The United States today is witnessing a new era of concentration and centralization of monopoly-finance capital marked by the artificial intelligence (AI) boom. S&P Global economists estimate that “80% of the increase in final private domestic demand” in the United States in the first half of 2025 was attributable to spending on “data centers and related high-tech capital expenditures.”[1] This massive investment in data centers is being carried out by giant high-tech corporations, the number of which could easily be counted on the fingers of one hand. These firms are commonly referred to in the industry as “hyperscalers,” standing for the mega-corporations that dominate cloud computing. Ranked according to data-center investment in early 2026, they include Microsoft, Amazon Web Services, Google (Alphabet), and Meta, making up “the Great Houses of AI.”[2] These giant monopolistic entities are also all among the top six U.S. corporations, measured by market value. (Nvidia, the largest firm by market valuation at the beginning of 2026, is not itself a leader in cloud computing, but instead monopolizes 80 to 90 percent of GPU supercomputer chips.) According to Bloomberg, Microsoft, Amazon Web Services, Alphabet/Google, and Meta, had a combined capital spending of $150 billion in 2022 and $360 billion in 2025, while they plan to spend $650 billion in 2026. By comparison, “the largest US-based automakers, construction equipment manufacturers, railroads, defense contractors, wireless carriers, parcel delivery outfits, along with ExxonMobil Corp., Intel Corp., Walmart Inc., and the spun-off progeny of General Electric—21 companies—are projected to spend a combined $180 billion in 2026.”[3]

Investment in AI is now on a scale that invites comparison with the U.S. railway boom in the nineteenth century.[4] As in the case of the railroads, AI expansion today is backed by centers of finance manipulating government support, freeing it from dependence on actual profits, while relying instead on what John Maynard Keynes called “animal spirits,” or expected profits on new investments. It would have taken many years for the hyperscalers to increase their data center investments to the current level based simply on the accumulation of actual profits, while monopoly finance via the credit-debt system has allowed this transformation to occur in “a twinkling of an eye.”[5] Social wealth, drawn from the population as a whole, is being channeled to the Great Houses of AI through a variety of financial mechanisms and neoliberal economic policies, further concentrating the economic surplus produced by society in the hands of an infinitesimally small number of billionaires, located in the high-tech, energy, and finance sectors of the economy. Nine of the top fifteen billionaires on the 2026 Forbes billionaire list are tech billionaires.[6]

The rush to build massive data centers, the largest of which occupy millions of square feet and consume gargantuan amounts of energy, water, and mineral resources, is driven by the goal of developing advanced forms of generative AI, a type of machine learning capable of replicating human intelligence while drawing on seemingly unlimited data. This offers those who own, manage, and profit from such immense computational systems the prospect of complete surveillance and discipline (in the Foucauldian sense) of the population as a whole, not only in workplaces and prisons but in all activities of life, in such a way as to extract ever larger shares of the economic pie. Here the famous adage commonly attributed to Francis Bacon, “knowledge is power,” takes on new meaning. As Oracle CEO Larry Ellison has said, these technologies allow for tracking and targeting everyone all the time. “Citizens will be on their best behavior, because we’re constantly recording and reporting everything that’s going on. And it’s unimpeachable…because AI is monitoring the video.”[7]

Not only does generative AI point to vastly increased surveillance of human activities throughout society, but it also massively threatens employment, with tens of millions of jobs potentially being lost in the United States alone, according to some estimates.[8] In February 2026, Mustafa Suleyman, CEO of Microsoft AI, exuberantly told the Financial Times: “White-collar work, where you’re sitting down at a computer, either being a lawyer or an accountant or a project manager or a marketing person—most of those tasks will be fully automated by an AI within the next 12 to 18 months.”[9] What makes this possible of course is the theft by AI of all past intellectual labor. At the same time, the AI rush presents unimaginable environmental dangers through the hyper-expansion of data centers, drawing upon exponentially rising rates of energy, water, and other resources, thereby setting aside the transition away from fossil fuels, and threatening a vast acceleration of carbon emissions and global environmental damage. What makes AI expansion in these extreme terms seem unstoppable is a technological determinism rooted in a fetishism of AI, in which it is seen as embodying a pure computational logic, combined with the naturalization of market relations, which suggests that the new technology will inevitably be subordinated to the interests of capital accumulation.[10] Indeed, it is the advent of AI as a new regime of computational power controlled by monopoly-finance capital that is the emerging matrix of class (and imperial) struggle in our time.

In reality, innovative forces of production, such as machine learning/artificial intelligence, are never to be conceived in simple technocratic terms, as in the ultimate dominance of AI’s “neural networks,” but rather have to be seen as articulated with the social relations of production. For Karl Marx, it was the combination of the forces and social relations of production in any given set of historical conditions that gave rise to the “social individual,” while automatic machinery pointed to the “general intellect” in which human knowledge was embodied in machine artefacts giving rise to the “collective worker.”[11] A socialist approach to AI therefore focuses above all on the historical and social relations that brought it about in conjunction with capitalism, thus demystifying the current AI fetishism and making it clear that the path ahead for humanity is ultimately up to us, requiring a struggle of revolutionary scale and content.[12]

