The American AI Boom and the British Industrial Revolution
Yifu Xing: Energy, Labor, and the Limits of First-Mover Advantage
Introduction:
Unfortunately, artificial intelligence does not yet appear capable of staging a Terminator-style takeover of humanity. But that does not mean the world has not begun to reorganize itself around machines that think, write, classify, code, recommend, and increasingly act on behalf of human users. The rise of generative AI has already changed how students write essays, how companies manage information, how software is produced, how scientific research is accelerated, and how states imagine technological power. The most important question is therefore not whether AI is powerful, but why some societies are better positioned than others to capture that power.
At present, the United States remains the leading center of global AI development. Stanford's 2026 AI Index reports that U.S. private AI investment reached $285.9 billion in 2025, more than twenty-three times China's $12.4 billion, and that the United States released more notable AI models in 2025 than any other country.[1] The basic explanation is familiar: the United States has elite universities, deep capital markets, large technology firms, a sophisticated venture-capital ecosystem, and a culture that rewards risky innovation. Yet this explanation is incomplete. AI is not only a matter of algorithms and entrepreneurs. It is also a matter of electricity, land, chips, cooling systems, gas turbines, transmission lines, and water. The cloud may seem weightless, but it rests on very heavy infrastructure.
This article argues that the American AI boom resembles the British Industrial Revolution in one crucial respect: both were made possible by a particular country's ability to mobilize energy at scale. Britain's early industrial dominance depended in part on its access to cheap coal and its ability to turn fossil energy into mechanical power. The United States' current AI advantage depends in part on its ability to turn electricity, natural gas, chips, and data centers into computational power. However, the comparison also reveals an important difference. Britain's advantage lasted for decades because industrialization required a deep transformation of energy, labor, machinery, and urban geography. The American lead in AI may prove less durable because models diffuse quickly, competitors can learn from the frontier, and China is already narrowing the technical gap. Energy still matters, but in the AI era it may determine the pace of competition more than the permanent winner.
I. Energy and the Logic of First-Mover Advantage
Using energy to explain why one place modernized faster than others is not new. One influential interpretation of the British Industrial Revolution emphasizes that Britain did not industrialize simply because it had clever inventors or a superior national character. It industrialized because its economy made certain inventions profitable. Robert Allen argues that eighteenth-century Britain combined high wages with unusually cheap energy. In that setting, labor-saving machinery, steam engines, and coal-based production were more profitable in Britain than in many other parts of Europe or Asia.[2] Innovation was therefore not only an intellectual breakthrough; it was a response to prices, resources, and incentives.
E. A. Wrigley describes the Industrial Revolution as an escape from the constraints of an “organic economy.” Before fossil fuels, most energy came from land-based flows: food for humans and animals, wood for heating and industry, wind for sails and mills, and water for mechanical power. Because these energy sources ultimately depended on land, economic growth faced ecological limits. Coal altered this constraint. It offered a concentrated store of ancient solar energy that could be extracted and burned without requiring an equivalent expansion of farmland or forest.[3] In this sense, the Industrial Revolution was not simply a revolution in machines. It was a revolution in the energy basis of society.
Emma Griffin's account of Britain's industrialization is useful here because it resists the idea of a single dramatic leap. She emphasizes that between 1700 and 1870 Britain experienced uneven stages of economic transformation. Population grew, textile technologies improved, and some sectors changed significantly during the eighteenth century. Yet the most dramatic restructuring of national income, urbanization, and employment came when industrial growth became more closely connected to fossil-fuel use, especially coal-powered steam.[4] This matters for comparison with AI because the early history of data centers followed a similar two-stage pattern. For years, the digital economy expanded without explosive electricity growth; then AI workloads began to change the energy profile of computing.
Britain's coal advantage did not mean that other countries lacked industry. France, for example, had important centers of textile production, engineering, luxury manufacturing, and scientific expertise. But French industrialization remained more geographically dispersed and more dependent on water power for longer. The relative persistence of hydraulic power made French industrial growth less concentrated around coalfields and less quickly reorganized around large steam-powered factories.[5] The contrast should not be exaggerated: Britain also used water power, and France also used steam. But the comparison shows why energy systems matter. A new technology does not spread simply because it exists. It spreads when infrastructure, geography, capital, and prices make it usable at scale.
II. The AI Boom as an Energy Boom
The American AI boom is often described as a software revolution, but its material foundation is the data center. Training and running large models requires densely packed servers, specialized chips, storage, networking equipment, and cooling infrastructure. The Lawrence Berkeley National Laboratory estimated that U.S. data-center electricity use grew from 58 terawatt-hours in 2014 to 176 terawatt-hours in 2023, reaching about 4.4 percent of total U.S. electricity consumption. The same report projected that data-center use could rise to between 325 and 580 terawatt-hours by 2028.[6] These numbers make clear that AI is not merely a digital phenomenon. It is becoming one of the major new loads on the American power system.
The International Energy Agency reaches a similar conclusion. In its 2025 report on energy and AI, the IEA found that the United States accounted for the largest share of global data-center electricity consumption in 2024, followed by China and Europe. It also projected that in the United States, data centers would account for nearly half of electricity-demand growth between 2024 and 2030.[7] That is an extraordinary claim. It means AI infrastructure is becoming one of the forces shaping grid planning, electricity prices, local permitting battles, and future energy investment.
This dependence gives the United States a real advantage. The country has large technology firms willing to spend enormous sums on compute; it has deep financial markets; it has an existing cloud-computing base; and it has abundant domestic energy production, especially natural gas. Natural gas matters because data centers require reliable, continuous electricity. Wind and solar are expanding rapidly, but without sufficient storage and transmission, they cannot always meet the around-the-clock demand of hyperscale computing. Gas-fired generation is therefore attractive because it can be built relatively quickly, run continuously, and adjust output more flexibly than many other resources.[8] This does not mean gas is environmentally ideal. It means that, under current grid constraints, it is politically and economically convenient.
The result is a strange contradiction. The technology often advertised as immaterial, intelligent, and futuristic is in practice pushing some regions back toward very traditional questions of fuel, water, and land. Data centers need electricity, but they also need cooling. LBNL estimated that U.S. data centers directly consumed billions of liters of water in 2023, and the environmental burden of new facilities has become increasingly controversial in communities where water supplies, local power prices, or land use are already politically sensitive.[9] AI therefore resembles the industrial revolution not because GPUs are steam engines, but because both revolutions convert a new energy regime into economic power while distributing the costs unevenly.
The Electric Power Research Institute has warned that data centers could consume between 9 and 17 percent of U.S. electricity by 2030, compared with roughly 4 to 5 percent today.[10] Even if the lower end of such forecasts proves more accurate, the policy implications are significant. The central question for AI leadership may become less “who has the best model?” and more “who can connect enough power, fast enough, without creating political backlash?” This is where the comparison with Britain becomes especially revealing. Britain's industrial advantage depended not only on inventing machinery, but on embedding that machinery in an energy system. America's AI advantage likewise depends not only on model design, but on whether the country can build the infrastructure that model design now requires.
III. Labor: From Factories to Automated Offices
Energy is only one side of the comparison. The other is labor. The British Industrial Revolution did not merely increase output; it reorganized work. Textile machinery changed where production happened, who performed it, and under what discipline. As spinning and weaving moved from domestic settings toward larger workshops and factories, households were exposed to new forms of competition and dependence. In some cases, entire families entered factory labor; in others, women and children became central to the industrial labor force.[11] The factory was not simply a building. It was a new social arrangement built around time discipline, supervision, wage dependence, and machine pacing.
Child labor illustrates the social violence hidden inside technological progress. Mechanized textile production created tasks that children could perform because machinery reduced the physical strength required for some operations. Children were cheaper than adult men, smaller in body, and often seen by employers as easier to discipline.[12] Early factory owners also drew on systems of parish apprenticeship, through which poor children were transferred from local poor-law authorities into industrial employment.[13] This was not an accidental by-product of industrialization. It was one of the ways industrial capitalism solved its labor problem: by moving vulnerable people into systems where their labor could be cheaply organized and intensively supervised.
