Talks

Shaping Technological Revolutions in Turbulent Times — Carlota Perez

56 min read

About this talk

Speaker: Carlota Perez (identified from her own site’s talk listing; the speaker is not named in the transcript)
Event: Luohan Academy Frontier Dialogue, “How Shall We Develop the Technologies to Embrace AI Technology” (event title from Perez’s site; talk listed there as 9 Dec 2025)
Date: 2025-12-09 (from Perez’s site listing; not confirmed from the video itself)
Length: 24:22

How to read this page. It has two kinds of content, always marked:

  • 🎙 In the talk. What the speaker says, with [MM:SS] timestamps. Only these parts are her views.
  • 📚 Deeper theory. Background from the traditions she draws on, including thinkers she does not name. This explains why she thinks what she thinks and who disagrees. It is not her claim unless a timestamp says so.

Suggested route. Summary → Map → Key terms → Part 2: The intellectual map (read once, slowly) → Deep dives in order → “Putting it together”.

Speakers

LabelWhoRole
PerezCarlota Perez (not named in the talk; see confidence note below)Economist of technology and development. Honorary Professor at UCL’s Institute for Innovation and Public Purpose and at SPRU (Sussex); adjunct professor at Tallinn University of Technology. Author of Technological Revolutions and Financial Capital (2002). Born Caracas, 1939 (Wikipedia)
HostNot heard. The talk opens with “thank you for the invitation” [00:01]Unnamed. The event is the Luohan Academy’s Frontier Dialogue

Identification confidence: high, but inferred. Nobody says her name in the audio. Three things point to her. (1) Her website lists a talk titled “Shaping Technological Revolutions in Turbulent Times: Lessons from History” at China’s Luohan Academy Frontier Dialogue on 9 December 2025, within the event “How Shall We Develop the Technologies to Embrace AI Technology” (carlotaperez.org). The title matches the opening line [00:01]. (2) The content is her framework: installation and deployment periods, the turning point, golden ages, five revolutions. (3) She uses her own coined terms, “creative construction” and “dynamic demand”. I could not find a recording or abstract to check the date against the video. Treat the exact date as (not confirmed).

Caption warnings (auto-captions; Perez2 gives the same text with small differences):

  • “bellipo progressive era” = Belle Époque (and the US Progressive Era) [09:46]. Perez2 has the same error.
  • “JBC’s idea about supply creating its own demand” = J.-B. Say (Say’s Law) [12:17].
  • “French shoring” (twice, [19:15] and [20:19]) = almost certainly friend-shoring. The captions were consistent in both versions, so it is a mishearing, not a slip.
  • “subraational arrangements” [17:38] = probably supranational (not confirmed; it could be “subnational”). I read it as supranational because she says that today it “looks impossible”, which fits global coordination.
  • “mid-9th century” [09:46] = mid-19th century (Perez2 fixes it).
  • “prochemicals” = petrochemicals. “reigns” at [07:45] is “range” in Perez2; “reigns” fits the sense (“production reigns rather than finance”).
  • “C2 next to the mine” [20:19] is unclear. Context says: process minerals at the mine using local renewable energy, not ship raw ore. I did not guess a word.
  • “this is the new revolution… coexists at the end with the next” [06:32] — the caption is garbled. She means each revolution overlaps with the one before and the one after.
  • “paradigm” appears as “parad.” at [03:44] in Perez1 (cut off).
  • Her list of previous revolutions [05:55] is muddled in speech. She says “the industrial revolution, the age of steam, coal, iron and railways. Before that [sic], the age of steel and heavy engineering… after that the age of the automobile, oil, petrochemicals and mass production.” She says “four previous revolutions”. Read it as four: (1) the Industrial Revolution, (2) steam and railways, (3) steel and heavy engineering, (4) oil, automobile and mass production. “Before that” is a slip of the tongue for “after that”, and she corrects herself.
  • Factual slips / flags. At [03:44] she dates mass production from 1913 (the moving assembly line). In her book she uses 1908 (Ford’s Model T) as the start date; the moving line at Highland Park began in 1913. Both are defensible; they are two markers, not a contradiction. At [09:46] she shows a chart of the top 1% income share and reads it as “20%” for the top 1% in installation periods against “20%” for the bottom 50% after the postwar golden age. I did not check the chart’s source or the country; it is presumably US data (not confirmed). See deep dive 3.

Summary

The question. Where is the world in the “information revolution”, and can it reach a global golden age instead of extended conflict? She frames it for a China and West audience.

The thesis, in five steps.

  1. A revolutionary technology (AI, plastics) is not a technological revolution. A revolution is a cluster of technologies plus cheap inputs, new infrastructure and a new “common sense” [00:39]–[05:19]. AI is part of the information revolution, so it can only be guided by guiding that whole revolution [05:19].
  2. Every revolution follows an S-curve in two periods. Installation is turbulent: creative destruction, polarisation, financial casino, crashes. Deployment is the golden age of “creative construction”. Between them is a turning point of instability [05:55]–[08:34].
  3. She says we are in the turning point now, “and we might be coming to the end” [08:34].
  4. Golden ages are not automatic. They need a proactive state that sets up a win-win game between business and society, and they depend on dynamic demand: demand that grows with the new core industries [07:45], [12:17]. The postwar golden age got it from suburbs, the Cold War, farm subsidies and the welfare state [13:20]–[16:05]. That demand was national.
  5. The information revolution is global, and it needs demand for capital goods and services. Today China and the West fight for markets that are too small [18:19]–[19:15]. The way out is to build demand in the global south, to rent rather than sell durable goods, and to build new welfare states, with AI used for health, education and public services [19:15]–[23:32].

The close. “A peaceful global golden age is possible with social progress, economic growth, and social justice in north, south, east, and west in a safe and healthy planet” [23:32]. “To get there, history is a useful guide” [24:15].

Her stance. Optimistic, but conditional on policy. She says social progress in capitalism “requires shaping action by the state, even though many free marketers will deny it” [08:34]. She names nobody and gives no numbers on AI itself.

Map — talking points

#PointTime
1Roadmap: what is a revolution, how they spread, the road ahead for China and the West[00:01]–[00:39]
2Definition of a technological revolution (cluster + cheap input + infrastructure + paradigm shift)[00:39]–[01:42]
3Definition of a revolutionary technology. AI belongs to the information revolution[01:42]–[03:04]
4Plastics as an example inside the mass production revolution[03:04]–[03:44]
5Mass production 1913 to the 1970s. Stagflation and offshoring as the sign of exhaustion[03:44]–[05:19]
6Why it matters: guide AI by guiding the whole revolution[05:19]–[05:55]
7Four earlier revolutions, one S-curve: installation, turning point, deployment[05:55]–[07:06]
8Installation = turbulent creative destruction, polarisation, financial casino[07:06]–[07:45]
9Deployment = creative construction, proactive state, production over finance[07:45]–[08:34]
10The turning point = instability, populism. “We are there… we might be coming to the end”[08:34]–[09:46]
11Inequality chart: top 1% high in installation, falls in deployment[09:46]–[10:42]
12Three golden ages (Victorian, Belle Époque/Progressive, postwar) and their three advances[10:42]–[12:17]
13Dynamic demand defined, against Keynes and Say[12:17]–[13:20]
14Postwar demand engines: suburbs, Cold War, farm subsidy[13:20]–[14:49]
15Welfare state as demand machine (mortgages, insurance, pensions, unions, free schooling)[14:49]–[16:05]
16Marshall Plan and World Bank. National vs global revolutions. Trade shares[16:05]–[17:38]
17ICT is global; needs supranational(?) arrangements; demand for capital goods and services[17:38]–[18:19]
18China is making the leap in two revolutions at once[18:19]–[19:15]
19Today’s problem: overproduction, zero-sum, risk of conflict[18:19]–[19:15]
20Way out: friend-shoring, global south demand, green development[19:15]–[20:19]
21Belt and Road. Rent durable goods. Process minerals at the mine[20:19]–[22:03]
22Domestic demand and welfare state in all countries[20:19]–[22:03]
23AI: the good life of services, not of waste[22:03]–[22:47]
24Forging ahead means mastering two revolutions (US and Germany over Britain; China now)[22:47]–[23:32]
25Welfare state as emergency in the West, essential complement in China. Close[23:32]–[24:22]

