The Economics of Transformative AI (2025)
The Economics of Transformative AI is the December 2025 volume edited by Ajay Agrawal (University of Toronto), Erik Brynjolfsson (Stanford), and Anton Korinek (University of Virginia). The book assembles sixteen studies from leading economists examining how AI of sufficient capability would reshape the fundamental structures of economic life. The volume's organizing concept — Transformative AI (TAI), defined as AI capable of increasing total-factor productivity growth by three to five times historical averages — was deliberately chosen over the more common artificial general intelligence. The distinction matters. AGI is a technical benchmark about whether machines can match human cognition across all domains. TAI is an economic benchmark about whether the technology is powerful enough to reshape the economy at a magnitude comparable to the steam engine or electrification. The economist's question is not whether the machine can think like a human. It is whether the machine can transform like a revolution.
The Capture Economics Frame — Contrarian ^ Opus
There is a parallel reading that begins from the political economy of who controls AI infrastructure rather than its aggregate productivity effects. The volume's TAI framing — focusing on whether AI can achieve 3-5x TFP growth — may miss the more immediate question of concentration. Even if AI delivers transformative productivity gains, those gains flow through infrastructure controlled by perhaps five companies globally. The steam engine and electrification, the volume's historical comparisons, emerged in eras of competitive markets and antitrust enforcement. Today's AI revolution unfolds in an environment where network effects, data moats, and capital requirements create natural monopolies that regulatory frameworks haven't caught up to.
The economists assembled in this volume bring sophisticated modeling of innovation dynamics and labor markets, but their frameworks assume competitive equilibriums that may not apply when the means of intelligence production are this concentrated. Acemoglu's pessimistic estimates of 0.53-0.66 percent TFP impact might be too optimistic if we consider not just the technology's capability but the extractive dynamics of its deployment. The SaaSpocalypse the entry mentions — where AI vendors capture value from their customers' markets — is not a bug but the predictable outcome when transformative technology meets platform economics. The volume's methodological shift from "can machines think" to "can they transform productivity" is valuable, but a third question looms: who captures the transformation? Historical technological revolutions eventually diffused their benefits through competition and regulation. This one may be different — not because the technology is more powerful, but because the mechanisms of capture are more sophisticated and the political economy more favorable to concentration.
In the AI Story
The TAI framing is methodologically important. It shifts the debate from contested philosophical questions about machine consciousness and general intelligence to empirical questions about economic transformation that can be studied with economic methods. Whether a system is "truly intelligent" is not resolvable by current science and may never be. Whether a system is powerful enough to produce TFP growth of 3-5x historical averages is empirically tractable — requiring data on adoption, complementary investments, and productivity outcomes that economic methods can assess.
The volume's structure reflects the breadth of concerns transformative AI raises. Chapters examine innovation dynamics, market structure effects, employment and wage distribution, inequality and distribution, macroeconomic implications, and the speculative frontier — what economics looks like if and when AI substitutes for human cognition at scale. Contributors include some of the most prominent economists working on AI, including Philippe Aghion, Daron Acemoglu, and Chad Syverson alongside the editors.
The volume appeared at a moment when the AI transition's economic implications were becoming central to policy debates. The productivity statistics were beginning to show movement — Brynjolfsson's February 2026 claim of 2.7 percent U.S. productivity growth in 2025 was close to the volume's publication. Corporate AI spending was surging. The SaaSpocalypse was unfolding. The economic discipline needed to develop frameworks for analyzing transformations that were happening faster than economic research could normally keep pace with.
The volume's tension between optimistic and pessimistic contributors reflects the broader economics debate. Brynjolfsson's own position — mindful optimism about transformative potential — coexists in the volume with Acemoglu's more pessimistic estimates of AI's TFP impact (0.53-0.66 percent over a decade). The editors' decision to include rather than resolve this tension was deliberate — signaling that the economics profession has not reached consensus on the magnitude or trajectory of AI's effects and that the framework for analyzing them is still being developed.
Origin
The volume emerged from an NBER conference organized by the three editors in the prior year. The editors brought complementary perspectives: Agrawal's work on AI and economic decision-making (author of Prediction Machines), Brynjolfsson's long research program on productivity and technology, and Korinek's work on macroeconomic modeling of transformative technologies.
The volume was published by the University of Chicago Press in December 2025, positioning it as an academic reference at the moment the AI transition was becoming central to economic policy debates.
Key Ideas
TAI over AGI. Defines transformative AI economically (TFP growth threshold) rather than philosophically (general intelligence benchmark).
Integrates 16 perspectives. Brings together leading economists across the optimism-pessimism spectrum, reflecting unresolved debates.
Multi-dimensional analysis. Examines innovation, markets, employment, inequality, and macroeconomic effects rather than single aspect.
Methodologically tractable. Makes AI's economic impact empirically studyable rather than conceptually contested.
Marks economics catching up. Represents the discipline's effort to develop frameworks for analyzing transformations happening faster than normal research cycles allow.
Appears in the Orange Pill Cycle
The Synthesis Through Time Horizons — Arbitrator ^ Opus
The right weighting between these perspectives depends fundamentally on time horizon. For immediate effects (2025-2027), the capture frame dominates — perhaps 80/20. The concentration of AI capabilities in a handful of firms shapes deployment patterns more than aggregate productivity potential. The SaaSpocalypse and platform dynamics the contrarian highlights are the lived reality of AI adoption, where vendors extract value before markets can adjust. Here, political economy trumps productivity economics.
For medium-term trajectories (2028-2035), the balance shifts toward 60/40 favoring the volume's productivity focus. If AI achieves even half the TFP growth the TAI threshold suggests, the sheer magnitude of value creation begins to overflow capture mechanisms. New entrants emerge, open-source alternatives mature, and regulatory frameworks adapt. The volume's inclusion of both optimistic and pessimistic estimates becomes more relevant as we see which adoption patterns materialize. The methodological contribution — making AI economically tractable rather than philosophically contested — proves its worth as data accumulates.
The synthetic frame that holds both views is that transformative technologies undergo distinct phases: concentration, diffusion, and democratization. The volume implicitly assumes we're already in diffusion, while the capture frame correctly identifies we're still in concentration. Both are right at different points in the cycle. The steam engine also began with patent monopolies and factory concentration before becoming infrastructure. The question isn't whether AI will transform productivity or remain captured, but how long the concentration phase lasts and what determines the transition. The volume provides tools for measuring transformation; the capture frame reminds us that measurement alone doesn't determine distribution. Together they map not just AI's economic potential but the political economy of its realization.
Further reading
- Agrawal, Ajay, Erik Brynjolfsson, and Anton Korinek (eds.). The Economics of Transformative AI. University of Chicago Press, 2025.
- Agrawal, Ajay, Joshua Gans, and Avi Goldfarb. Prediction Machines. Harvard Business Review Press, 2018.
- Korinek, Anton and Joseph Stiglitz. Artificial Intelligence and Its Implications for Income Distribution and Unemployment. NBER, 2017.