AI is not reproducing the same labor system. It is not sending orphaned children into cotton mills. But it may be creating a different kind of labor reorganization. Instead of replacing muscle with steam-powered machinery, generative AI automates, accelerates, or cheapens cognitive tasks: summarizing documents, drafting emails, writing code, producing marketing copy, generating legal templates, translating text, analyzing spreadsheets, and answering customer-service queries. The immediate effect is not always job destruction. Sometimes it is task restructuring. A worker may keep the same job title while the composition of the job changes radically.
This is why the labor-market effects of AI are difficult to measure. McKinsey Global Institute estimated that generative AI and other forms of automation could accelerate occupational transitions in the United States, with office support, customer service, food service, and production work among the categories most exposed to declining demand.[14] Yet recent empirical work also suggests that the early effects are uneven. Some studies find productivity gains and widespread experimentation, while others caution that economy-wide effects remain modest so far.[15] Daron Acemoglu argues that if AI affects only a limited share of tasks profitably, its macroeconomic impact may be significant but far smaller than the most optimistic forecasts suggest.[16] The lesson from the Industrial Revolution is that the social impact of a technology is rarely visible from the invention alone. It depends on how firms reorganize labor around it.
The likely result is not a simple division between jobs that disappear and jobs that survive. A more useful distinction is between workers who can use AI to increase their bargaining power and workers whose tasks become easier to monitor, standardize, or replace. Senior professionals may use AI as leverage: a lawyer can draft faster, a programmer can test faster, a consultant can analyze more data, and a manager can produce more polished documents. Junior workers, however, may face the opposite problem. If AI can perform the entry-level tasks through which young workers once learned, firms may hire fewer beginners or demand more from them immediately. This would resemble the Industrial Revolution less as direct substitution and more as a change in the ladder of skill formation.
IV. Who Gets Rich? Capital, Compute, and the New Industrial Geography
The Industrial Revolution enriched factory owners, coal interests, machinery manufacturers, merchants, financiers, and eventually many workers through long-term growth. But the gains were not evenly distributed, especially in the early decades. Industrialization created new forms of wealth precisely because it concentrated control over machinery, capital, energy, and labor. The same dynamic appears in the AI boom. The largest gains flow first to firms that own models, cloud platforms, chips, data centers, proprietary data, and distribution channels. AI may make individual users more productive, but the rents tend to collect where compute, capital, and market access are concentrated.
This helps explain why the AI race is not simply a race of talent. It is a race of vertically connected systems. The most powerful firms can buy advanced chips, sign long-term power-purchase agreements, build or lease data centers, recruit researchers, collect user data, and deploy products through existing platforms. Smaller firms and countries may use open-source models or rent cloud access, but they often remain dependent on infrastructure controlled elsewhere. In this respect, AI power looks like industrial power: those who control the productive apparatus can shape the terms on which others use it.
Yet the analogy also has limits. Britain's industrial lead endured partly because the barrier to replication was high. Industrialization required mines, canals, railways, machine tools, factories, skilled mechanics, imperial markets, finance, and urban labor pools. AI has high barriers too, especially at the frontier, but it also diffuses through software. Models can be copied, compressed, distilled, fine-tuned, and adapted. Research circulates quickly. Open-weight models make advanced capabilities available outside the original labs. This does not eliminate American advantage, but it makes the lead more contestable than Britain's nineteenth-century coal-and-steam advantage.
China is the most important example. Although U.S. private AI investment remains far higher, the Stanford AI Index notes that the performance gap between leading U.S. and Chinese models has narrowed sharply.[17] This means America's advantage may not be a stable empire of technology, but a moving frontier. In Britain's case, coal helped create a durable industrial center. In the AI case, energy and compute may create temporary acceleration, but the knowledge embedded in models can travel faster than heavy machinery once did.
Still, energy may slow that diffusion. AI may be software, but frontier AI requires enormous capital expenditure. If advanced models become increasingly expensive to train and run, then only a few firms and states will be able to compete at the highest level. If, by contrast, algorithmic efficiency improves and smaller models become good enough for most commercial tasks, then AI power may spread more widely. The future of the American lead therefore depends on a tension between two trends: the centralization of compute and the democratization of model access.