Kate Crawford and the Mapping of AI

The foremost figure in socially mapping AI is Kate Crawford, a senior principal researcher at Microsoft Research and a research professor at the University of Southern California at Annenberg. Crawford adopts an approach that is historical, materialist, ecological, and concerned with mapping AI as a regime of power operating in conjunction with corporate hegemony, representing an era of “computational capitalism.”[13] Her work draws on a wide range of thinkers, including such figures as Charles Babbage, Marx, William Stanley Jevons, Max Weber, Lewis Mumford, Harry Braverman, E. P. Thompson, Stephen Jay Gould, Christian Fuchs, and Vandana Shiva, along with contemporary analyses of monopoly capital, global capitalism, and the metabolic rift. Crawford’s principal works on AI include (1) her interactive graphic “Anatomy of an AI System: An Anatomical Case Study of the Amazon Echo as an Artificial Intelligence System Made with Human Labor” (with Vladen Joler, 2018); (2) her book, Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence (2021); (3) Calculating Empires—a 24-meter fresco of AI (2023); (4) her Long Now Foundation lecture, “Mapping Empires” (2025); and (5) her article, “Eating the Future: The Metabolic Logic of AI Slop” (2025).[14]

The fetishism of AI, heavily promoted today by corporations and the media monopoly, is a reflection of what Crawford calls “enchanted determinism,” portraying AI as a “cloud” technology occupying some ethereal dimension, with only secondary connections to the material world and to the realm of production.[15] She inverts this dominant mystifying view, adopting a critical materialist perspective. “AI,” she writes, “is neither artificial nor intelligent.” Instead, it is “a registry of power.” Although utilizing the term “AI,” she defines it as a “massive industrial formation that includes, politics, labor, culture, and capital.”[16] As Tung-Hui Hu states in A Prehistory of the Cloud, “today’s dominant metaphor for digital space, ‘the cloud,’ is actually a metaphor for private ownership” and the exclusion of public material access.[17] In Crawford’s words, “Artificial intelligence…is an idea, an infrastructure, an industry, a form of exercising power, and a way of seeing; it’s also a manifestation of highly organized capital backed by vast systems of extraction and logistics, with supply chains that wrap around the entire planet.” She adds: “AI systems are built with the logics of capital, policy, and militarization and this combination further widens the existing asymmetries of power.”[18]

The concept of “enchanted determinism” is used to address the commodity fetishism and mystical godlike qualities imputed to AI. “AI systems,” Crawford explains, “are seen as enchanted, beyond the known world, yet deterministic in that they discover patterns that can be applied with predictive certainty to everyday life.” This enchanted determinism takes two main forms, each of which is dialectically related to the other. The first is a “tech utopianism,” while the second is a “tech dystopian” outlook. “These dystopian and utopian discourses,” she writes, “are metaphysical twins: one places its faith in AI as a solution to every problem, while the other fears AI as the greatest peril.” The answer to both is a historical and materialist critique that uncovers the social roots of AI and explains that it is ultimately a question of social relations, not simply technology. “The fantasy that AI systems are disembodied brains that absorb and produce knowledge independently from their creators, infrastructures, and the world at large…distract[s] from the far more relevant questions: Whom do these systems serve? What are the political economies of their construction? And what are the larger planetary consequences?”[19]

In exploring the various dimensions of AI, Crawford begins with the material base in the form of mining for lithium, cobalt, and rare earth metals. She explores the Silver Peak lithium mine in Nevada and Tesla’s nearby battery factories. Tesla is now tapping into a considerable portion of the planet’s lithium reserves.[20] The production of each metric ton (2,205 pounds) of lithium requires the evaporation of about 2 million liters (528,000 gallons) of water, threatening water tables and water supplies. At the level of extraction, the labor behind AI is rooted in the long history of colonialism and imperialism. Most of the extraction occurs in the Global South. In the cobalt mines in Congo, workers receive the equivalent of a dollar or two a day for working in inhuman conditions, exposed to toxic cobalt dug out with picks and shovels in trenches and tunnels. Workers have no alternative as “the mines have taken over everything.”[21]

In “Anatomy of an AI System,” Crawford and Joler, following Marx, present production at each stage of the overall process as based on the appropriation of “the surplus of value” over labor cost, out of which arises the profits of capital.[22] Capitalist AI is aimed at the displacement of highly paid labor and its replacement by a combination of machine automation and cheaper labor globally outsourced. The globalized nature of the AI system, with its complex supply chains, makes the overall transnational employment effects extraordinarily difficult to ascertain. Although aimed at displacing labor in the current centers of production, the hidden abode of AI is to be found in the hiring of masses of low-paid machine trainers, image taggers, and AI platform service workers, the actual existence of which dispels the myth of artificial intelligence. Thus, AI currently requires enormous numbers of “crowdworkers” involved in “crowdsourcing,” that is, online workers, commonly in their twenties and scattered around the globe, carrying out a kind of “ghost work.” For example, OpenAI in 2022 relied on outsourced workers in Kenya, who were paid less than $2 per hour to examine and label tens of thousands of toxic images and passages associated with child sexual abuse, bestiality, rape, etc., as part of “cleaning up” ChatGPT, while similar work was being done by outsourced workers in Uganda and India.[23]

Vast numbers of workers are used to monitor and adjust the content of AI chatbots. Jeff Bezos has cynically referred to this reality of workers behind the machines as “artificial artificial intelligence.” “Until there is another way to create large-scale AI that doesn’t use extensive behind-the-curtain work by humans,” Crawford observed in 2021, “this is a core logic of how AI works.” It should be remembered that, between 2005 and 2015, 94 percent of new jobs in the United States were for “alternative work” rather than traditional employment.[24]

While “intelligent machines” today require ghost work by crowdworkers located mainly in the Global South, Crawford also looks at the devastating role of AI and robots in existing industry. In Amazon warehouses, the labor process and time at work are hierarchically controlled as never before. The worker is not only an “appendage to the machine,” as Marx wrote, but increasingly an appendage to “intelligent” robots, while being placed under constant surveillance and control.

