Capitalism in the AI-Powered Economy
Market structure and institutions will determine whether surplus flows to capitalists, consumers, or workers
Last week I argued that AI-induced economic surplus would flow differently across sectors: largely to consumers in services, except where regulatory capture enables rent-seeking, whilst AI-savvy manufacturers would be better positioned to retain surplus and reinvest it. The distinction, I suggested, ran along the manufacturing-services divide.
That framing was too simple. Both situations coexist. Some service sectors exhibit increasing returns to scale that allow firms to retain surplus and compound value. Conversely, some manufacturing sectors face such intense competition that firms must cut prices, passing gains to consumers. Thus the real divide is not between manufacturing and services, but between businesses that can capture producer surplus and those that cannot. AI creates supply-side value only when producers retain part of the surplus—a dynamic that exists in some manufacturing contexts but not all, and in some service sectors but not most.
This edition refines last week’s argument. It examines how AI will reshape capitalism itself: strengthening some firms’ ability to escape the market economy’s competitive pressures, whilst trapping others within it. It explores three pathways for AI to enter legacy value chains, analyses why client-facing services cannot scale like software, and considers how AI might transform proximity services and the social contract. Finally, it assesses how these dynamics will play out across the three major economic blocks: the US, China, and Europe.
The stakes are high. AI will reshape not just industries but the formula for generating prosperity itself.
1/ In last week’s edition, I tried to answer a set of simple questions
How will the deployment of AI translate into widespread prosperity? Through which sectors, and in which regions?
I received some pushback, mostly regarding my attempt to draw a line between services on one hand and manufacturing on the other. Specifically, the competitive analysis I outlined implied that the AI-induced economic surplus in services would largely flow to consumers—except where regulatory capture enables rent-seeking—whereas AI-savvy manufacturers would be better positioned to retain a significant share of that surplus, allowing them to reinvest and compound value creation.
In fact, both of those situations coexist:
Some service sectors exhibit increasing returns to scale that confer a durable competitive advantage, enabling firms to retain most of the surplus and reinvest it.
Conversely, some manufacturing sectors face such intense competition that firms must cut prices to maintain market share, passing nearly all the surplus to consumers.
In other words, the distinction is not as clear-cut as I initially suggested. The divide between high- and low-producer-surplus businesses is orthogonal to the manufacturing–services split. AI will create value on the supply side only when producers can capture part of the surplus—but this is not unique to manufacturing, and there are manufacturing contexts where it is not possible1.
What I am offering today is therefore a refinement of last week’s argument. The previous edition still stands as useful scaffolding, but this version presents a more definitive and self-contained analysis.
2/ At the highest level, this is a discussion about capitalism itself
My friend Bill Janeway, with whom I had lunch earlier this week, reminded me that Peter Thiel’s remark that “competition is for losers” ultimately traces back to Adam Smith and The Wealth of Nations. Yet the (more recent) framework I find most useful comes from the French historian Fernand Braudel, who divided the economy into three distinct layers:
Material life, the largest and least measurable layer, composed of the countless informal transactions that make up daily existence.
The market economy, where individuals and small firms trade goods and services, linking production to consumption. It corresponds today to the world of freelancers, artisans, and small family businesses who operate on razor-thin margins because competition is relentless. It is the domain of merchants who buy and sell, earning only modest spreads as they perform services and/or sell goods that merely pass through their hands.
Capitalism, a set of processes that allow actors to rise above the day-to-day grind of commerce and pursue increasing returns to scale. Unlike the merchant’s world, capitalism is that of financiers and entrepreneurs who realise they can weaken market competition by deploying capital within production itself, thus generating those crucial increasing returns.
When increasing returns produce a surplus that a firm can retain—having freed itself from the pressures of the market economy—the result is compounding: more capital leads to a larger surplus, which can then be distributed among shareholders, workers, consumers, and the state. In turn, this enables further investment and greater value creation. Capitalism, in this sense, is the only true engine of prosperity. Its output is a rising return on invested capital, which drives higher labour productivity—the single economic indicator most closely correlated with long-term development2.
3/ The question then becomes: how will AI affect capitalism and reshape the formula for generating prosperity?
The first dimension concerns AI being used to strengthen existing industries. As Dan Wang, author of Breakneck, wrote this week in the Financial Times, we should expect this to be AI’s primary role over the coming years:
AI is not a simple “race”. What matters is not just creating the technology but what each country does with it. Silicon Valley has been obsessed with superintelligence, as if it were possible to build God in a box. Beijing has been interested less in treating AI as a supernatural goal, and more as a technology to be harnessed — Chinese academics and policymakers consistently talk about AI as a practical tool for enhancing existing industries.
AI will help both countries deepen their specialisations. Think about it this way: America is better in service sectors such as consulting and litigation; with AI, it might be able to generate even more lawsuits. China, which has far superior training data on manufacturing, might grow even better at producing electronics, drones and munitions.
This idea of AI deepening and entrenching existing positions suggests that some firms in sectors with high barriers to entry may further consolidate their position and capture an even larger share of economic surplus. In relatively stable industries where competition has been softened, AI will enable firms to reinvest retained surplus rather than being forced to cut prices.
In contrast, in highly competitive merchant sectors, AI-induced gains are likely to flow immediately to consumers via lower prices, because relentless competition prevents firms from retaining much surplus.
It is notable that Dan uses professional services (dominant in the US) and manufacturing (China’s strength) as examples. This explains why, over the past few years, Western debates have focused on AI destroying white-collar jobs—lawyers and accountants first—while China has quietly applied AI to manufacturing, where there are fewer jobs to displace and substantial productivity gains to be secured. For China, that is even an advantage given the demographic challenges it faces.
The second dimension is whether AI will reshape the boundary between capitalism and the market economy. In some sectors, AI can shift the balance, enabling some firms to retain more surplus and reinvest, while others see their historical advantages eroded:
Will AI allow firms in competitive markets to escape the constant pressure of price competition and enter capitalism’s compounding cycle?
Conversely, will AI undermine entrenched capitalist advantages, eroding incumbents’ market power as new entrants leverage AI more effectively?
The third dimension concerns the geographic distribution of wealth and power. Several factors come into play: the availability of capital, access to the compute-energy stack—from electricity and data centres to chips—state support for capitalist expansion through industrial policy, the regulatory approach to capitalism’s potential dead ends (a race to the bottom where no firm profits, or dominance by a single firm that can exploit customers), and the broader institutional context. AI will shape which paths sectors and regions follow, depending on how it interacts with existing market and capital structures.
The next sections will examine these dimensions from different angles and conclude with a provisional assessment of where various regions stand within the emerging AI economy.
4/ There are three ways for AI to enter legacy value chains
About ten years ago, I designed a simple model to understand how technology penetrates legacy industries and showed how incumbents find new reasons to be in denial at every stage. The first stage is when incumbents do not care at all. For AI, this stage was brief, as the hype hit the market with full force from November 2022 onward. The second stage is where we are now: incumbents convince themselves they are experimenting with the new technology, while in reality they mostly run small-scale projects with consultants that have no impact on the core business3.
Even though it’s still early, I believe we have seen enough to be able to conclude that AI will penetrate industries in three different ways, depending on the industry’s economics as well as its regulatory and institutional context: