capital more valuable because companies would rely more heavily on AI systems, computing infrastructure and other resources used to produce economic output.
The total economic pie gets bigger.
But the share going to workers gets smaller.
That means GDP growth alone would not tell the full story.
Americans Expect Something Close to the Middle Scenario
Anthropic also surveyed US adults about their expectations for AI.
The survey included 10,980 people and was conducted through Morning Consult between August 11 and August 23, 2026.
Anthropic used the responses to estimate what economic scenario the typical respondent’s expectations would imply.
The median respondent’s answers landed close to the substantial scenario.
Anthropic says the typical respondent’s expectations imply GDP about 10% higher by 2030 than it would be without AI, with overall unemployment around 5%. Around 10% of respondents held views that were closer to the extreme scenario.
This does not mean most Americans expect 17.9% unemployment among knowledge workers.
Instead, their responses suggest that the typical person expects AI to have a significant economic effect by 2030, but not the full transformation described by the extreme scenario.
Why the $44.4 Trillion Figure Comes With a Catch
The $44.4 trillion figure is striking because it shows how much economic output could be generated under an aggressive AI scenario.
But the same scenario produces the report’s biggest labor-market problems.
The model shows an economy with much higher GDP, faster growth and higher wages in some occupations.
At the same time, many knowledge workers face falling wages or unemployment.
This is why the report does not treat economic growth as the only issue.
A country could become much wealthier while large groups of workers struggle to benefit from that wealth.
The distribution of the gains becomes as important as the size of the gains.
Anthropic says the central challenge in the extreme scenario is ensuring that the benefits of economic growth are broadly shared and that the costs are not concentrated among particular groups.
The Model Also Has Important Limits
Anthropic does not present the explorer as a complete picture of the future.
The company says the current model leaves out several economic and technological factors.
For example, the model does not include scenarios involving highly capable physical robots. It also does not fully capture policy responses, financial market disruptions, business cycles or some effects linked to data-center construction.
External economists also reviewed an early version of the research. Anthropic says reviewers raised questions about whether AI-exposed occupations would shrink or grow, how much AI could accelerate technological progress and how well the model captures the experience of individual workers.
Those limitations are important because the economy does not operate like a fixed mathematical system.
Workers change occupations. Companies change strategies. Governments introduce new policies. Consumers change how they spend money.
AI itself also continues to develop.
A model built around today’s understanding of AI therefore cannot establish exactly what the US economy will look like four years from now.
What the Report Means for the AI Economy
Anthropic’s research points to a future where AI could increase economic output under every scenario it models.
The question is how large the gains become and who receives them.
The modest scenario produces $34.1 trillion in GDP.
The substantial scenario reaches $36.3 trillion.
The extreme scenario reaches $44.4 trillion.
The gap between those outcomes is enormous.
But the labor-market differences are just as important.
The substantial scenario produces moderate disruption. The extreme scenario produces unemployment levels far outside normal conditions, particularly among cognitive workers.
For businesses, the research highlights the importance of AI adoption and productivity.
For workers, it shows why skills and the ability to move into growing occupations could become increasingly important as technology changes the task mix inside jobs.
For policymakers, the findings raise questions about education, worker transitions, income distribution and how the gains from AI should be shared.
The report therefore offers more than a giant GDP number.
It shows that the economic impact of AI depends heavily on how the technology develops and how quickly society adopts it.
Conclusion
Anthropic’s new economic model presents three very different pictures of the US economy in 2030.
The modest scenario produces $34.1 trillion in GDP and limited labor-market disruption. The substantial scenario takes GDP to $36.3 trillion and assumes AI performs about half of knowledge work. The extreme scenario pushes GDP to $44.4 trillion, with annual growth reaching about 15.4%.
That final scenario is also where the warning becomes harder to ignore.
Knowledge-worker unemployment reaches 17.9%, overall unemployment reaches 11.9%, knowledge-worker wages fall below the no-AI path and labor’s share of income drops to 45.2%.
Anthropic is not saying this outcome will happen.
The company is showing what could happen if AI reaches a particular level of capability and adoption.
That distinction matters.
The report’s central message is that AI could make the economy much larger without guaranteeing that workers share the gains equally. A richer economy does not automatically mean a more secure labor market.
By 2030, the biggest economic question might therefore not be whether AI creates wealth.
It could be who gets to benefit from the wealth AI creates.
Frequently Asked Questions
1. Does Anthropic predict that US GDP will reach $44.4 trillion by 2030?
No. The $44.4 trillion figure belongs to Anthropic’s extreme scenario. The company does not assign probabilities to its three scenarios and says the model should be used to examine possible economic outcomes rather than predict the future.
2. What happens to jobs in Anthropic’s extreme AI scenario?
The model estimates cognitive-worker unemployment at 17.9% and overall unemployment at 11.9%. Knowledge-worker wages fall 11.5% below the no-AI path, while wages in other occupations rise substantially.
3. What does Anthropic’s AI economic model mean for workers?
The model suggests that AI could change the demand for different types of work. Some knowledge-work tasks could become automated while demand rises in occupations less exposed to AI. The report therefore highlights worker transitions, new skills and the distribution of AI-generated income as major issues for the coming years.