Part 1 — Before you read

Key terms

  • Technological revolution. Her definition [00:39]: “a set of interrelated and interdependent revolutionary technologies and technology systems”, with new dynamic products and industries, a low-cost input (“chips and data… before oil… before steel… before coal”), and a new infrastructure (internet; before that roads and electricity; before that ships and railways). It lowers costs and raises productivity “of every single sector”, causing “a complete paradigm shift” in common sense and demanding “a major reorganization of the whole social institutional framework” [01:42].
  • Revolutionary technology. “A major scientific and technological breakthrough” that can transform many sectors. It “follows and strengthens the paradigm of the prevailing revolution” [01:42]. Examples: plastics (inside mass production), AI (inside the information revolution).
  • Paradigm (techno-economic paradigm). A shared “common sense” about what is efficient, modern and profitable. In the book this is the techno-economic paradigm. The word is Thomas Kuhn’s, applied to firms and policy. See 2.2. Watch out: it means different things in different fields (box below).
  • Installation period / deployment period / turning point. The two halves of the S-curve and the hinge between them [06:32]. In the book, installation has two phases (irruption, frenzy), and deployment has two (synergy, maturity). She does not name these four in this talk.
  • Creative destruction. Schumpeter’s phrase. The new destroys or modernises the old [07:06].
  • Creative construction. Her own term, the golden-age opposite: “every single innovation brings more and more innovations upstream and downstream” because the paradigm is now shared [07:45].
  • Financial casino. Her phrase for the finance of installation, when money seeks money and not production [07:06].
  • Golden age. A deployment-period boom of social progress, growth and social justice [10:42]. A gilded age (the contrast term in the book, not used in the talk) is installation-style wealth for the few.
  • Dynamic demand. The central economic idea of the talk. Not “aggregate demand” (Keynes) and not “supply creates its own demand” (Say). It is “demand that responds to the nature of the new technologies and grows at the same rhythm as the potential production capacity of the core industries” [12:17]. It is made possible by “implicit or explicit government policy”.
  • Welfare state. Used here as a demand and security machine, not mainly as charity [14:14]–[16:05].
  • Forging ahead / catching up. A country surpasses the leader by mastering the new revolution and modernising the old one with it [22:47].
  • Friend-shoring (caption: “French shoring”). Moving supply chains and investment to politically friendly countries. She uses it here to mean investment into the global south [19:15].
  • Global golden age. The goal in the title: a golden age that spans north, south, east and west [23:32].

Words that mean different things in different fields

WordIn this talkElsewhere
ParadigmA shared business and policy common senseKuhn: a scientific community’s shared problem-solving model. Dosi: a technological paradigm (a pattern of problem-solving within a technology)
RevolutionA 50–70 year wave of diffusionPolitics: a sudden break. History: “Industrial Revolution” as one event in Britain
DemandDemand that grows with the core industriesKeynes: total spending, often cyclical. Micro: a curve of price against quantity
CapitalFinancial capital vs production capital (implied)Marx: capital as a social relation. Piketty: stock of wealth
Golden ageSocial-peace, growth and justice togetherEconomic history: the postwar boom, c.1945–73. Also the myth of a lost golden past

People named (a reading list)

PersonWhy they come upStart with
John Maynard KeynesPerez says her “dynamic demand” is not his “aggregate demand” [12:17]General Theory (1936), ch. 1–3, 12
J.-B. Say”Supply creates its own demand” (caption: “JBC”) [12:17]Traité d’économie politique (1803)
(Not named) Joseph Schumpeter”Creative destruction” is his term [07:06]Capitalism, Socialism and Democracy (1942), ch. 7
Marshall Plan; World BankFunded world demand and infrastructure [16:05]Institutions, not persons

The speaker names almost no thinkers. That is why Part 2 is long: the talk is a compressed version of a school of thought.

Perez’s own frame

Perez built this framework over four decades. 1983: an article in Futures on “Structural change and assimilation of new technologies in the economic and social systems” (from memory; not checked). 1988: “Structural crises of adjustment, business cycles and investment behaviour”, with Chris Freeman, in Dosi et al.’s Technical Change and Economic Theory (from memory). 2002: Technological Revolutions and Financial Capital: The Dynamics of Bubbles and Golden Ages, her main book. A later paper (2015, European Journal of Economic and Social Systems) drops the “Kondratiev long wave” label for “great surges” (paper, noted from its section titles in search results). Her site says she is writing a new book on the social shaping of technological revolutions and the state’s role in the move to golden ages (carlotaperez.org; title not given). Her recent policy slogan is that growth should be “smart, green, fair and global” (Wikipedia).

The five revolutions in the book (from memory; the dates are the book’s start markers, not checked here):

#RevolutionStart markerCore input / infrastructure
1Industrial Revolution1771, Arkwright’s mill, CromfordCotton, iron, canals
2Steam and railways1829, the Rocket trialCoal, steam, railways
3Steel, electricity, heavy engineering1875, Carnegie’s Bessemer plantSteel; telegraph, electric grids
4Oil, automobile, mass production1908, Ford Model TOil; roads, highways
5Information and telecommunications1971, Intel microprocessorChips; internet

In the talk she uses a different marker for 4 (1913) and the fifth is “now”, with chips and data as inputs [00:39].

The current debate (2025–26)

  • Is AI a new revolution or part of the fifth one? The talk takes a clear side: AI is “a revolutionary technology” inside the information revolution [01:42]. A summary of her own writing agrees: she places AI inside the ICT revolution, treats electrification as a comparable “revolutionary technology” within the steel revolution, and suggests a sixth revolution may come from AI converging with biotech and new materials (Irving Wladawsky-Berger, Dec 2025). The same post notes the weakness: the category call is a judgment, and the “not yet in deployment” claim rests on the idea that a financial correction has not happened, which is hard to test.
  • Will AI crash before it pays? Press coverage in 2025 used her model to ask whether AI is in the installation frenzy. A Financial Times column by John Thornhill, reposted on 22 Aug 2025, was headlined “Brace for a crash before the golden age of AI” (BizNews repost). I could read only the headline and subtitle: “History teaches that tech revolutions follow a predictable cycle of installation followed by creative destruction”. The column’s body is (not confirmed). Note the headline is the editor’s, not a quote from her.
  • Where is the turning point? Her 2002 book put the turning point after the dot-com crash. By 2019 she wrote that the present is “the 1930s, the turning point of the IT surge” and counted the Great Recession in it (Stratechery, 2021). In this talk she says we may be “coming to the end” of the turning point [08:34]. So the claim has moved: the same model has been used to say “the turning point has happened” (2002), “it is long” (2019) and “it may be ending” (2025). That is a fair test question for any dated theory (see Part 3, deep dive 2).
  • Steering technology. Perez appears with Daron Acemoglu in a 2025 CEPS podcast, “Steering the technological revolution”, and in a 6 Oct 2025 Oxford TIDE interview comparing the present with the 1930s (both listed on her interviews page, via a search summary; I did not hear them). Acemoglu and Johnson’s Power and Progress argues “there is nothing automatic” about technology bringing shared prosperity; it is “an economic, social, and political choice” (MIT Shaping Work, via search summary). This is the closest live partner to her argument that the state must shape golden ages.
  • China’s demand problem. Her claim that China and the West “are competing for their own insufficient markets” [18:19] matches the IMF’s Feb 2026 conclusion on China: “private domestic demand remained lackluster”, the 2025 current account surplus was about 3.3% of GDP, and the Board’s “overarching priority” was a “consumption-led growth model” (IMF press release 26/053, from a fetched summary of the release; the word “overcapacity” is not used there). This supports her diagnosis. It does not test her remedy.

