Paper 03

Gen-AI, Power, and Labor

Two class discussion responses: how Gen-AI's capital concentration may end the “Engels' pause,” and how the U.S. is building its strategic AI playbook.

Status
Discussion Post
Date
Jan 17, 2026
Reading
7 min

Abstract

A two-part class discussion post. Part I argues that Gen-AI's combination of extreme capital concentration and unusually rapid diffusion may break the historical pattern in which wages eventually catch up to productivity after a general-purpose technology arrives. Part II examines how current U.S. AI strategy converges around four levers, innovation control, economic leverage, narrative leadership, and military deployment, and what that reveals about how nations now project power through general-purpose technologies.

00 / Listen

Audio companion to this paper.

00 / Paper

I. Gen-AI Is Likely to Cause Massive Global Unemployment

Historically, general-purpose technologies (GPTs) have diffused slowly, disproportionately affected blue-collar labor, and often emerged through relatively open or accessible development pathways. The steam engine, patented in 1769, took nearly a century to become the dominant means of production (Crafts, 2004). As the primary GPT of the British Industrial Revolution, it generated major output gains while wages for factory workers stagnated or declined, allowing owners to capture the bulk of the resulting profits for roughly half a century (Allen, 2009).

The dot-com era presents a contrasting case. While still capital-intensive, the core technologies were comparatively accessible. Entrepreneurs with limited resources could build transformative companies using open-source tools and local servers. Amazon, for example, emerged from this environment, as Jeff Bezos founded Amazon with an initial $10,000 investment (Stone, 2013).

Generative AI represents a sharp departure from both of these trajectories. OpenAI's first public release was backed by a $1 billion investment from Microsoft in specialized supercomputing infrastructure, placing the foundational development of the technology far beyond the reach of individuals or small firms (Microsoft, 2019). Today, over 60 percent of foundational AI platforms are controlled by just two firms, Microsoft and Amazon (IoT Analytics, 2025). At the same time, diffusion has been unusually rapid, with an estimated 40 percent of the global workforce exposed to AI tools within two years of ChatGPT's release (Cazzaniga et al., 2024).

This combination of extreme capital concentration and rapid diffusion is historically unusual for a GPT. It suggests that the distributional consequences of Gen-AI may be more severe than those observed in earlier technological transitions. Allen (2009) describes an Engels' pause, a period in which wages lag productivity before eventually recovering.

Gen-AI may produce a prolonged or even permanent decoupling of productivity gains from labor income, an Engels' stop, characterized by persistent wage stagnation and widening inequality across both white-collar and blue-collar labor markets.

II. The Strategic Battle for AI Dominance: Power and Control

First, it's important to understand the United States' interest in winning the AI battle before we look at current AI strategies. The National Security Commission on Artificial Intelligence Final Report (2021) states that allowing regimes like China to obtain AI superiority would threaten democratic values, citing examples of mass surveillance and military power.

Innovation control and export policy

Looking at the current strategy of the US, it appears that current actions revolve around control and power. As the US is leading the AI industry in innovation, it looks for ways to maintain that innovative edge (The White House, 2025). One of those ways to maintain innovative controls is to limit the exports of GPUs needed to continue building AI with China by implementing a “volume cap” policy that restricts the quantity of advanced chips, such as NVIDIA's H200, that can be sold to Chinese entities to 50% of the volume shipped to U.S. customers (McGuire, 2026).

Narrative leadership

The US also wants to control the narrative and global usage guidelines surrounding AI. They are doing so by leading international safety consortiums and enforcing the “AI Action Plan,” which prioritizes American-led standards in international diplomacy (The White House, 2025).

Economic power

The larger the economy of a country, the more power it exerts on a global stage (Kroese, 2024). The US sees its heavy investment in AI as an essential driver towards economic growth (Bank of America Institute, 2025). In fact, economists from the Penn Wharton Budget Model (2025) project that Generative AI adoption will increase US GDP by 1.5% by 2035 and nearly 3% by 2055.

Military and national security

And of course, a government strategy isn't complete without discussing military implications and national security. Our current administration isn't shy about testing AI within the military as it is actively experimenting with Grok AI to operate on Pentagon networks (Associated Press, 2026). Furthermore, the Department of Defense has launched the “Replicator” initiative, aiming to field thousands of autonomous AI-driven systems by August 2025 to counter China's military scale (Defense Innovation Unit, 2025).