In this context, Crawford explores the late eighteenth-century innovations of engineer Samuel Bentham, who first conceived of the panopticon system for the surveillance and control of the movements of labor (later applied to prisons by his older brother, Jeremy Bentham).[25]

Computational capitalism, Crawford contends, is deeply rooted in and runs on the exploitation of human bodies through time and the imposition of work-discipline. She discusses Thompson’s work on how industrialization and capitalism changed time itself within work in the nineteenth century—then turns to Braverman’s critique of Taylorism and the degradation of the labor process under monopoly capitalism.[26] Algorithms now determine both the times and the spaces of the workers. The new world of AI algorithms represents the realization of the “real subsumption of labor” to capital discussed by Marx, as in the relentless domination of “the rate,” standing for the pace of work in Amazon warehouses. Here she quotes from Marx’s critique of capital’s time versus nature’s time found in Capital: “Time is everything, man is nothing; he is at the most, time’s carcase.”[27]

After approaching AI from the material ground, beginning with mining and the exploitation of workers in both extraction and production, Crawford proceeds to discuss the new regime of data that lies at the core of this new registry of power. The AI regime feeds on the notion that absolutely everything is data, which is to be harvested regardless of the social and environmental costs. The new computational capitalism promotes unrelenting data accumulation in the forms of text, image, sound, and video, with the entire human world serving as raw data for AI systems.[28] Social media platforms are conduits for gargantuan amounts of data feeding AI systems, which also penetrate into nearly every sphere of the commons and private life:

There are gigantic data sets full of people’s selfies, tattoos, parents walking with their children, hand gestures, people driving their cars, people committing crimes on CCTV, and hundreds of everyday human actions like sitting down, waving, raising a glass, or crying. Every form of biodata—including forensic, biometric, sociometric, and psychometric—is being captured and logged into databases for AI systems to find patterns and assessments…. Voice data is gathered from devices that sit on kitchen counters or bedroom nightstands; physical data comes from watches on wrists and phones in pockets; data about what books and newspapers are read comes from tablets and laptops; gestures and facial expressions are compiled and assessed in workplaces and classrooms….

Fundamentally, the practices of data accumulation over many years have contributed to a powerful extractive logic, a logic that is now a core feature of how the AI field works. This logic has enriched the tech companies with the largest data pipelines, while the spaces free from data collection have dramatically diminished.[29]

Data has to be categorized. The subjective impressions of crowdworkers are used to set up classifications of people on the basis of race, ethnicity and gender.[30] Race signifiers that are the product of historic racist classification systems are incorporated. Gender is always restrictively regarded as binary. As Crawford notes, “machine learning systems are, in a very real way, constructing race and gender: they are defining the world in the terms they have set.” The categories used in AI machine training and classification reinforce existing prejudices and perpetuate invidious comparisons, as well as replicating the dominant political-economic ideology.[31]

Although the promise of increased productivity through the more efficient and total exploitation of labor is the basis for claims as to the future profitability of AI systems, it is also based on the prospect of the extraction of profits from all forms of human action. The goal is to universalize exploitative/expropriative systems, promoting the accelerated amassing of capital and its further concentration and centralization by a few all-dominating companies that have become almost synonymous with “the market.”

On top of this sits the capitalist state, which monopolizes the laws of property and violence. The state is a major accumulator of data, working in conjunction with, rather than in opposition to, computational capital. The monopoly-capitalist state is heavily organized around military and police functions that grow hand-in-hand with surveillance capitalism within the private sector. For Peter Thiel, the founder of Palantir, and a key billionaire supporter of the Donald Trump administration, AI is essentially a military technology geared to surveillance and targeting, applicable both to war and to domestic control operations. “These tools,” he writes, “are…valuable to any army—to gain an intelligence advantage, for example,” while such “machine learning tools,” he adds, “have civilian uses, too.” During the first Trump administration, Palantir’s contracts with U.S. government agencies amounted to more than a $1 billion. Palantir has become a key outsourced surveillance company for Immigration and Customs Enforcement (ICE), aiding ICE in its racially motivated deportation campaign. According to a Bloomberg report in 2018, Palantir “is an intelligence platform designed for the global War on Terror,” which is mainly “weaponized against ordinary Americans at home,” working in conjunction with state agencies.[32]

Similarly, the Neighbors app, which relies on Amazon’s Ring doorbell cameras, classifies footage into categories such as “Crime,” “Suspicious,” or “Stranger,” and the videos are shared through contracts with police and ICE. Ring is also used to police workers who deliver packages. In the words of Tung-Hui Hu, such apps have become “freelancers” for the state’s military and security apparatus.[33]

Military use of AI is now pervasive, as used in drone warfare and cyberwarfare, and is now integrated into all war operations. In 2017, the U.S. Department of Defense launched its Algorithmic Warfare Cross-Functional Team, which was code-named Project Maven, aimed at using AI as an “automated search engine of drone videos” for surveillance and targeting. The initial contract going to Google led to over three thousand employees signing a letter of protest demanding that the contract be canceled. Google responded by turning the subject of the debate not into a protest of AI being used in warfare, but rather into a question of whether the technology was being used “to kill people incorrectly,” something that the corporation indicated could be avoided through the AI technology itself, which provides the basis for killing people correctly. The United States has utilized Claude, Anthropic’s AI model, as well as others, in its war on Iran, in alliance with Israel, which began on February 28, 2026. In the first twenty-four hours of the U.S. and Israeli attack on Iran, Anthropic generated as many as a thousand prioritized targets, in which it synthesized satellite imagery, surveillance feeds, and signals intelligence, providing GPS coordinates in real time for both human and strategic targets, while automating the legal justifications with respect to each strike.[34]

Yet the role of the state with respect to AI goes beyond its outsourcing of domestic surveillance, population control, and its military operations. The capitalist state has given a green light to a system of monopolistic computational capital aimed at unlimited data accumulation as a basis for boundless capital accumulation, with few, if any, real legal constraints. This reflects a government of the corporations, by the corporations, and for the corporations. The lack of state regulation has allowed the AI rush to proceed with little concern for the destructive consequences, from the prospect of a bursting of the AI bubble to eventual widespread social and ecological implosions.