V. Will AI Accelerate the Energy Transition or Delay It?
A final comparison concerns the environment. Coal made the British Industrial Revolution possible, but it also locked industrial society into a carbon-intensive development path. AI faces a parallel dilemma. On one hand, AI could help optimize grids, design batteries, discover new materials, improve forecasting, reduce waste, and make industrial systems more efficient. On the other hand, the race to build data centers could increase short-term demand for fossil-fuel generation and delay decarbonization if utilities respond by building new gas infrastructure. The IEA argues that renewables are likely to be the fastest-growing source of electricity for data centers through 2030, but it also emphasizes that fossil fuels and nuclear power will continue to play important roles in meeting data-center demand in different regions.[18] AI may therefore accelerate the energy transition in some contexts while intensifying fossil-fuel dependence in others.
This ambiguity is exactly what makes the industrial analogy useful. In both cases, technology does not determine the energy system by itself. Institutions do. Britain did not simply “discover” coal; it built a society that extracted, transported, financed, and burned coal on a vast scale. The United States will not simply “use AI”; it will build a political economy around AI infrastructure. The outcome will depend on grid regulation, energy markets, local permitting, water policy, chip supply chains, climate commitments, antitrust policy, and the bargaining power of communities asked to host the infrastructure.
This also changes the meaning of national power. A country that leads in AI must do more than produce clever algorithms. It must coordinate electricity supply, industrial policy, research funding, immigration, education, infrastructure finance, and public legitimacy. If data centers raise electricity bills or intensify water scarcity, AI may provoke local resistance. If states subsidize AI infrastructure too generously, the public may ask why ordinary consumers are paying for private compute. If AI raises productivity but concentrates profits, it may intensify inequality rather than produce shared prosperity. These are not secondary problems. They are the politics of industrialization returning in digital form.
Conclusion
The American AI boom and the British Industrial Revolution are not identical. Coal-powered steam engines transformed textile mills, mines, transport, and factories; AI transforms information, prediction, writing, coding, scientific research, and administration. Industrialization reorganized manual labor and urban space; AI reorganizes cognitive labor and digital infrastructure. Yet the comparison reveals a shared structure. Both revolutions depend on the conversion of energy into productive capacity. Both create first-mover advantages for countries able to mobilize resources at scale. Both concentrate profits around those who own the machinery of production. Both promise progress while shifting costs onto workers, communities, and environments.
The crucial difference is durability. Britain's coal-based industrial advantage was difficult to copy quickly because it required a large-scale transformation of energy, labor, capital, and geography. The American AI advantage is formidable, but more fragile. It rests on compute, capital, talent, chips, energy, and corporate platforms, but it exists in a world where software travels quickly and competitors learn rapidly. The United States may remain the center of the AI revolution, but it cannot assume that its current lead will last simply because it began ahead.
The deeper lesson is that no technological revolution is purely technological. The British Industrial Revolution was not just about steam engines; it was about coalfields, wages, factories, children, cities, and empire. The AI revolution is not just about models; it is about electricity, gas turbines, water, chips, data centers, platform power, and the workers whose tasks are being reorganized. If Britain's industrial rise teaches us anything, it is that the winners of a revolution are not merely those who invent new machines. They are those who control the systems that allow machines to reshape society.
Bibliography
Acemoglu, Daron. “The Simple Macroeconomics of AI.” NBER Working Paper 32487, 2024. https://doi.org/10.3386/w32487.
Allen, Robert C. The British Industrial Revolution in Global Perspective. Cambridge: Cambridge University Press, 2009.
Allen, Robert C. “Why the Industrial Revolution Was British: Commerce, Induced Invention, and the Scientific Revolution.” Economic History Review 64, no. 2 (2011): 357-384. https://doi.org/10.1111/j.1468-0289.2010.00532.x.
Brynjolfsson, Erik, Danielle Li, and Lindsey R. Raymond. “Generative AI at Work.” Quarterly Journal of Economics 140, no. 2 (2025): 889-942.
Caron, François. An Economic History of Modern France. Translated by Barbara Bray. New York: Columbia University Press, 1979.
Electric Power Research Institute. Powering Intelligence 2026: Analyzing Artificial Intelligence and Data Center Energy Consumption. Palo Alto, CA: EPRI, 2026. https://powering-intelligence.epri.com/executive-summary.html.