01 / Why I explored this

These two class discussion responses sit next to each other because they ask the same underlying question from opposite ends, one from the labor market, one from the state. Writing them together forced me to notice that the same feature of Gen-AI (concentrated infrastructure plus rapid diffusion) is what makes both the wage story and the geopolitical story unusual.

02 / The question I was wrestling with

Does Gen-AI's combination of capital concentration, rapid diffusion, and state-level strategic competition break the historical pattern of broadly-shared gains from general-purpose technologies?

03 / Key insights

  1. 01

    Historical GPTs diffused over decades; Gen-AI has exposed roughly 40% of the global workforce within two years of ChatGPT's release.

  2. 02

    Foundational Gen-AI development sits behind a $1B+ infrastructure floor, with over 60% of platforms controlled by Microsoft and Amazon, a profile no prior GPT has had this early.

  3. 03

    If productivity gains accrue mainly to infrastructure owners, Allen's “Engels' pause” may become an “Engels' stop”: a permanent decoupling of productivity from labor income.

  4. 04

    U.S. AI strategy is converging on four levers at once, export caps on advanced chips, economic-growth framing, international norm-setting, and military deployment (Replicator, Pentagon Grok pilots).

  5. 05

    Both stories share a mechanism: control of the underlying compute and models is becoming the primary site where power, economic and geopolitical, is concentrated.

06 / Citations

14 citations
  1. Allen (2009)

    Allen, R. C. (2009). Engels' pause: Technical change, capital accumulation, and inequality in the British industrial revolution. Explorations in Economic History, 46(4), 418–435.link

  2. Associated Press (2026)

    Associated Press. (2026, January 13). Pentagon embraces Musk's Grok AI chatbot as it draws global outcry. PBS News Hour.link

  3. Bank of America Institute (2025)

    Bank of America Institute. (2025, October). Economic shifts in the age of AI. Bank of America.link

  4. Cazzaniga et al. (2024)

    Cazzaniga, M., Jaumotte, F., Li, L., Melina, G., Panton, A. J., Pizzinelli, C., Rockall, E., & Tavares, M. M. (2024). Gen-AI: Artificial intelligence and the future of work (Staff Discussion Note SDN/2024/001). International Monetary Fund.link

  5. Crafts (2004)

    Crafts, N. (2004). Steam as a general purpose technology: A growth accounting perspective. The Economic Journal, 114(495), 338–351.link

  6. Defense Innovation Unit (2025)

    Defense Innovation Unit. (2025). The Replicator initiative: Progress update. U.S. Department of Defense.link

  7. IoT Analytics (2025)

    IoT Analytics. (2025, March 4). The leading generative AI companies.link

  8. Kroese (2024)

    Kroese, B. (2024, December). GDP in the future. Finance & Development, 61(4). International Monetary Fund.link

  9. McGuire (2026)

    McGuire, C. (2026, January 14). The new AI chip export policy to China: Strategically incoherent and unenforceable. Council on Foreign Relations.link

  10. Microsoft (2019)

    Microsoft. (2019, July 22). Microsoft invests in and partners with OpenAI to support us building beneficial AGI [Press release].link

  11. NSCAI (2021)

    National Security Commission on Artificial Intelligence. (2021). Final report.link

  12. Penn Wharton Budget Model (2025)

    Penn Wharton Budget Model. (2025, September 8). The projected impact of generative AI on future productivity growth. University of Pennsylvania.link

  13. Stone (2013)

    Stone, B. (2013). The everything store: Jeff Bezos and the age of Amazon. Little, Brown and Company.

  14. The White House (2025)

    The White House. (2025, July 28). New federal AI action plan.link

08 / Future questions

  • What policy interventions could re-couple productivity gains and labor income before an “Engels' stop” sets in?
  • How durable are U.S. export-control levers as Chinese domestic chip capacity grows?
  • If compute is the new strategic resource, what does “non-alignment” look like for countries that depend on both U.S. and Chinese AI stacks?
← All research

End of paper 03