Part 2 — The intellectual map

Perez speaks for 24 minutes to an audience that knows some of this. She names almost no one. Seven families of ideas meet in the talk. They are laid out here, with the thinkers she does not name.

 LONG-WAVE TRADITION ──────┐                  ┌────── FINANCE & BUBBLES
 Kondratiev · Schumpeter    │                  │       Hilferding · Veblen · Keynes
 Freeman · Mensch           ▼                  ▼       Minsky · Kindleberger
                     ┌──────────────────────────────────┐
 PARADIGMS & SYSTEMS │  PEREZ: a revolution is a cluster│ ◄── INSTITUTIONS & THE STATE
 Kuhn · Dosi · Nelson│  that needs a new common sense,  │     Polanyi · Gerschenkron
 & Winter · Hughes   │  and golden ages need the state  │     Regulation school · Mazzucato
 · GPT theory        │  to create dynamic demand        │     Acemoglu
                     └──────────────────────────────────┘
                       ▲              ▲               ▲
    DEMAND, WAGES      │   CATCH-UP & │   INEQUALITY  │  AI, PRODUCTIVITY
    Say · Keynes       │   GLOBAL ORDER│   Kuznets     │  & THE LIMITS
    Ford · Kalecki     │   List · Gerschenkron  Piketty │  Solow · David
    Kaldor-Verdoorn    │   Perez & Soete · Chang  Goldin-Margo  Brynjolfsson · Gordon
    Hirschman · Pasinetti  Rodrik                       │

2.1 The long-wave tradition — from Kondratiev to “great surges”

The problem it addresses. Why does capitalism grow in long swings of several decades, not just in 7–10 year business cycles? And why do big new technologies arrive in clusters, not one at a time?

The chain of thinkers

  1. Nikolai Kondratiev, “The Long Waves in Economic Life” (1925; Eng. 1935). A Russian economist who studied price and production series in Britain, France and the US. He claimed there are long waves of about 50 years. He was arrested and executed in the Stalinist purges (1938) (from memory; dates not checked). The idea survived by being renamed “Kondratiev waves”.
  2. Joseph Schumpeter, The Theory of Economic Development (1911; Eng. 1934), Business Cycles (1939), Capitalism, Socialism and Democracy (1942). He gives the cause. Entrepreneurs introduce innovations. Innovations come in swarms because once one entrepreneur succeeds, others follow. The swarm drives a boom; the old economy is destroyed by creative destruction. He linked Kondratiev’s waves to clusters of innovation: railways, steel and so on. Key point for Perez: she describes her own work as neo-Schumpeterian (Wikipedia). The line “the new destroys the old or modernizes it” [07:06] is Schumpeter.
  3. Gerhard Mensch, Das technologische Patt (1975; Eng. Stalemate in Technology, 1979). He argued that basic innovations bunch up in depressions, because falling profits push firms to take risks. Freeman challenged him. This debate (do depressions cause innovation, or does innovation cause the boom?) is one reason Perez puts the cause on the technology side.
  4. Chris Freeman (1921–2010), SPRU at Sussex, with John Clark and Luc Soete, Unemployment and Technical Innovation (1982). He proposed that each long wave has a techno-economic paradigm, a bundle of technologies, organisational forms and “common sense”, and that economic crises are mismatches between the new technology system and the old institutions. Freeman and Perez (1988) developed “structural crises of adjustment”. Freeman and Louçã, As Time Goes By (2001), gave the long history.
  5. Perez, Technological Revolutions and Financial Capital (2002). She makes the model concrete, with five revolutions, an S-curve with two periods, and a role for financial capital. She later moves away from the “Kondratiev” label, to “great surges”, partly because the waves are not regular enough to be called cycles (from the title of a section in her 2015 paper; reason not checked).

How it feeds the talk. The whole structure [05:55]–[08:34] is this tradition. What she adds to Schumpeter is the institutional half: swarms of innovation cause a boom, but whether the boom becomes a golden age depends on a social and political response.

2.2 Paradigms, trajectories and technological systems

The problem it addresses. Why do firms and governments keep doing things the old way after a technology has changed what is efficient? Why does “common sense” lag?

  1. Thomas Kuhn, The Structure of Scientific Revolutions (1962). A scientific paradigm is a shared set of examples, methods and assumptions. It changes by revolution after anomalies pile up. Perez borrows the word and the idea of a shift in common sense [01:42].
  2. Giovanni Dosi, “Technological paradigms and technological trajectories” (1982). He transfers Kuhn to engineering. A technological paradigm defines what problems are worth solving; a trajectory is the path of progress along it. So firms search in a bounded area. Key point: innovation is path-dependent, not a random draw.
  3. Richard Nelson and Sidney Winter, An Evolutionary Theory of Economic Change (1982). Firms follow routines; change is slow because routines are sticky. This helps explain why “organizations” must also change [01:42].
  4. Thomas Hughes, Networks of Power (1983). Technologies become large technical systems (electric grids, rail). Systems have “momentum” and need “system builders”, including utilities, regulators and financiers. Perez’s “new infrastructure” is a technical system in Hughes’s sense.
  5. Paul David, “Clio and the Economics of QWERTY” (1985). Standards lock in by history, not merit. This matters for why paradigms are hard to dislodge.
  6. General-purpose technologies (GPTs). Bresnahan and Trajtenberg (1995) defined GPTs as technologies that are pervasive, improve over time and spawn complementary innovations. Steam, electricity and computers are standard examples; AI is now argued to be one. Key link to the talk: a GPT is close to Perez’s “revolutionary technology”, and her “technological revolution” is the larger cluster. She would say GPT research is too narrow because it studies single technologies, not the bundle plus institutions.
  7. Carlota Perez and Luc Soete, “Catching up in technology: entry barriers and windows of opportunity” (1988). New paradigms briefly lower entry barriers for latecomers, because everyone must learn something new. This is the root of her “forging ahead” claim about China [22:47] (see 2.6).

How it feeds the talk. The sentence “a complete shift in common sense for technology and for business and for the economy and even for organizations” [01:42] is a compact version of Kuhn plus Dosi plus Nelson and Winter. It also explains why she sees policy as part of the technology, not outside it.

2.3 Finance and bubbles — the half the talk mostly skips

The problem it addresses. Why do bubbles and crashes keep arriving with new technologies? Is finance a help or a hazard for innovation?

The chain of thinkers

  1. Rudolf Hilferding, Finance Capital (1910). Banks and industry fuse; finance directs production. He wrote about German capitalism, but the idea of a separate finance power over industry comes from him.
  2. Thorstein Veblen, The Theory of Business Enterprise (1904). A split between industry (making things) and business (making money from ownership and credit). Perez’s “financial capital” vs “production capital” echoes this (my reading; she does not cite him in the talk).
  3. John Maynard Keynes, General Theory (1936), chapter 12. Stock markets are like a beauty contest: investors guess what others will guess. “When the capital development of a country becomes a by-product of the activities of a casino, the job is likely to be ill-done.” The phrase “financial casino” [07:07] is close to this, though I cannot say she is quoting it.
  4. Hyman Minsky, Stabilizing an Unstable Economy (1986). The financial instability hypothesis: stability breeds risk-taking; debt moves from hedge to speculative to Ponzi finance; a crash follows.
  5. Charles Kindleberger, Manias, Panics and Crashes (1978). A bubble starts with a displacement, a shock such as a new technology, which opens new profit chances. Then euphoria, then a crash. This is almost exactly the “installation frenzy” template. (He builds on Minsky.)
  6. Perez’s addition. She argues the bubble is functional. Financial capital floods into the new technology, builds the infrastructure faster than demand would justify (canals, railways, fibre), then crashes. The bust leaves cheap, usable infrastructure, and at that point finance and production are forced to realign. The turning point is where regulation reconnects finance to the real economy. A FT summary describes the same logic: bubbles “are often productive, funding the building of vital infrastructure” (BizNews repost, via a search summary).