AI and the Metabolic Rift

Crawford’s Atlas of AI was published in 2021, a year before the introduction of ChatGPT, which accelerated the AI craze and led to an enormous expansion of investment in data centers. Informed by these developments, Crawford’s more recent work focuses on the core contradictions of AI as a registry of power. In her 2023 “Calculating Empires” interactive artwork, she singles out monopoly capital and globalized capital as defining the political-economic mode in which digital-AI technology has emerged. However, the startling innovation in her 2025 “Mapping Empires” lecture is to focus on the internal and external contradictions of AI. Here she draws her core argument from the concept of metabolic rift, as developed in the nineteenth century by Marx, based in part on the work of the German chemist Justus von Liebig. Crawford, in her lecture, provides a detailed discussion of the rift in the soil nutrient cycle in nineteenth-century England, due to the sending of food and fiber containing soil nutrients, such as nitrogen, phosphorus, and potassium, to the new, highly populated industrial cities hundreds and even thousands of miles away, where these nutrients ended up as pollution, with people throwing “excrement into the streets and the waterways.” As a consequence, these essential elements were not returned to farms to replenish the soil. As Crawford herself puts it, “Europe was literally eating itself into depletion.” Here she draws on Liebig’s concept of Raubbau or the robbery culture/economy.

Given the general inability to produce synthetic fertilizers at the time, particularly those incorporating nitrogen, the “guano craze” set in, with European countries and the United States competing for guano (nitrogen-rich bird droppings). Massive quantities of guano were imported to Europe from the guano-rich Chincha islands off the coast of Peru. Although synthetic fertilizers were later developed, this merely shifted the contradiction around, leading to the present-day rifts in the nitrogen and phosphorous cycles, with the result that the general metabolic rift associated with a disjuncture between human resource extraction and the conditions of ecological sustainability only deepened. Today, the emergence of the Anthropocene is seen as representing an “anthropogenic rift” in the biogeophysical cycles of the Earth System.[35]

Recognizing that AI is a material system that has emerged historically as a result of human-social action and is an embodiment of natural and human relations, Crawford argues that it is necessary to view it as a metabolic system following “metabolic patterns” or cycles. Contradictions in the form of metabolic rifts necessarily arise between the conditions of material existence and reproduction and the internal imperatives of AI capital. Thus, the mining of essential materials and resources, unlimited “data ingestion,” and the final content in the form of “AI slop” can all be seen as phases in a metabolic cycle. This is driven by the imperatives of computational capitalism, leading at some point, since this is unsustainable, to “model collapse.”[36]

In Crawford’s conception, AI’s destructive data ingestion is equivalent to Raubbau. Mineral extractivism and energy and water use are exponentially raising demands on the natural environment, disrupting the human relation to nature on an accelerating scale, in line with Marx’s classic notion of the metabolic rift. Moreover, it is now recognized that there is a self-generating rift within AI, known in the scientific literature as “AI autophagy” (after dysfunctional metabolic autophagy—self-eating—in cells). Here AI, by relying more and more on its own synthetic data, or AI slop, essentially eats itself, leading to “model collapse,” with disastrous consequences for the entire AI-alienated world.[37]

AI data ingestion today is enormous beyond measure, already equal to what can be scraped from the web, encompassing countless terabytes of data and seeking to enclose the entire world of information in all of its forms. The totality of human creativity over thousands of years and all human behavior and expressions are grist for its mill—all to be incorporated into machine learning commanded by a system of political-economic power. All this, however, is materially embodied, placing limits on the system’s operations.

“AI’s mineral demands,” Crawford tells us, “are driving another metabolic rift, extracting the minerals that took billions of years to form in the earth’s crust in deep time for AI chips that are used generally for one to two years.” The biggest environmental costs associated with the new AI Raubbau, however, are energy and water consumption, which already point to levels of use comparable to those of the wealthiest countries. Estimates from the International Energy Agency and Bloomberg project that the amount of electricity needed for AI will be equivalent to that of states like Japan and India, or as much as 25 percent of U.S. electricity, by 2030.[38] Hyperscale data centers require cooling systems that consume millions of gallons of water daily, with the demand constantly increasing. None of this is sustainable. Although some say that greater efficiency can solve the problem, Crawford turns here to the famous Jevons paradox, based on William Stanley Jevons’s work The Coal Question (1865), in which it was argued that increased efficiency in the use of coal never reduced the amount of coal used, since the increased efficiency always led to expansions in the level of production—a phenomenon inherent in the system of capital accumulation.[39]