Ellingrud, Kweilin, Saurabh Sanghvi, Gurneet Singh Dandona, Anu Madgavkar, Michael Chui, Olivia White, and Paige Hasebe. Generative AI and the Future of Work in America. McKinsey Global Institute, July 2023. https://www.mckinsey.com/mgi/our-research/generative-ai-and-the-future-of-work-in-america.
Griffin, Emma. A Short History of the British Industrial Revolution. 2nd ed. London: Palgrave Macmillan, 2018.
Honeyman, Katrina. Child Workers in England, 1780-1820: Parish Apprentices and the Making of the Early Industrial Labour Force. Aldershot: Ashgate, 2007.
Humlum, Anders, and Emilie Vestergaard. “Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI.” NBER Working Paper 33777, 2025. https://doi.org/10.3386/w33777.
International Energy Agency. Energy and AI. Paris: IEA, 2025. https://www.iea.org/reports/energy-and-ai.
Kirby, Peter. Child Labour in Britain, 1750-1870. Basingstoke: Palgrave Macmillan, 2003.
Sajadieh, Sha, Loredana Fattorini, Raymond Perrault, Yolanda Gil, Vanessa Parli, Lapo Santarlasci, Juan Pava, Nestor Maslej, Russ Altman, Erik Brynjolfsson, Carla Brodley, Jack Clark, Virginia Dignum, Vipin Kumar, James Landay, Terah Lyons, James Manyika, Juan Carlos Niebles, Yoav Shoham, Elham Tabassi, Russell Wald, Toby Walsh, and Dan Weld. Artificial Intelligence Index Report 2026. Stanford, CA: Stanford Institute for Human-Centered Artificial Intelligence, 2026. https://hai.stanford.edu/ai-index/2026-ai-index-report.
Shehabi, Arman, Sarah J. Smith, Nathaniel Horner, Ines Azevedo, William M. P. Smith, and others. 2024 United States Data Center Energy Usage Report. Berkeley, CA: Lawrence Berkeley National Laboratory, 2024. https://doi.org/10.71468/P1WC7Q.
Tuttle, Carolyn. Hard at Work in Factories and Mines: The Economics of Child Labor during the British Industrial Revolution. Boulder, CO: Westview Press, 1999.
U.S. Department of Energy. “DOE Releases New Report Evaluating Increase in Electricity Demand from Data Centers.” December 20, 2024. https://www.energy.gov/articles/doe-releases-new-report-evaluating-increase-electricity-demand-data-centers.
U.S. Energy Information Administration. “Natural Gas Explained: Where Our Natural Gas Comes From.” Accessed July 2026. https://www.eia.gov/energyexplained/natural-gas/where-our-natural-gas-comes-from.php.
Wrigley, E. A. Energy and the English Industrial Revolution. Cambridge: Cambridge University Press, 2010.
[1] Sajadieh et al., Artificial Intelligence Index Report 2026.
[2] Allen, British Industrial Revolution in Global Perspective; Allen, “Why the Industrial Revolution Was British.”
[3] Wrigley, Energy and the English Industrial Revolution.
[4] Griffin, A Short History of the British Industrial Revolution.
[5] Caron, An Economic History of Modern France.
[6] Shehabi et al., 2024 United States Data Center Energy Usage Report; U.S. Department of Energy, “DOE Releases New Report.”
[7] International Energy Agency, Energy and AI.
[8] U.S. Energy Information Administration, “Natural Gas Explained.”
[9] Shehabi et al., 2024 United States Data Center Energy Usage Report; U.S. Department of Energy, “DOE Releases New Report.”
[10] Electric Power Research Institute, Powering Intelligence 2026.
[11] Griffin, A Short History of the British Industrial Revolution.
[12] Tuttle, Hard at Work in Factories and Mines, 95-99.
[13] Honeyman, Child Workers in England, 15, 56.
[14] Ellingrud et al., Generative AI and the Future of Work in America.
[15] Humlum and Vestergaard, “Still Waters, Rapid Currents.”
[16] Acemoglu, “The Simple Macroeconomics of AI.”
[17] Sajadieh et al., Artificial Intelligence Index Report 2026.
[18] International Energy Agency, Energy and AI.