Historical crashes in her story (from memory, not checked): the canal mania (1790s), the railway mania (1840s), the 1890s heavy-engineering boom and its crises, 1929, and the dot-com bubble (2000), and then 2008.

How it feeds the talk. The book’s title has Financial Capital in it, but the talk mentions finance only briefly: “from financial creativity to financial casino and major crashes along the way” [07:06] and “production reigns rather than finance” [07:45]. Treat the talk as the policy half of the model. The finance half lives in the book.

2.4 Institutions and the state — why golden ages need policy

The problem it addresses. If technology raises productivity, why does shared prosperity not follow by itself?

  1. Karl Polanyi, The Great Transformation (1944). Pure market societies cause a double movement: the market spreads, then society protects itself (labour law, social insurance), sometimes in ugly forms (fascism). Perez’s “populism” and “tensions” at the turning point [08:34] is a Polanyian reading.
  2. Alexander Gerschenkron, Economic Backwardness in Historical Perspective (1962). Latecomers use state and banks as substitutes for missing preconditions. Germany used banks, Russia the state. Relevant to China (2.6).
  3. The French Regulation School (Michel Aglietta, Robert Boyer, Alain Lipietz, 1970s–80s). Capitalism works through a regime of accumulation (how production and consumption fit) held together by a mode of regulation (laws, wage norms, money, the state). Fordism is the postwar example: mass production tied to mass consumption by wage growth with productivity. Perez’s account of the postwar golden age [13:20]–[16:05] is almost a Regulationist account. Aglietta’s A Theory of Capitalist Regulation (1976) is the first source.
  4. Social Structures of Accumulation (SSA). David Gordon, Richard Edwards, Michael Reich (1982) and Samuel Bowles and Thomas Weisskopf describe the same idea for the US.
  5. Douglass North, Institutions, Institutional Change and Economic Performance (1990). Institutions are the “rules of the game”; they change slowly. This matches her “institutional lag”.
  6. Mariana Mazzucato, The Entrepreneurial State (2013). The state funds the riskiest early technology (the internet, GPS, touch screens, pharma) and then does not share the rewards. Later, her “mission-oriented” approach says government should set the direction. Perez’s “directionality” is similar. Mazzucato is her colleague at UCL IIPP (from memory; check). The talk says the state sets “a win-win game between business and society” [07:45].
  7. Daron Acemoglu and Simon Johnson, Power and Progress (2023). The gains from technology are not automatic; they depend on who has power to steer the technology and share the gains (MIT Shaping Work). Acemoglu won the 2024 Nobel Prize with Johnson and James Robinson, as the publisher’s text says. This is the strongest mainstream cousin of her view.

How it feeds the talk. Her statement that “social progress in capitalism requires shaping action by the state even though many free marketers will deny it” [08:34] puts her in the Polanyi–Regulation–Mazzucato line, against the “market will sort it out” view.

2.5 Demand, wages and structural change — where “dynamic demand” comes from

The problem it addresses. Output can grow only if somebody buys it. Who buys, and why would that grow in step with productivity?

  1. Say’s Law (J.-B. Say, 1803). “Supply creates its own demand”: the income from production buys the production. She rejects it [12:17]. Keynes’s attack on it in 1936 is the better-known rejection.
  2. Keynes’s aggregate demand. Total spending can fall short of capacity, so output and jobs collapse. The cure is public spending. Perez says her idea is different: it is “not just quantitative” [12:17]. It cares about what kind of demand, tied to the new core industries.
  3. Henry Ford and the “five-dollar day” (1914). Ford doubled wages to cut turnover and to make workers into buyers. The story is often used to explain Fordism: mass production needs mass consumers. The talk’s unions point [14:49] (“the unions forced this which was good for business but they don’t necessarily understand that”) is in this spirit.
  4. Michał Kalecki (1930s–40s). Workers spend what they earn, capitalists save; so wage share drives demand. Kalecki also wrote on the political limits of full employment.
  5. The Kaldor–Verdoorn law. Nicholas Kaldor and Petrus Verdoorn: productivity growth rises with output growth (scale, learning). So demand and productivity feed each other. It is the nearest formal relative of her statement that demand must grow “at the same rhythm as the potential production capacity” [12:17]. (My reading; she does not cite it.)
  6. Albert Hirschman, The Strategy of Economic Development (1958). Backward and forward linkages: some industries create demand for suppliers and users. Her “creative construction… upstream and downstream” [07:45] is a linkage argument.
  7. Luigi Pasinetti, Structural Change and Economic Growth (1981). As productivity rises, demand moves across sectors. If demand does not shift to new sectors, growth stalls. This is a formal version of “demand has to respond to the nature of the new technologies”.
  8. Walt Rostow, The Stages of Economic Growth (1960). “Leading sectors” pull the rest of the economy. The postwar age of “high mass consumption” is his last stage, and the suburbs-and-cars picture is Rostow’s.
  9. Embedded liberalism (John Ruggie, 1982) and Bretton Woods (1944). The postwar order let states run welfare and full-employment policies while trade opened slowly. She notes that the postwar world was “international rather than global” [17:38] and that exports and imports stayed under 20% of GDP [16:05]. That is Ruggie’s point: the national economy was protected.

How it feeds the talk. Her idea of dynamic demand is a synthesis: Keynes (demand matters), Schumpeter (technology sets what grows), Regulation school (institutions tie wages to productivity), Hirschman (linkages). The postwar list [13:20]–[16:05] is the best example, so see Part 3.

2.6 Catch-up, global order and China

The problem it addresses. How do latecomers become leaders, and what does the world order look like when they do?

  1. Friedrich List, The National System of Political Economy (1841). Protect infant industries to catch up. His idea that Britain “kicked away the ladder” after it got ahead is also in Ha-Joon Chang’s Kicking Away the Ladder (2002).
  2. Gerschenkron (1962) and Moses Abramovitz (1986). The advantage of backwardness, with conditions. Catch-up needs “social capability”.
  3. Perez and Soete (1988). New paradigms open windows of opportunity. Latecomers can leapfrog if they build skills early. In the talk this is “forging ahead to challenge the leader as that of the United States and Germany surpassing Britain in the third revolution”, which “involve making the leap with two revolutions” [22:47]. China “has done” this, “upgrading mass production with the information revolution” [22:47].
  4. Alice Amsden, Asia’s Next Giant (1989), Robert Wade, Governing the Market (1990). East Asian states used credit, protection and export discipline in a deliberate way.
  5. Hegemonic transitions. Britain to the US, in the third revolution. Charles Kindleberger (the Great Depression as a failure of leadership), and Graham Allison’s “Thucydides trap” (2017) are the debate she is stepping into when she warns of “extended conflict” [19:15]. (My reading; she does not cite them.)
  6. Dani Rodrik on premature deindustrialisation (2016) and the “globalisation trilemma” (2011). States cannot have hyper-globalisation, national democracy and national sovereignty together. This is relevant to her claim that the ICT revolution “is likely to require subraational arrangements” [17:38].
  7. China’s data. The IMF’s 2026 conclusion says growth leans on exports and private demand is weak; it urges a consumption-led model and stronger social protection to “lower precautionary savings” (IMF, summary). That line is nearly her own: “security of the welfare state… so you’re secure for your old age” [20:19].

How it feeds the talk. The whole third part of the talk (China and the West) is catch-up theory plus a demand diagnosis. What is new in her argument: dynamic demand is the missing link between industrial policy and the welfare state. She asks China to build both “for the old and for the new” [18:19].

2.7 Inequality, productivity and the AI question

The problem it addresses. Do technology booms raise or lower inequality, and why has the information revolution not (yet) delivered the productivity boom its size would suggest?

