American productivity surged 2.7% in 2025 according to Tyler Cowen, the economist who had diagnosed the “great stagnation” in 2011. This increase, the strongest in fifteen years, coincides with a downward revision of 403,000 jobs and real GDP growth of 0.7% in the fourth quarter. For Cowen, this decoupling signals the return of productivity gains that artificial intelligence has been promising for years.
Thirteen years after theorizing the miring of American innovation, the George Mason University economist reverses course. But other voices point to the concentration of gains among technology platforms and the absence of wage spillovers. The debate now centers on who captures these productivity gains: companies, workers, or shareholders.
The Essentials
- Tyler Cowen calculates 2.7% growth in American productivity in 2025, a first since 2009
- Job creation revised downward by 403,000 positions, real GDP up 0.7% in Q4
- The reversal thesis: AI finally generates measurable gains after years of investment
- Critics point to capture by tech platforms and absence of wage gains
- The stakes: determining whether this productivity will benefit workers or remain concentrated
Cowen Reverses His Great Stagnation Thesis
Tyler Cowen made waves in 2011 with “The Great Stagnation,” an essay diagnosing the miring of American innovation since the 1970s. His thesis: the United States had exhausted the most accessible technological fruits and struggled to recover the productivity gains of the Thirty Glorious Years.
Today, the George Mason University economist makes amends. In an analysis published on his blog Marginal Revolution, he calculates a productivity increase of 2.7% for 2025, the strongest progression since the 2009 financial crisis. This figure rests on the decoupling between moderate growth, 0.7% in the fourth quarter of 2025, and massive downward revisions to job creation.
The Bureau of Labor Statistics indeed subtracted 403,000 positions from employment statistics, suggesting that economic growth relies less on hiring than on productive efficiency. For Cowen, this development marks the culmination of a decade of investments in artificial intelligence and automation.
“We are finally witnessing the inverted Solow paradox,” he writes, referencing the famous observation by economist Robert Solow that “you see computers everywhere except in the productivity statistics.” This time, technological gains would translate concretely into economic data.
AI Emerges from Its Pure Investment Phase
This acceleration in productivity coincides with the maturation of artificial intelligence tools in the real economy. After years of massive investment — OpenAI raised $6.6 billion in October 2024 — American companies are integrating these technologies into their operational processes.
The most exposed sectors show signs of transformation. In financial services, Goldman Sachs reports a 40% reduction in time spent on regulatory compliance tasks thanks to automation. In consulting, McKinsey & Company estimates that its consultants handle 30% more cases with the same workforce.
More broadly, American technology platforms — Meta, Google, Microsoft — display record profit margins despite colossal infrastructure investments in AI. Meta generated $62.4 billion in net profit in 2024 while injecting $38 billion into its data centers and language models.
This dynamic contrasts with the growing precarity observed in other sectors of the American economy, where automation destroys more jobs than it creates. The gap widens between a hyper-productive technology elite and a workforce whose added value stagnates.
Gains Remain Concentrated Among Giants
This productivity renaissance masks an unequal distribution of benefits. According to Federal Reserve Bank of San Francisco data, the ten largest American technology companies capture 60% of total productivity gains, compared to 40% in 2019.
Paul Krugman, Nobel Prize-winning economist, tempers Cowen’s optimism. In a New York Times op-ed, he emphasizes that American median wages advanced 2.1% in 2024, less than residual inflation of 2.4%. Productivity gains do not mechanically translate into wage increases, contrary to dynamics observed in the post-war period.
This capture is partly explained by the monopolistic structure of digital industries. Alphabet controls 91% of global internet searches, Amazon 38% of American e-commerce, Apple 57% of high-end smartphones. These dominant positions allow gains in efficiency to be retained rather than shared through competition.
Traditional companies struggle to replicate these performances. In manufacturing, productivity advances only 0.8% according to Bureau of Labor Statistics data. The gap widens between a technological vanguard and an economic fabric struggling to absorb innovation.
The Test of Technological Diffusion
The central issue now concerns the ability of these gains to diffuse. Historically, technological revolutions take a generation to irrigate the entire economy. Electrification required forty years to transform American industry, computerization thirty years to reshape services.
Does artificial intelligence follow the same pattern? Early signs remain mixed. Unlike China, which is massively industrializing humanoid robots, the United States prioritizes software integration in high value-added sectors. This approach concentrates benefits but limits their scope.
American small and medium-sized enterprises, which employ 47% of the private workforce, have difficult access to advanced AI tools. Training, integration, and maintenance costs hinder adoption. Only 23% of companies with fewer than 500 employees use automation tools, compared to 78% of firms with more than 5,000 employees according to a MIT Sloan School of Management study.
This asymmetry reproduces inequalities observed during previous transitions. Productivity gains first benefit the companies best endowed with capital and skills, before progressively diffusing — if political and regulatory conditions permit.
Redistribution Policies in Question
Facing this concentration, economists debate redistribution mechanisms. Gabriel Zucman, an inequality specialist at Berkeley, proposes progressive taxation of AI-related profits to finance public investments in education and training.
Other voices, like that of Daron Acemoglu at MIT, argue for strengthened antitrust regulation of technology platforms. The objective: force competition so that productivity gains translate into price reductions or wage improvements.
The Biden administration embarked on this path with lawsuits against Google and Meta, but proceedings span years. Meanwhile, giants consolidate their dominant positions and internalize the benefits of innovation.
Tyler Cowen, for his part, bets on natural market dynamics. In his vision, productivity acceleration will eventually benefit the entire economy through job creation and service improvements. This confidence in market mechanisms divides economists, particularly on the ability of American capitalism to spontaneously share the fruits of innovation.
The Post-Stagnation Economy Takes Shape
If Tyler Cowen is right, the American economy enters a new phase of its development. After fifteen years of sluggish growth and disappointing productivity gains, AI could restart the American productive machine. The 2025 figures suggest this shift is beginning, even if its durability remains to be proven.
The stakes transcend economic statistics. They concern the American development model: can growth driven by technological innovation generate shared prosperity, or does it reinforce structural inequalities? The answer will determine whether the United States regains a virtuous cycle of inclusive growth or sinks into a two-speed society.
The coming quarters will tell whether this productivity increase confirms itself and irrigates the real economy. For now, Tyler Cowen at least had the merit of burying his own thesis — a sign that the American economy retains its capacity to surprise its observers.