At CES 2026, the world's largest electronics and IT exhibition held in Las Vegas last month, Siemens drew attention by presenting a new form of manufacturing plant. The German company plans to introduce "Agentic AI" at its Erlangen factory to handle the entire process—from design, procurement, and logistics to production and quality control—in place of humans. This is fundamentally different from the Adidas smart factory that closed in 2019. The AI at the Adidas facility was at an early stage, merely connecting robots that repeated predetermined movements. In contrast, the Erlangen factory's AI organically controls robots that can make judgments autonomously and work like humans. If the Erlangen factory operates as planned, it is expected to become a completely different type of facility from existing factories. The door is opening to the "dark factory"—once thought impossible—where only a small number of supervisors exist and workers are absent.
How would Adam Smith, who laid the foundation of capitalist economics with "The Wealth of Nations" some 250 years ago, react if he saw a "dark factory"? The most advanced machines Smith ever witnessed were merely spinning machines and steam engines, so he would likely be astonished to see a factory producing goods without human workers. But the greater fear would be having to confess that the "labor theory of value"—which he proposed and countless economists after him accepted, holding that "all wealth originates from labor"—"may have been wrong."
Until now, even as technology advanced and superior machines were deployed in factories, the basic premise of the labor theory of value was never completely negated. However, as humanoid robots enter factories and the main agents of production shift from humans to robots, we must consider how to explain product prices when human labor input converges to "zero" yet prices remain stable.
It is not just the labor theory of value. The theory of marginal productivity could also be shaken. This theory is based on the premise that "labor and capital are each compensated according to the additional production (marginal contribution) generated when additionally invested." However, the performance of AI systems operating in factories is difficult to divide into individual contributions. Data, algorithms, graphics processing units (GPUs), computing power such as cloud services, and on-site application and operation must be combined as a single unit. When the source of efficiency comes not from "individual additional contribution" but from "the combination of the entire system," the traditional formula of "compensation according to contribution" becomes difficult to apply in reality.
David Ricardo's comparative advantage theory explains that differences in opportunity costs between countries generate trade. Thus, labor-intensive industries develop in underdeveloped countries while service and knowledge industries flourish in advanced nations, maintaining trade between them. However, in the era of humanoid robots and AI, competitiveness is increasingly likely to depend on whether a country possesses an ecosystem of hyper-scale AI companies like Google and Nvidia rather than natural resources or labor costs. As production increasingly occurs in countries with such ecosystems, polarization between nations will only intensify.
Additionally, the Keynesian effective demand theory—which holds that "demand, not supply, ultimately drives the economy"—faces criticism that while humanoid robots can rapidly expand production capacity, reduced wage income for humans means consumption cannot keep pace, potentially making "demand shortage" chronic. The Phillips curve, which suggests that "when unemployment falls, wages and prices rise, and when unemployment rises, they fall," may also flatten. This is because humanoid robots effectively increase labor supply, weakening upward pressure on wages. Particularly if the linkages between prices, wages, and employment change, policy prescriptions from central banks and governments will inevitably require redesign.
Furthermore, Joseph Schumpeter's "creative destruction"—the idea that "innovation tears down old industries while creating new ones"—and human capital theory, which holds that education leads to productivity improvement (technology accumulation) and higher wages, also become subjects for revision.
"Traditional economics holds that as scale increases, marginal productivity declines and exchange value arises in the market," said Cho Joon-mo, professor of economics at Sungkyunkwan University. "In the AI era, winners like Google and Nvidia supply goods while also becoming monopolistic leaders. This situation is tantamount to destroying all traditional economic logic."
The movement to create dark factories is not merely a question of "how smart can factories become." It is also a signal that we must establish new principles together, as labor markets could be shaken before new economic orders are even discussed. "For capitalism to be healthy, social safety nets and market economy must be guaranteed together," Cho said. "It is urgent to prepare a blueprint for how new AI capitalism will develop and how Homo sapiens should respond to this capitalism."