  1. Simon Kuznets, “Economic Growth and Income Inequality” (1955). Inequality rises at first, then falls with growth: the Kuznets curve. Later work found the postwar fall was not automatic.
  2. Thomas Piketty and Emmanuel Saez, “Income Inequality in the United States, 1913–1998” (2003); Piketty, Capital in the Twenty-First Century (2013). Top income shares in the US form a U: high in the 1920s, low in 1950–70, high again from the 1980s. This is almost certainly the data type behind her chart [09:46] (not confirmed). It also fits her claim that inequality is high in installation and low in deployment: the top-1% peak of 1928 is just before the 1929 crash, and the 2000s peak is near the 2008 crash.
  3. Claudia Goldin and Robert Margo, “The Great Compression” (1992). The postwar narrowing of US wages was a policy-and-war event: the New Deal, the war labour board, unions and a surge in education supply. This supports her point that golden ages are made, but also warns that war was part of the cause.
  4. The Solow paradox (1987). “You can see the computer age everywhere but in the productivity statistics.” Paul David, “The Dynamo and the Computer” (1990): electricity too took about 40 years to lift productivity, because factories had to be redesigned. This is the same logic as her “deployment needs a new common sense”. Erik Brynjolfsson’s productivity J-curve (2021) is the modern version.
  5. Robert Gordon, The Rise and Fall of American Growth (2016). The great-inventions era (1870–1970) is a one-off; digital technology adds much less. This is the main sceptic of her optimism about deployment.
  6. Automation and labour. Acemoglu and Restrepo (2018–2022) argue that automation without new tasks lowers wages and the labour share. Brynjolfsson’s “Turing trap” (2022) argues that AI built to replace humans concentrates gains.
  7. Winner-take-all economics. Haskel and Westlake, Capitalism without Capital (2017): intangibles scale cheaply, so a few firms win. This contrasts with the steel-and-cars era, where firms had to hire many workers to scale. If this holds for AI, the postwar link between productivity and wages is weaker.
  8. Climate and materials. Her point about rental and in-place processing of minerals [20:19] belongs to the circular economy, which Walter Stahel (the “performance economy”, 1970s–80s) first proposed: sell the use of a product, not the product (from memory). Kate Raworth’s Doughnut Economics (2017) is the broader frame. See Part 3, deep dive 5.

How it feeds the talk. The most contested part of her view is the idea that deployment will repeat the golden-age pattern. Gordon, Haskel–Westlake and Acemoglu give three reasons it might not.

2.8 Two shorter threads

  • The “gilded age” vs “golden age”. The original Gilded Age (US, 1870s–1890s) was an installation-style era of vast fortunes and corruption; Twain and Warner coined the name in 1873. Perez uses “gilded” for a boom whose gains stay with few, and “golden” for broad ones. The talk uses only “golden”.
  • Synergy can be authoritarian. In the book’s account (as relayed in the Stratechery piece) the deployment “synergy” phase is “amoral” and can take authoritarian forms, such as Nazi Germany, and, in Ben Thompson’s reading, China (Stratechery). The talk does not say this. It speaks only of a “peaceful” golden age [23:32]. Keep the two apart: a golden age is not automatically a democratic one.

Part 3 — Deep dives

1. A revolutionary technology is not a technological revolution (AI inside the information revolution)

🎙 In the talk. A technological revolution is “a set of interrelated and interdependent revolutionary technologies and technology systems including new and dynamic products and industries as well as a low-cost impetus [input]… and a new infrastructure” [00:39]. It lowers cost and raises productivity “of every single sector of the economy, leading to a complete paradigm shift” [01:42]. A revolutionary technology “follows and strengthens the paradigm of the prevailing revolution and acts as an integral part of it” [01:42]. Her example is plastics, with “successive systems downstream” (consumer durables, industrial supplies) and “technologies upstream” (injection, compression, blow molding), all inside the mass production revolution [03:04]–[03:44]. She then states the aim: “I’m trying to distinguish the revolutionary technology that AI is from the information revolution to which it belongs” [01:42]. The reason it matters is that guiding AI “can only succeed in the context of shaping and providing directionality to the whole information revolution” [05:19].

She also says the mass production revolution ran from 1913 to exhaustion in the 1970s, and that stagflation and offshoring were signs it was “spent”: firms “could no longer increase the productivity of mass production” so they went abroad for cheap labour [03:44]–[04:21]. The information revolution “is still exercising modernization and creative destruction” on those old systems [04:21].

📚 Deeper theory

  • A taxonomy of “big”. At the bottom: an innovation (a product). Then a technology system (plastics). Then a GPT (steam, electricity). At the top, Perez’s technological revolution: a cluster of GPTs and industries plus a paradigm (2.2). Her claim is about level: AI is level three, the revolution is level four.
  • Why the level matters for policy. If AI is the revolution, then regulate AI. If AI is a technology inside the ICT revolution, then the questions are wider: platforms, data, finance, labour and public services. She pushes the second view. Compare Wladawsky-Berger’s note that the classification is a judgment (post).
  • Plastics is a good choice. Plastics depended on petrochemicals, which depended on cheap oil. The oil input and the automobile infrastructure are the revolution’s. Plastics only found its demand inside that cheap-oil economy.
  • The exhaustion claim. Her story of the 1970s: productivity in mass production stopped rising, profits fell, firms moved out. This links to the profit squeeze and productivity slowdown literature (Brenner, Gordon, the Regulation school). Others blame oil shocks, the end of Bretton Woods and wage militancy. She does not weigh these.
  • Overlap. Each revolution overlaps with its neighbours [06:32]. So at any date, old and new paradigms coexist, and that is why data from the 1970s to now can be read as installation of the new and modernisation of the old.

Where others push back

  • Classification is arbitrary. Is the “ICT revolution” one revolution since 1971, or has the internet (1995) or the smartphone (2007) or AI (2022) started a new one? The test is the paradigm shift in “common sense”, which cannot be measured. Critics call this unfalsifiable.
  • Dates are fitted. Moving the start of revolution four from 1908 to 1913 (she uses both) shows how soft the markers are.
  • Revolution two and three are less distinct. Historians of technology (Joel Mokyr, Nathan Rosenberg) see slow, continuous change, not discrete bursts (from memory; not checked).

What she leaves out

  • Any evidence on AI’s productivity so far, or how AI differs from earlier GPTs (it is software, so it spreads fast and cheaply, and the “new infrastructure” is data centres and power). The talk treats AI only as an instance.
  • The role of finance in this part. The book’s central idea is missing from the talk (see 2.3).

Check yourself

  1. What are the three elements of a technological revolution in her definition, and what is each one today?
  2. Why does she say plastics is not a revolution, even though it replaced so many natural materials?
  3. What policy difference follows if AI is “a revolution” rather than “a revolutionary technology inside one”?

2. The S-curve: installation, turning point, deployment

🎙 In the talk. “We have four previous revolutions to learn from… They are all profoundly different, but their propagation has followed the same basic pattern. Their trajectory has been an S-curve with two contrasting periods” [05:55]. The new revolution “always coexists with the previous and coexists at the end with the next” [06:32] (as captioned). The installation period is “a turbulent and unequal creative destruction time”. Listed features: industries disappear; “rise and decline of industries, regions and countries”; jobs and skills gained and lost; “income polarization, the prosperity of the few, the desperation of the many”; “from financial creativity to financial casino and major crashes” [07:06]. The turning point is “when all the problems created by the installation period… come to the surface”, and deployment is “like the solution to all those problems” [06:32]. At the turning point: “instability and uncertainty, political divisions, recessions, tensions, conflicts, resentment, populism. Do you recognize what we’re going through now?” [08:34]. Then the key claim: “I think not only that we are there, but we might be coming to the end and we might have ahead the golden age” [08:34].

📚 Deeper theory

  • What the S-curve measures. Diffusion (the share of firms or users who have adopted a technology) is an S-shaped logistic: slow, then fast, then saturated. Zvi Griliches (1957, hybrid corn) and Everett Rogers (Diffusion of Innovations, 1962) are the standard sources. Perez applies it to the whole paradigm, not one product.
  • Four phases (book). Irruption, frenzy, [turning point], synergy, maturity. Installation is about 2–3 decades, deployment similar. At irruption the new tech spreads against a still-dominant old paradigm; in frenzy finance goes into the new tech; after the crash, in synergy, production capital leads and the state sets the rules; in maturity markets saturate and the next revolution starts to appear. In the Stratechery piece, her phrase for maturity is “the twilight of the golden age, though it shines with false splendor” (Stratechery).
  • The turning point is a choice point. It can end in a golden age or a gilded age, depending on whether policy aligns finance with production. The 1930s New Deal and the postwar settlement are her examples of a good resolution (see Stratechery’s summary: “the greed of some is for the benefit of the many”).
  • The 1930s analogy. She has said since at least 2019 that the present resembles the 1930s, the turning point of the IT surge (Stratechery). Her 2025 Oxford TIDE interview reportedly makes the same comparison (via search summary; not heard).
  • Why populism? Installation polarises income and leaves regions behind; the political system lags. Compare Polanyi’s double movement (2.4) and recent work on the “China shock” (Autor, Dorn, Hanson, 2013) and populist voting.
  • The first-order puzzle. The previous turning point (1929–1940s) was resolved with a world war and a New Deal. If today is a turning point, what plays the part of war and New Deal? She says policy; critics worry about the war.

Where others push back

  • Unfalsifiable timing. Her 2002 book put the turning point at the dot-com crash. In 2019 it became 2000 to the 2008 crisis and beyond; in 2025 it “might be” ending [08:34]. A theory that moves its own dates is hard to test. Defenders say turning points are long and messy by design.
  • Empirical doubt about long waves. Economic historians such as Solomou, and Kuznets earlier, doubted that regular long swings exist in the data (from memory; not checked). Perez’s move to “great surges” is a partial reply.
  • The ICT revolution is already ~55 years old (from 1971), longer than the 40–60 years of earlier ones. Wladawsky-Berger flags this timing pressure (post).
  • Is it one global curve? Revolutions arrive at different times in different countries. China’s own installation and deployment are not in phase with the West’s.

What she leaves out

  • The finance mechanism: why the frenzy, why the crash, why a crash opens a golden age. The talk names “financial casino” and moves on.
  • How long the turning point lasts and what ends it. “We might be coming to the end” has no signal attached [08:34].
  • Who loses within a golden age (colonies, women’s unpaid work, the environment). Critics note the postwar golden age relied on cheap oil and on exclusion.

Check yourself

  1. List four features of the installation period she gives. Which one is most visible to you today?
  2. Why does she say the turning point is when installation’s problems “come to the surface”?
  3. Name one reason her timing claims are hard to test.

3. What a golden age is: three advances and the proactive state

🎙 In the talk. Deployment is “the golden age of creative construction”, where every innovation breeds more upstream and downstream, because “the new technology is already the common paradigm” and a proactive state promotes “economic and social progress by setting up a win-win game between business and society”. “Production reigns rather than finance with clear dynamic demand and stable reliable regulation” [07:45]. Her inequality evidence: the income earned by top 1% of taxpayers… is high in installation periods and it comes down in deployment… we went from the top 1% having 20% of income to the… bottom 50% getting [the] 20% [08:34]–[09:46]. She names three previous golden ages: the “Victorian boom in mid-19th century”, the “Belle Époque [and] progressive era in the US”, and the “postwar golden age” [09:46]. Each is “about the perception and reality” of three advances [10:42]:

  1. Social progress → social peace. A growing middle class and the feeling that “everybody’s moving up”.
  2. Economic growth “based on guaranteed and increasing dynamic demand” and “multiplication of opportunities for business up and downstream”.
  3. Social justice. “Increasing rights for a greater proportion of the population and social trust in the fairness of the system” [11:40]. “It’s a feel-good time”, like the French “glorious thirty years”.

📚 Deeper theory

  • The three golden ages in history. Victorian boom (c.1850–73): railways, steam, free trade; also the Great Depression of 1873 that ended it. Belle Époque / Progressive Era (c.1895–1914): steel, electricity; trusts, labour laws, antitrust, public health. Postwar (1945–73): the Fordist compromise. Economic historians usually call the third the “Golden Age” (Marglin and Schor, eds., The Golden Age of Capitalism, 1990; Eichengreen, The European Economy since 1945, 2007).
  • The “perception and reality” point. Her wording is that golden ages are felt and real [10:42]. This is close to what political scientists call legitimacy (Weber; Habermas on legitimation) and what Polanyi would call social protection. It is also close to Hirschman’s tunnel effect (1973): when others move up, you tolerate inequality because you expect your turn.
  • Why inequality falls in deployment: three standard explanations. (a) Policy and institutions: union power, taxes, social insurance (Piketty; Goldin–Margo’s Great Compression). (b) War and shocks: capital destroyed, labour scarce (Scheidel’s The Great Leveler, 2017). (c) Technology demand: skill-biased demand early, mass-education supply later (Goldin and Katz’s “race between education and technology”). Perez emphasises (a) and, via dynamic demand, (c).
  • A caution on the chart. US top-1% share hit about 20% in 1928 and again in the 2010s, per Piketty–Saez data, and the bottom-50% share was around 20% in the 1970s (from memory; I did not open the source). It is plausible that she is showing US data, but the talk does not say, and “the top 1% having 20% of income” to “the bottom 50% getting the 20%” is a compression of two series into one sentence [09:46]. Treat the numbers as indicative.
  • The proactive state in each age. Victorian: the Factory Acts, Poor Law reform, railway regulation. Progressive: antitrust (the Sherman Act 1890, the Clayton Act 1914), food and drug law, income tax. Postwar: the GI Bill, the Interstate Highway Act (1956), FHA mortgage insurance, Social Security (from memory; she names a different list in deep dive 4).
  • Compare “creative construction” with “endogenous growth”. Romer (1990): ideas are non-rival, so new ideas make more ideas. Her “every innovation brings more innovations” is the same loop, but she adds that it needs a shared paradigm.

Where others push back

  • Golden age for whom? The postwar boom excluded women’s equal pay, Black Americans in many programmes (redlining; the GI Bill’s uneven delivery), and colonial subjects. Feminist and postcolonial writers query the nostalgia.
  • War and reconstruction. Europe and Japan grew fast partly because they were catching up after war damage (Eichengreen; Crafts and Toniolo). So the golden age was not only a product of institutions.
  • Market critics. Some economists read the 1970s stagflation as a failure of the postwar state, not of technology (Hayek, Friedman, and later public-choice work).
  • Ends in the same place? Authoritarian “golden ages” exist (see 2.8). A golden age does not guarantee democracy.

What she leaves out

  • Costs to nature. The postwar boom’s pollution and oil use. She raises this only as a goal for the next age [20:19]–[22:03].
  • The causal order. Does the state create the golden age, or does the golden age’s prosperity make the welfare state affordable? She says the first. Her own examples (e.g., suburbs, mortgages) can be read either way.

Check yourself

  1. What are the three advances? Give a one-line sign of each.
  2. What does “production reigns rather than finance” mean in practice?
  3. Name one standard explanation of falling inequality that she does not stress.

4. Dynamic demand and the postwar welfare state

🎙 In the talk. “The issue of dynamic demand is crucial for golden ages. It is not just quantitative like the Keynesian notion of aggregate demand. Nor does it respond to [Say’s] idea about supply creating its own demand. It doesn’t exactly happen like that. It is demand that responds to the nature of the new technologies and grows at the same rhythm as the potential production capacity of the core industries” [12:17]. It is “enabled by implicit or explicit government policy”.