What Crawford refers to as an emerging metabolic rift, rooted in capitalist social relations, thus has to do with AI’s insatiable appetite, ingesting, digesting, and excreting data in ways that lead to its own cannibalization. As in the Greek myth of King Erysichthon as told in Ovid’s Metamorphoses—in which Erysichthon, consumed with desire for wealth and consumption, sold his own daughter and then ate himself—today’s AI systems, driven by capital accumulation and by their own inner technological logic, will end up consuming themselves.[40] Increasingly ingesting its own synthetic outputs, full of phantasmagoria and hallucinations, along with the general flattening of knowledge, the result will be a kind of structural degradation. “The latest metabolic rift between AI and humans,” Crawford writes, “threatens multiple forms of cascading failure: moral collapse, financial collapse, ecological collapse, and, depending on who you believe, cognitive collapse.”[41]

The rifts in the human relation to nature in modern society are manifestations of the alienated and destructive logic of capital accumulation and crisis. Meta, Amazon, Microsoft, Alphabet (Google), and Tesla together spent $561 billion on AI capital investments in 2023–2025, while generating income, not profits, from these investments of $35 billion. The AI bubble is sustained by debt and by the relentless bidding up of the assets of these firms, as investors seek to be part of this modern gold rush—though recently the market value of all these firms has been falling. Referring to the debt taken on by the hyperscalers in their rush to build data centers, Bloomberg says this is taking the form of “blue-chip bonds, junk debt, private credit and complex asset-backed pools of loans,” amounting to as much as $200 billion or more. The AI acceleration is integral to monopoly-finance capital itself, which expects, if a crash should occur, to be bailed out by Washington on a scale that would dwarf all previous bailouts. To solve the problem of the lack of a sufficient market for AI, computational capital intends to force the adoption of generative AI by implanting it in innumerable apps. This is an accumulation model fraught with risk.[42]

The rise of the neofascist movement associated with Trump’s “Make America Great Again” (MAGA) politics has been heavily financed by Silicon Valley high-tech billionaires such as Musk, Thiel, and Ellison, posing threats to the entire body politic. The announcement of the Trump administration’s Stargate initiative on his first full day in office in his second term, which aims to pour $500 billion into data centers, was designed to boost Oracle and OpenAI (the developer of ChatGPT), headed by Ellison and Sam Altman, respectively, both heavy contributors to Trump’s MAGA political interests. Some commentators have seen these developments as pointing to an emerging state-leveraged cartel, stretching from media to AI to “cloud” technology, dominating both communications and the economy—while at the same time promoting a dictatorial political regime.[43]

Marx’s “General Intellect” and Socialism

If AI is more than a mere epoch-making technology, but rather is to be understood, as Crawford says, as a “registry of power,” then the only viable response is to exert genuine social power over its development, rooted in substantive democracy. The potential ramifications of AI point to what István Mészáros called “the necessity of social control,” a social control that needs to be exerted if a tendency toward ecological, military, and social exterminism is to be avoided. Here not only the forces of production are to be questioned, but even more the social relations of production.[44]

In his “Fragment on Machines” in the Grundrisse, Marx commented on how the transfer of human knowledge and activities—that is, the essence of human labor—to machines via automation led to the embodiment in machines of the “general intellect” of society, that properly belonged to and stood for the “social individual” and, as he explained in Capital, to “the collective worker.”[45] The monopolistic appropriation of this general intellect as the property of the capitalist, meant that it would be utilized for one end and one end only: the accumulation of capital, benefiting very few. The incorporation of the general intellect within capital was, for Marx, a deadly contradiction for capital itself. Any attempt on the part of capitalists to wield the general intellect on behalf of their own narrow, accumulative ends, would create crisis upon crisis. Quoting from the scene titled “Auerbach’s Cellar” in Johann Wolfgang von Goethe’s Faust (Part 1, Scene 5), Marx subtly alluded to a grisly, ribald song about poison fed to a cellar rat, causing it to act “as if its body were by love possessed,” ending in its death—signifying living labor transformed into dead labor: a mere “animated body,” incapable of directly creating labor value. This could be seen as standing in our time for AI capital’s absorption of all the knowledge generated by creative labor and the entire digitalized world within itself, producing a robotic body, leading to AI autophagy and model collapse.[46]

The very potential for the expansion of disposable labor time (leisure) due to automation, Marx explained in his day, contradicts capital’s incessant need to expand surplus labor time. The system therefore seeks to promote, by way of automation—based on the leverage provided by an expanding industrial reserve army—the increased degradation and material dependence of labor, forcing “labour to work longer than the savage does, or than he himself did with the simplest, crudest tools,” now as a mere “appendage of a machine.”[47]

Yet, the reality of the general intellect embodied in automation at the same time makes possible the rise of “the collective worker as the dominant subject” of production and the decisive movement to a society of associated producers.[48] The necessity of social control and planning means putting general social relations in charge while ending the reign of monopoly-finance capital.

Some signs of what is possible are prefigured in China today. China rivals the United States in AI development. China’s open-source DeepSeek AI model is more energy efficient and cost effective than U.S. chatbots. While the Great Houses of AI in the United States are in a race for some godlike “superintelligence” through large language models, Beijing’s “socialism with Chinese characteristics” has focused its machine-learning technology—not without its own contradictions—more directly on manufacturing, logistics, energy, public finance, and public services. Automakers utilize robots with minimal human intervention. AI tools are used heavily in hospitals, where a “simpler, ‘narrow AI’” is employed that is designed for specific tasks. AI in China is integrated primarily into a manufacturing rather than with a developed service economy as in the United States today. Naturally, the very heavy use of robots in Chinese manufacturing leads to displacement of labor. Data banks in China, as in the United States and elsewhere, use vast resources, and are dependent on mining of lithium, cobalt, and rare earth metals. Like the United States, China’s contemporary military modernization is based in AI. Nevertheless, the regulatory controls on AI under “socialism with Chinese characteristics” hold out hope for a more rational social approach to the entire phenomenon.

