“Productivity isn’t everything, but in the long run it is almost everything.”
This observation by Nobel laureate Paul Krugman captures the foundational premise of modern economics. At its core, "total factor productivity" (TFP) serves as the primary metric economists use to quantify the contribution of technological innovation to economic growth. It represents the sustained ability of an economy to generate a higher volume of output while utilizing the exact same amount of inputs. Because of this dynamic, TFP is arguably the most critical measurement in the economic toolkit, as the continuous process of producing more with fewer resources is the fundamental mechanism through which human living standards improve.
As Krugman has frequently explained, a country’s long-term capacity to elevate its standard of living depends almost entirely on its ability to steadily raise its output per worker. Technology is the primary engine that makes that upward trajectory possible, and TFP is the instrument used to measure it.
To grasp the tangible impact that technology-generated productivity can have on macroeconomic stability, one only needs to look at a recent research paper published by the National Bureau of Economic Research (NBER). According to the study’s findings, an additional 0.5% of annual TFP growth would be enough to completely stabilize the United States government’s finances, holding the debt-to-GDP ratio right at its current level.
While 0.5% may sound modest, the compounding effect over time is staggering. If that rate of growth is sustained over the next decade, the NBER estimates it would successfully reduce the baseline forecast for US government debt by a massive $2 trillion.
Over a 30-year horizon, a sustained 0.5% boost to TFP would drive the US government’s debt-to-GDP ratio down to a level 42 percentage points lower than the NBER’s baseline forecast. Compared to the organization’s pessimistic economic forecast, that same productivity boost would result in a debt-to-GDP ratio that is an astounding 80 percentage points lower.
Given the seemingly intractable and hopeless state of current government finances, maintaining today’s level of indebtedness looks like a dream scenario—one that appears almost too good to be true. Yet, artificial intelligence researchers at Anthropic believe that the modern economy might actually be capable of achieving an even more dramatic transformation.
Anthropic recently conducted a comprehensive study analyzing 100,000 real-world conversations on Claude.ai. The primary objective of the research was to estimate the precise amount of time tasks within those conversations would require with and without AI assistance, thereby shedding light on the broader productivity implications for the wider economy.
The conclusion reached by the researchers is striking: large language models (LLMs) could potentially raise total factor productivity by a full 1.1 percentage points.
If a 0.5% increase is capable of stabilizing US government finances for decades to come, the natural question is what a 1.1% boost might achieve. Economically speaking, an injection of that magnitude would likely solve almost every major long-term fiscal challenge on the horizon.
Of course, there are plenty of valid reasons to approach such optimistic forecasts with a healthy dose of skepticism.
The Anthropic study illustrates, for instance, that Claude can save educators approximately four hours of labor by generating comprehensive lesson plans and curricula in a mere 11 minutes. However, translating those localized time savings into broader economic output requires complex macroeconomic modeling—a practice notoriously fraught with best-guess assumptions and a degree of false precision.
Consequently, even if Anthropic’s calculations regarding time savings are entirely accurate, the company’s projections about overall productivity might still miss the mark. It is entirely possible that society will simply utilize the time saved by artificial intelligence to engage in activities that are economically unproductive, such as consuming more leisure media, watching short-form video content, or reading additional newsletters.
Under that specific scenario, the widespread adoption of AI would undeniably elevate human welfare by granting individuals more free time, but it would fail to expand aggregate wealth through higher economic output. That outcome would still represent wonderful news for individual well-being, but it would offer zero assistance to governments desperately searching for a silver bullet solution to their mounting debt crises.
Conversely, there are equally compelling arguments suggesting that Anthropic’s economic model may actually be far too pessimistic. As the researchers themselves note, their study deliberately excludes critical variables like the speed of technology adoption and the compounding productivity gains that will inevitably arrive with vastly more capable future AI systems.
In other words, the study’s baseline model assumes that society will continue utilizing artificial intelligence exactly as it does today, and that the workforce will rely on today’s generation of language models—completely unimproved—for an entire decade.
Given that language models exhibit noticeable leaps in capability every few months, and society has only just begun to understand how to effectively integrate them into daily workflows, Anthropic’s caution is well-founded. The researchers readily acknowledge that their 1.1% estimate likely represents an approximate lower bound on the true productivity effects of AI.
If that assessment proves accurate—if 1.1% is merely the floor for AI-induced productivity—society might successfully pay down national debt obligations while simultaneously enjoying significantly more leisure time. Furthermore, that projection accounts solely for AI’s impact on non-physical, desk-based work, leaving out the eventual macroeconomic contributions of physical robotics as automation hardware matures.
To dismiss this level of optimism entirely requires believing that the trillions of dollars corporations currently plan to invest in artificial intelligence capital expenditure and research and development will be completely squandered.
While that remains a distinct possibility—technological revolutions have historically struggled to arrive neatly on schedule—the most compelling reason for optimism is that Anthropic’s 1.1% estimate is derived solely from AI making existing tasks faster to complete. The model explicitly excludes AI’s capacity to fundamentally reinvent how those tasks are executed.
As Anthropic points out, historical analysis shows that truly transformative productivity improvements—such as those delivered by electrification, computing, or the internet—did not stem simply from accelerating legacy tasks. Instead, they emerged from the fundamental reorganization of production itself.
While it is impossible to accurately model these entirely new ways of working before they materialize, the ultimate macroeconomic impact is widely expected to eclipse the figures Anthropic has attempted to measure.
The researchers at Anthropic remain appropriately cautious, taking great care to caveat their hopeful findings by explicitly detailing the methodological limitations and numerous underlying assumptions of their study. Even if all those optimistic assumptions materialize and AI-driven productivity successfully rescues the US government from its fiscal quagmire, political realities suggest lawmakers might simply spend their way right back into the same deficit.
Nevertheless, given the near-universal consensus pointing toward inevitable fiscal disaster, even a small probability that Anthropic’s projections are correct provides a compelling reason to update economic priors. The financial position of the US government may not be as intractable as conventional wisdom suggests, and the long-term outlook for the US dollar may not be as perilous as many fear. In the long run, productivity remains almost everything—and artificial intelligence may be on the verge of making the global workforce dramatically more efficient.
Comments (0)