For the postwar West she names two main directions: suburbanisation (“the automobile created conditions… cheap land… a cheap house on cheap land”, which guaranteed markets for mass-produced consumer goods and required “the welfare state and public infrastructure”) and the Cold War (guaranteed demand for military goods and technology, “public procurement and funding of science”) [13:20]. A third: subsidised agriculture, which freed income for manufactured goods and made markets for farm machinery and petrochemicals [14:14].

Then she lists how the welfare state made mass consumption [14:14]–[16:05]:

  • public infrastructure for suburbia
  • public backing for mortgages, so banks would lend to wage-earners
  • unemployment insurance, so credit and mortgage payments continued
  • pensions, so people “spend salaries without fear of old age”
  • strong unions, so salaries rose with productivity (“the unions forced this which was good for business but they don’t necessarily understand that that is so”)
  • time for leisure and shopping (“you really need time in order to be able to consume properly”)
  • free education and health, so wages could go to consumption
  • farm subsidies
  • the Marshall Plan and the World Bank, for world demand and infrastructure [16:05]

She adds that mass production thrived on national markets: trade was under 20% of world GDP [16:05]–[16:45].

📚 Deeper theory

  • What “dynamic” adds. In static terms, demand is a level. In her usage it is a growth path matched to a supply path. If the core industries double productivity every decade, demand for their goods must double too. Otherwise firms face overcapacity, profits fall, and finance flees to speculation. This is the realisation problem (Marx; Luxemburg; Keynes) in a growth setting. Compare Kaldor–Verdoorn and Pasinetti (2.5).
  • The Fordist compromise (2.4). Wages rise with productivity, so workers can buy what they make. Evidence: in the US, real wages and productivity moved together roughly until the 1970s, then they diverged (Economic Policy Institute data; contested: Lawrence 2016 argues the gap is smaller after adjusting for prices and depreciation; from memory).
  • Housing and the car as the demand engine. Suburbs need cars, roads, houses, appliances, furniture and fuel; the demand chain is long. Kenneth Jackson, Crabgrass Frontier (1985), is the standard history. Public mortgage backing: the FHA (1934) and the 1944 GI Bill.
  • Military Keynesianism. The Cold War created a permanent demand for aerospace, electronics and computing. Much of the early computer and internet industry was funded by the Pentagon (Mazzucato 2013; Fred Block’s “developmental network state”). This is a clear case of the state steering a revolution.
  • Say’s Law and Keynes. Say: no general glut. Keynes: glut is possible without policy. Perez: not only possible, but the composition of demand has to match the new technology. A relevant line: Hyman Minsky’s “big government” and “big bank” as stabilisers.
  • Why she says it needs “policy”. Markets alone give demand to whoever has money. In installation, income concentrates (the top 1% in the chart), and the rich do not buy mass-produced goods in mass. So wage and welfare policy create the mass of buyers. This argument is more familiar from underconsumption theory (Hobson 1902; Keynes) than from standard growth models.
  • The “consumption” worry. She wants demand growth, yet also says “waste as a form of market expansion” was the old golden age’s weakness [22:03]. That is the tension between growth and the environment, which she resolves by changing what is consumed (services, rental) (see deep dive 6).

Where others push back

  • Supply-side view. Growth comes from investment, skills, and innovation; demand adjusts. Pushing demand without supply reform makes inflation (the 1970s).
  • Chicken and egg. Did welfare spending create the demand, or did the booming economy fund the welfare state? Both.
  • Public choice. State-guaranteed demand (defence, farms) creates lobbying and waste.
  • Fiscal limits today. Western public debt is higher and population older; a new welfare state is costly. She calls an update “an emergency” [23:32] but gives no financing.
  • “Demand for what?” Her list of growth sectors in the talk (AI, smart plants, green energy, biotech, microelectronics, new materials [19:15]) is a wish-list. The postwar boom had a concrete mass good (the house and car). What is the mass good of the information revolution? She suggests services [22:03] without naming one product.

What she leaves out

  • A measure. How would we know dynamic demand is on track? No indicator is given.
  • The household debt part of postwar demand (mortgage credit), which later helped create the 2008 crisis.
  • Women’s labour. The postwar demand model assumed a male breadwinner.

Check yourself

  1. How is dynamic demand different from Keynes’s aggregate demand and from Say’s Law?
  2. Pick two items on her welfare-state list and explain how each created demand.
  3. Why was the postwar revolution “national” and the information revolution “global”? What follows?

5. The road ahead for China and the West

🎙 In the talk. The information revolution is globalising: “globalized markets actually need world demand” [16:45]. It “is likely to require subraational [supranational?] arrangements. Right now it looks impossible but if we don’t have it it might be very difficult to have a golden age with a globalizing paradigm” [17:38]. The new dynamic demand “must be for capital goods and services” [17:38]. China “is making the leap in both revolutions” and so needs “dynamic demand in basic consumer goods and services too… for the old and for the new” [18:19].

The problem today. “China and the west are competing for their own insufficient markets. They have over production of things that they cannot consume themselves and exports are not absorbing yet.” The overproduced items: “capital goods and services, metals, especially steel but high-tech consumer goods, infrastructure projects, green tech, high-tech pharma” [18:19]. “Currently, it is a zero-sum game that could lead to extended conflict” [19:15].

The way out. “Massive new growing demand growing markets are required.” Friend-shoring “to favor the development of the global south” [19:15]. Demand for “energy, food and resources for green development”, for “enhancing the purchasing power of the global south”, and for “high level services and capital goods, AI, smart plants, green energy, biotech, microelectronics, new materials” [19:15]. Her critique: China’s Belt and Road and friend-shoring investment have started, the West “is beginning to follow, but neither seems to have taken on board the dynamic demand issue as a clear goal” [20:19]. Her proposals: consumer goods assembly with lower-cost labour in the south, rental and recycling (“we can’t have every home in the world have a refrigerator if… we are throwing them out”), and processing minerals in situ, near the mine, “because we have local renewable energy” [20:19]. Also, all countries must raise domestic demand, which “depends on salary levels, on free time to consume and on the security of the welfare state” and on the “productivity of the services of the welfare state itself” [20:19]. “The successful welfare state set up by the west for the mass production revolution is no longer adequate whereas China is broadening its own” [20:19].

📚 Deeper theory

  • Global imbalances. China runs large surpluses with low household consumption (about 40% of GDP, vs ~65–70% in the US; from memory, not checked). Economists (Michael Pettis, Trade Wars Are Class Wars, 2020; Matthew Klein) say the cause is a mix of weak wage share, high savings and financial repression. The IMF’s 2026 advice matches her diagnosis: consumption-led growth, stronger social protection to “lower precautionary savings” (IMF). Note a limit: the IMF summary I read does not use the word “overcapacity”.
  • Why a zero-sum game? Two big exporters sell similar goods (steel, EVs, solar) into limited markets. Trade barriers rise (tariffs). That is the fallacy of composition problem in export-led growth: not every country can run a surplus.
  • Global-south demand. The south’s potential is large but constrained by debt, currency and infrastructure. Her idea resembles the Brandt Report (1980) and the “global Marshall Plan” proposals (Al Gore, 1992; Mazzucato). It also resembles a “Green New Deal for the south”. The money question is the hard part, and the talk does not address it.
  • Rental and circularity. Selling the service of a product (rent a fridge) rewards durable design (Stahel, 2.7). The firm earns more if the product lasts. Practical examples: Rolls-Royce “power by the hour” jet engines, and Michelin tyre-as-a-service (from memory).
  • In-situ processing. Moving smelting and refining to resource-rich countries (Indonesia’s nickel export ban from 2020 is a recent example; Chile and Zimbabwe lithium moves). Critics note capture by foreign firms and heavy environmental cost; India has parallel debates about critical minerals.
  • Supranational arrangements. Candidates: a global tax floor, a digital trade regime, a green-transition fund, an IMF/World Bank reform. Rodrik’s trilemma says each reduces national autonomy. She says it “looks impossible” [17:38] and offers no design.
  • Forging ahead and Thucydides. Britain to the US was peaceful in the main; Germany versus Britain was not. Her fear of “extended conflict” [19:15] is the second case.
  • What China says about itself. China’s own plans (the 15th Five-Year Plan, “new quality productive forces”, “dual circulation”) stress domestic demand and technology self-reliance. The IMF notes the Plan’s focus on consumption. I did not read the Plan.