Indeed, where China differs most from the United States and the West with respect to AI is in its leadership in AI governance, which emphasizes that machine learning must be subordinated to a “people-centered” path of development and to the welfare of the population. Beijing has introduced targeted rules for deep synthesis technologies (known as deepfakes) and for generative AI. All deepfakes require conspicuous labeling or watermarking to ensure transparency, accuracy, and reliability. Any company wanting to offer generative AI has to register its algorithms with the Cyberspace Administration of China, the main regulatory body. Every major body of data that developers wish to include in their AI model must be randomly sampled for discriminatory, antisocial content. Regulations are expressly designed to protect individuals who have definite “rights of portrait, reputation, honor, privacy and personal information.” Most regulations apply to large language models offered to the public, while regulations are less strict for machine learning within industry to support innovation. Nevertheless, the social character of China’s approach, while clearly not sufficient and itself raising difficult questions, contrasts favorably with the more privatized and predatory development of the technology in the United States, where meaningful federal regulations are notoriously absent.[49]

Not surprisingly, China is also the leader in promoting global AI governance, with its Global AI Governance Initiative, introduced in October 2023, and its Shanghai Declaration on Global AI Governance at the World AI Conference in 2024. In these global initiatives, Beijing insists on a “people-centered approach” as a “common task” with respect to the regulation of AI to cope with the “unpredictable risks and complicated challenges” of these technologies, which are frequently being used “for purposes of manipulating public opinion, spreading disinformation, intervening in other countries’ internal affairs, social systems and social order, as well as jeopardizing the sovereignty of other states.” Among the specified dangers are “technological monopolies and unilateral coercive measures”; biases related to discrimination of “ethnicities, beliefs, nationalities, genders, etc.”; the acceleration of environmental harm; and blocking the spread of machine-learning technology throughout the Global South, thus inhibiting global sustainable development. China insists that the goal should be human development and the use of these technologies in such fields as “healthcare, education, transportation, agriculture, industry, culture, and ecology.” The negative effects of AI on employment have to be carefully watched and “mitigated.” All countries are invited to join, in accordance with their own national needs, in establishing “a testing and assessment system based on AI risk levels and a sci-tech ethical review system.” In Xi Jinping’s words, it is necessary “to make sure that AI serves the common good and benefits all, and [that] it is not a plaything of the rich countries and the wealthy.”[50]

Various struggles with respect to AI are arising around the world. A notable demand being promoted is to “pause” AI development until the dangers associated with its further advance can be ascertained, so that rational regulation can play a role in its development.[51] However, the U.S. federal government under the Trump administration is not only attempting not to regulate AI, but it is also actively fighting those individual states and localities all over the country that are attempting to introduce AI regulations.[52] The AI cartel, which can now be seen as encompassing the hyperscalers within high tech, backed by monopoly finance and the energy sector as well as the state, is currently completely in charge. Attempts to socially control AI within monopoly capitalism therefore necessarily point to the need for a more revolutionary movement away from capitalism and toward socialism.

The Great Houses of AI are divided against themselves and cannot stand. They rely for their very existence on an increasingly centralized, coercive, and corrupt class-based capitalist state (and cultural) apparatus, constituting an overall logic that—if allowed to continue—will be nothing less than catastrophic. If humanity is to flourish, the forces and relations of production must be revolutionized together, along with the development of human capacities, creating a world of sustainable human development. This requires the formation under socialism of a true “whole-process democracy” informed by the general intellect, in which “the associated producers govern the human metabolism with nature in a rational way…accomplishing it with the least expenditure of energy and in conditions most appropriate for their human nature.”[53]

