Artificial intelligence saves the US economy! Anthropic: TFP increases by 1.1% and reduces $2 trillion in debt

The research from the National Bureau of Economic Research (NBER) indicates that maintaining a stable debt-to-GDP ratio can be achieved by increasing the total factor productivity (TFP) growth by just an additional 0.5 percentage points annually, which would be enough to stabilize public finances. If productivity growth continues for ten years, the projected debt level would decrease by approximately $2 trillion. AI company Anthropic’s analysis suggests that AI-assisted productivity has the potential to boost TFP by about 1.1 percentage points, which is more than twice the amount needed for fiscal stability.

The 0.5% Rescue Equation for US Fiscal Policy

The severity of the US fiscal crisis needs no further elaboration: debt-to-GDP ratios continue to rise, and interest payments are consuming an increasing share of the budget. However, NBER’s research offers a seemingly simple solution: achieve fiscal stability by enhancing total factor productivity (TFP). TFP reflects whether, under unchanged capital and labor inputs, an economy can continuously generate more output through technological advances and efficiency improvements.

A 0.5% increase may seem minor, but its impact is profound. According to NBER estimates, if such productivity gains can be sustained over ten years, the baseline forecast of US government debt would decrease by about $2 trillion. Over a 30-year horizon, the debt-to-GDP ratio would be 42 percentage points lower than the baseline forecast, and even 80 percentage points lower than a pessimistic scenario. The logic is straightforward: as the economy’s efficiency improves, the same tax rate yields more revenue, and the debt relative to GDP naturally decreases.

Nobel laureate Paul Krugman has pointed out that whether a country can raise living standards over time depends almost entirely on its ability to increase per capita output. Technological progress is at the core of everything. The importance of TFP is not only reflected in abstract growth theories but also directly related to the sustainability of US fiscal policy. The question is, where does the additional 0.5% TFP growth come from? Historically, such productivity leaps are usually associated with major technological revolutions—electricity, computers, the internet—each fundamentally transforming economic operations.

Anthropic’s 1.1% Revolutionary Finding

Anthropic, an AI company, provides a startling answer to this question. They analyzed approximately 100,000 actual conversations with Claude.ai, attempting to estimate the time difference in completing the same tasks with or without AI assistance. The study concludes that AI-assisted productivity has the potential to increase TFP by about 1.1 percentage points, more than double the 0.5% needed for US fiscal stability.

This figure has far-reaching implications. If a 0.5% productivity boost can stabilize government finances for decades, then a 1.1% increase could theoretically have revolutionary effects on the economy and public finances. Anthropic’s research provides a concrete example: Claude can complete a course design in 11 minutes, saving teachers about 4 hours of work. If such time savings could be replicated across the entire economy, the cumulative effects would be substantial.

However, Anthropic also acknowledges the limitations of their study. Whether time saved translates directly into increased output remains highly uncertain. Critics point out that the saved time might not necessarily be invested in higher-value economic activities but could instead be used for entertainment or leisure, such as scrolling social media or reading reports. In this scenario, AI indeed enhances people’s welfare and leisure time but may not necessarily increase overall wealth, thus offering limited help in solving US debt issues.

The Three Key Mechanisms and Challenges of AI-Driven TFP Growth

Time Savings Effect: Claude saves teachers 4 hours of work, but whether this time is converted into increased productivity remains to be seen.

Potential for Structural Change: Historically, technological revolutions not only accelerate tasks but also fundamentally change how work is done. However, modeling such effects is difficult.

Speed of Adoption: The research assumes the capabilities of current models without considering the ongoing evolution of AI, which could bring additional productivity gains in the future.

Cautious Optimism Behind Conservative Estimates

It is worth noting that Anthropic believes their 1.1% estimate may be conservative. The study does not account for the acceleration in AI adoption speed nor the potential for future models’ capabilities to continue evolving and boosting productivity. In other words, the research assumes that humans will continue to use current methods and language models at existing levels for the next decade. Given that large language models (LLMs) have shown significant improvements every few months, and that human application methods are rapidly evolving, 1.1% might be just a “lower bound” approximation of AI’s productivity effect.

More importantly, the study measures only the impact of AI on speeding up existing tasks, not on fundamentally restructuring workflows and production methods. Anthropic points out that major productivity leaps in history—electricity, computers, the internet—weren’t just about doing old tasks faster but about transforming how things are done. Such structural changes are hard to model but often produce the most profound impacts.

Nevertheless, the researchers remain cautious, clearly listing methodological limitations and assumptions. They also admit that even if AI truly creates more fiscal space for the US, future policymakers might again increase spending and accumulate debt. However, given the widespread perception that fiscal risks are imminent, even a small realization of this optimistic scenario is worth looking forward to. The potential contribution of AI to US fiscal health may far exceed our current expectations.