Where others push back

  • The zero-sum frame is the mainstream view in Washington and Brussels, and is not accepted in Beijing (“overcapacity” is called a Western label). On one side, firms say Chinese subsidies are the problem; on the other, China says the West’s own protectionism is the problem. Her “win-win” is a third position.
  • Friend-shoring is not what she means. In policy, “friend-shoring” (a US term from 2022) means moving supply chains from rivals to allies. She uses it to mean investment into the south. The two uses conflict.
  • Dependence. Some southern scholars say that “developing the south” via rich-country demand repeats a colonial pattern (dependency theory: Prebisch, Singer, Frank). Self-directed industrial policy may do better.
  • Climate limits. Rising mass consumption in the south may still raise emissions. Her answer is green growth and rental, which is a hope, not a proof.

What she leaves out

  • India, Africa and the south’s voices. The talk addresses only China and the West. The global south is a demand source, not an actor.
  • Debt, currency and tariffs. Who finances the new demand? What if the US and EU keep raising tariffs?
  • Geopolitics of chips and AI controls.
  • Whether China’s model is a success of the state or the market: relevant to what the West should copy.

Check yourself

  1. Why does she say the information revolution “needs” world demand when mass production did not?
  2. What does she mean by “making the leap in both revolutions”, and why does it matter for China’s welfare state?
  3. Pick one of her proposals (rental, in-situ processing, friend-shoring). What would block it?

6. What AI is for: the good life of services, and the welfare state as innovation

🎙 In the talk. “Here at last I talk about AI. AI can play a central role. The good life with mass production was about mass consumption of manufactured goods with waste as a form of market expansion. The good life with information technologies can be about optimal private and public services while protecting the planet. A whole new world of innovation in government and social services, health, education, longevity, food production, sustainability, etc. is ready to be created in all countries with the help of artificial intelligence” [22:03]. The welfare state “must also do both” (modernise the old and advance the new): “It has been catching up and can now forge ahead, reaching towards the affordability of a new good green life for all while guaranteeing dynamic consumer demand for producers” [22:47]. “For the West, bringing the welfare state up to date is an emergency to overcome social unrest and confront populism. For China, establishing a universal welfare state is an essential complement of the successful forging ahead process for all. It is an immense innovation opportunity” [23:32].

📚 Deeper theory

  • A “services” golden age. This links to Baumol’s cost disease (1966): services like health and education resist productivity gains, so their cost share keeps rising. If AI can raise service productivity, it could break the pattern. Evidence is early and mixed; some trials report gains in customer support and coding; others report little change.
  • Public services as the demand base. Here the welfare state is both the buyer and the innovator, in a way that echoes Mazzucato’s “mission” approach. Compare the Cold War as demand source [13:20]: now health and education become the missions.
  • Digital public infrastructure. India’s Aadhaar, UPI and similar systems are one working example of public digital rails for services (not in the talk; my addition). Whether this model generates “dynamic demand” for domestic firms is an open question.
  • Longevity and ageing. The talk says “longevity” among the targets [22:03]. Ageing raises welfare demand in the West and China; a services economy of care is a plausible mass sector.
  • AI and jobs. She says nothing on job loss. Acemoglu’s warning (2.7): AI built to substitute labour may cut wages. The direction of AI (augment vs replace) is the policy question. Her “directionality” fits, but she does not say how to steer it.
  • The risk of an AI bubble. If AI is in the installation frenzy, a crash could come before the services economy. Her model says the crash can help (infrastructure is left behind) but the pain is real. The talk does not mention the AI investment boom at all. Compare the FT headline (BizNews repost).

Where others push back

  • Techno-optimism of public services. Past digital government projects often ran over budget and failed (a pattern known from UK, US and Indian cases). AI may repeat this without institutional capacity.
  • The “waste as market expansion” point is a contested history. Planned obsolescence is real (the Phoebus cartel, 1920s–30s), but much consumer spending was also real welfare gain.
  • Service demand needs income. A service economy of care pays low wages unless policy changes that. She names salary levels [20:19] but does not say how.
  • Rights and surveillance. AI in welfare administration has produced wrongful denials and bias (e.g., the Dutch childcare benefits scandal; Robodebt in Australia — from memory; check before citing). The talk is silent on this.

What she leaves out

  • Any detail on how AI delivers services, who owns the models, or what regulation is needed.
  • Energy and water use of AI, though she cares about the planet [22:03].
  • The timeline: will deployment come in five years or twenty?

Check yourself

  1. What does she mean by “waste as a form of market expansion”?
  2. Why does she call the welfare state both a demand source and an innovation opportunity?
  3. What would you need to see to believe AI is moving into deployment and not still in the frenzy?

Putting it together

The mental model in one paragraph. A technological revolution is not a gadget but a package: new industries, a cheap input, a new infrastructure, and a new common sense about how to be efficient (Kuhn, Dosi, Freeman). It spreads on an S-curve in two halves. In installation, finance rushes in, builds the infrastructure fast, polarises income and ends in a crash (Schumpeter, Minsky, Kindleberger). At the turning point, society is unstable and must decide how to align finance with production (Polanyi). In deployment, a proactive state makes sure demand grows as fast as the new industries can produce, and the gains are shared, so a golden age follows (Regulation school, Keynes, Hirschman). Perez says the fifth revolution is near that choice point. AI is a technology inside it, and its fate depends on the whole package. The global twist is that this revolution is global, so demand has to be built across borders, for capital goods and services, with a stronger welfare state in China and a modernised one in the West. History, she says, is a useful guide [24:15], not a forecast.

Three tensions to keep open

  1. Pattern or story? Five cases fitted to a template, with soft dates and a moving turning point (Part 3, deep dive 2). Is it a theory or a lens?
  2. State as engine, or state as result? Does the proactive state create golden ages, or does a booming economy make it possible (deep dives 3 and 4)? Acemoglu says it is political choice; others say it follows from growth.
  3. Is the postwar template transferable? Her golden ages were national or (in the first) imperial, with war and reconstruction in the background. The information revolution is global, intangible and winner-take-all. A new template may be needed, and she says so (“an equivalent but different set of policies”, [17:38]).

Reading path

  • Start: Perez, Technological Revolutions and Financial Capital (2002), chapters 1–3 and the last two. Then her 2015 paper on “great surges” (PDF). Ben Thompson’s Stratechery piece is a short, readable bridge.
  • Then: Schumpeter, Capitalism, Socialism and Democracy, chapters 7–8. Kindleberger, Manias, Panics and Crashes. Polanyi, The Great Transformation, Part 1. Paul David, “The Dynamo and the Computer” (1990). Acemoglu and Johnson, Power and Progress.
  • Deeper: Freeman and Louçã, As Time Goes By. Minsky, Stabilizing an Unstable Economy. Aglietta, A Theory of Capitalist Regulation. Mazzucato, The Entrepreneurial State. Gordon, The Rise and Fall of American Growth (the sceptic). Pettis, Trade Wars Are Class Wars (for the China demand question). The IMF’s 2026 China conclusion (IMF) is the current data check.

Not confirmed in this page. The exact talk date and host (taken from Perez’s site only). The meaning of “subraational” and “C2” in the captions. The chart’s source and country at [09:46]. The FT column’s body text. Items marked “from memory” in Part 2 and Part 3 (book dates, thinkers’ dates, side facts such as Rolls-Royce and Michelin examples, household consumption shares). Check these before quoting them.

Q&A

(Add questions and answers here only if worth keeping.)

Graph View

Backlinks