Notes

1. Paul Gruenwald and Satyam Panday, “How Data Centres and AI Are Becoming a New Engine of Growth,” World Economic Forum, December 17, 2025. See also Nick Licthenberg, “Without Data Centers GDP Growth was 0.1% in the First Half of 2025, Harvard Economist Says,” Fortune, October 7, 2025.
2. Kate Crawford, Mapping Empires: Power, Politics, and the Planetary Costs of Artificial Intelligence (New Haven: Yale University Press, 2021), 20.
3. Matt Day and Annie Bang, “Big Tech to Spend $650 Billion This Year as AI Race Intensifies,” Bloomberg, February 5, 2026; Marty Hart-Landsberg, “AI and the Economy: A Losing Bet for Working People,” Reports from the Economic Front, February 16, 2026.
4. Paul A. Baran and Paul M. Sweezy, Monopoly Capital (New York: Monthly Review Press, 1966), 220–21.
5. John Maynard Keynes, The General Theory of Employment Interest and Money (London: Macmillan, 1936), 161–62; Karl Marx, Capital, vol. 1 (London: Penguin, 1976), 780.
6. Paul Krugman, “The Economics of Technological Change,” Substack, March 1, 2026, paulkrugman.substack.com; Forbes World Billionaire’s List, 2026, forbes.com/billionaires.
7. Matt Seybold, “The Ellisons are Beta-Testing Big Brother,” American Vandal, October 10, 2025.
8. Bernie Sanders, The Big Tech Oligarchs War Against Worker: AI and Automation Could Destroy Nearly 100 Million U.S. Jobs in a Decade, Ranking Member Minority Staff Report, Health, Education, Labor and Pensions Committee, October 6, 2025.
9. Melissa Heikkilä, “Mustafa Syleyman Plots AI ‘Self-Sufficiency’ as Microsoft Loosens OpenAI Ties,” Financial Times, February 12, 2026.
10. For a critical analysis of technological determinism, see Merritt Roe Smith and Leo Marx, eds., Does Technology Drive History?: The Dilemma of Technological Materialism (Cambridge, Massachusetts: MIT Press, 1994).
11. Karl Marx, Grundrisse (London: Penguin, 1983), 706; Marx, Capital, vol. 1, 279–80; John Bellamy Foster, “Braverman, Monopoly Capital, and AI: The Collective Worker and the Reunification of Labor,” Monthly Review 76, no. 7 (December 2024): 1–13; Matteo Pasquinelli, The Eye of the Master: A Social History of Artificial Intelligence (London: Verso, 2023).
12. John Bellamy Foster, Breaking the Bonds of Fate: Epicurus and Marx (New York: Monthly Review Press, 2025), 17. Although “up to us,” social struggle, though requiring agency, cannot be presented in voluntaristic terms. Rather, it has to be conceived in terms of what Roy Bhaskar called “the transformative model of social activity,” which encapsulated the essence of Marx’s concept of historical change. Roy Bhaskar, Reclaiming Reality (London: Routledge, 2011), 74–81; Karl Marx, The Eighteenth Brumaire of Louis Bonaparte (New York: International Publishers, 1963), 15.
13. Kate Crawford, “Eating the Future: The Metabolic Logic of AI Slop,” e-flux Architecture, September 2025, e-flux.com.
14. Kate Crawford, Atlas of AI: Power, Politics, and the Planetary Costs of Artificial Intelligence (New Haven: Yale University Press, 2021); Kate Crawford and Vladen Joler, “Anatomy of an AI System,” 2018, anatomyof.ai; Kate Crawford, “Calculating Empires,” Knowing Machines, November 23, 2023, knowingmachines.org; Kate Crawford, “Long Now Talks: Mapping Empires,” recorded November 12, 2025; Crawford, “Eating the Future.”
15. “Fetishism” is used here in the sense of Marx’s theory of commodity fetishism. See Marx, Capital, vol. 1, 163–77. On enchanted determinism, see Crawford, Atlas of AI, 213–15.
16. Crawford, Atlas of AI, 8.
17. Tung-Hui Hu, A Prehistory of the Cloud (Cambridge, Massachusetts: MIT Press, 2015), 147.
18. Crawford, Atlas of AI, 18–19. On the metabolic rift, see John Bellamy Foster, Marx’s Ecology (New York: Monthly Review Press, 2000), 141–77; John Bellamy Foster and Brett Clark, The Robbery of Nature (New York: Monthly Review Press, 2020), 12–34.
19. Crawford, Atlas of AI, 213–15; Alexander Campolo and Kate Crawford, “Enchanted Determinism: Power with Responsibility in Artificial Intelligence,” Engaging Science, Technology, and Society 6 (2020): 2.
20. Crawford, “Long Now Talks: Mapping Empires,” 27:07.
21. “World Water Day: The Water Impacts of Lithium Extraction,” Wetlands International Europe, March 22, 2023, europe.wetlands.org; Terry Gross, “How ‘Modern-Day Slavery’ in the Congo Powers the Rechargeable Battery Economy,” NPR, February 1, 2023.
22. Crawford and Joler, “Anatomy of an AI System,” Section XI.
23. Billy Perrigo, “Exclusive: OpenAI Used Kenyan Workers on Less than $2 an Hour to Make ChatGPT Less Toxic,” Time, January 18, 2023; Chinmayi Arun, “Transnational AI and Corporate Imperialism,” Carnegie Endowment of International Peace, October 8, 2024.
24. Crawford, Atlas of AI, 64–68; Lawrence F. Katz and Alan B. Krueger, “The Rise and Nature of Alternative Work Arrangements in the United States 1995–2015,” NBER Working Paper Series, National Bureau of Economic Research, Washington DC, September 2016: 7; Hu, Prehistory of the Cloud, 89; Martin Gonzalez-Cabello, Auyon Siddiq, Charles J. Corbett, and Catherine Hu, “Fairness in Crowdwork: Making the Human Supply-Chain More Humane,” Business Horizons 68, no. 5 (September–October 2025): 645–57.
25. “Since Michel Foucault’s Discipline and Punish, it has become commonplace to consider the prison as the origin point of today’s surveillance society, with the elder Bentham as it its progenitor. In fact, the panoptic prison owes its origins to the work of the younger Bentham in the context of the early manufacturing facility. The panopticon began as a workplace mechanism well before it was conceptualized for prisons” (Crawford, Atlas of AI, 61).
26. Crawford, Atlas of AI, 59–62, 72; E. P. Thompson, “Time, Work-Discipline, and Industrial Capitalism,” Past and Present, no. 38 (December 1967): 56–97; Harry Braverman, Labor and Monopoly Capital (New York: Monthly Review Press, 1998).
27. Crawford, Atlas of AI, 74; Karl Marx and Frederick Engels, Collected Works (New York: International Publishers, 1975), vol. 6, 127; Marx, Capital, vol. 1, 1034–38; István Mészáros, The Challenge and Burden of Historical Time (New York: Monthly Review Press. 2008), 43–49; Ian Angus, Facing the Anthropocene (New York: Monthly Review Press, 2016), 111–25.
28. Crawford, “Eating the Future”; Crawford, Atlas of AI, 95.
29. Crawford, Atlas of AI, 119.
30. On crowdworkers, see Crawford, Atlas of AI, 63–64.
31. Crawford, Atlas of AI, 123–36, 145–46.
32. Crawford, Atlas of AI, 193–99; Peter Waldman, Lizette Chapman, and Jordan Robertson, “Palantir Knows Everything About You,” Bloomberg, April 19, 2018.
33. Hu, Prehistory of the Cloud, 115; Crawford, Atlas of AI, 202.
34. Crawford, Atlas of AI, 189–92; Ed Pilkington, “US Military Reportedly Used Claude in Iran Strikes Despite Trump’s Ban,” Guardian, March 1, 2026; Gary Wilson, “Anthropic Is Already at War,” Struggle La Lucha, March 5, 2026, struggle-la-lucha.org.
35. Crawford, “Long Now Talks: Mapping Empire,” 5:40–18:17; Crawford “Eating the Future”; Clive Hamilton and Jacques Grinevald, “Was the Anthropocene Anticipated?,” Anthropocene Review 2, no. 1 (2015): 67.
36. Crawford, “Long Now Talks: Mapping Empire,” 13:18, Crawford, “Eating the Future.”
37. Crawford, “Long Now Talks: Mapping Empires, 37:40–39:46; Crawford, “Eating the Future.”
38. Crawford, “Long Now Talks: Mapping Empires,” 28:08–29:20; Peter Landers, “Artificial Intelligence’s ‘Insatiable’ Energy Needs Not Sustainable, Arm CEO Says,” Wall Street Journal, April 9, 2024.
39. Crawford, “Long Now Talks: Mapping Empires,” 31:16–32:37; John Bellamy Foster, Brett Clark, and Richard York, The Ecological Rift (New York: Monthly Review Press, 2010), 169–82; William Stanley Jevons, The Coal Question (London: Macmillan, 1865), 102–16.
40. Ovid, Metamorphoses, trans. Charles Martin (New York: Norton, 2004), 298; Richard Seaford, Ancient Greece and Global Warming, Classical Association Presidential Address (London: Classical Association, 2009), 6; John Bellamy Foster, Foreword in Fred Magdoff and Chris Williams, Creating an Ecological Society (New York: Monthly Review Press, 2017), 7–9.
41. Crawford, “Long Now Talks: Mapping Empires,” 40:11.
42. Matt Day and Amy Bang, “Big Tech to Spend $650 Billion This Year as AI Race Intensifies,” Bloomberg, February 6, 2026; Hart-Landsberg, “AI and the Economy.”
43. Frank Vogl, “Trump’s Return to the Robber Baron Age,” Globalist, October 13, 2025.
44. István Mészáros, The Necessity of Social Control (New York: Monthly Review Press, 2015), 23–51.
45. Marx, Grundrisse, 706; Marx, Capital, vol. 1, 464–69, 544–45; Michael Heinrich, “The ‘Fragment on Machines’: A Marxian Misconception in the Grundrisse and Its Overcoming in Capital,” in Marx’s Laboratory: Critical Interpretations of the ‘Grundrisse,’ Riccardo Bellofiore, Guido Starosta, and Peter D. Thomas, eds. (Chicago: Haymarket, 2013), 197–212; John Bellamy Foster, “Braverman, Monopoly Capital, and AI,” Monthly Review 76, no. 7 (December 2024): 1–13. See also Te Li, “From Classic Labor to the Labor of the ‘General Intellect’: The Impact of the Digital Intelligence Era on Socialist Labor Theory,” Monthly Review 77, no. 11 (April 2026): 46–62.
46. Marx, Grundrisse, 704; Marx, Capital, vol. 1, 302; Johann Wolfgang von Goethe, Collected Works, vol. 2, Faust, Parts I and II, ed. and trans. Stuart Atkins (Princeton: Princeton University Press), 54; Sami Khatib, “The Drive of Capital: Of Monsters, Vampires, and Zombies,” Coils of the Serpent 8 (2021): 101–13.
47. Marx, Grundrisse, 708–9; Marx, Capital, vol. 1, 799.
48. Marx, Capital, vol. 1, 544–45.
49. Vanessa Bates Ramirez, “The U.S. and China Are Pursuing Different AI Futures,” IEEE Spectrum, February 19, 2026; “AI Watch: Global Regulatory Tracker—China,” White and Case, September 22, 2025.
50. Ministry of Foreign Affairs, People’s Republic of China, “Global AI Governance Initiative,” October 20, 2023; Ministry of Foreign Affairs, People’s Republic of China, “Full Text: Shanghai Declaration on Global AI Governance,” July 4, 2024; Xi Jinping, The Governance of China, vol. 5 (Beijing: Foreign Languages Press, 2025), 553.
51. Darko Suvin, “I Am Afraid of AI: A Politico-Epistemological Exasperation,” Historical Materialism (blog), 2026, historicalmaterialism.org; Anna Gordon, “Why Protestors Around the World Are Demanding a Pause on AI Development,” Time, May 13, 2024; Anthony Elmo, “Data Center Moratorium Bills Are Spreading in 2026,” Good Jobs First, February 19, 2026.
52. The White House, “Ensuring a National Policy Framework for Artificial Intelligence,” Executive Order, December 11, 2025.
53. Karl Marx, Capital, vol. 3 (London: Penguin, 1981), 959.

[John Bellamy Foster is the editor of Monthly Review and a professor of sociology at the University of Oregon. He is the author of several books, which have been published in at least twenty-five languages. Courtesy: The Monthly Review, an independent socialist magazine published monthly from New York City since 1949, whose present editor is John Bellamy Foster.]