BadHabitus

← Back to Essays

Essay · AI, value theory, and labor
Part of the series AI and Political Economy

Labor Will Not Become Inconsequential

Why David Harvey’s own theory contradicts his prediction about artificial intelligence

Will Charles

David Harvey recently predicted that artificial intelligence would further diminish the power of labor until labor became “almost inconsequential,” because the work would eventually be automated. The claim is striking not because it is unusually radical, but because it contradicts Harvey’s own account of capitalism. His historical work treats technological change as the result of competition, monopoly, state policy, labor conflict, and class strategy. His theory of value also explains how highly automated firms remain dependent on value produced by workers elsewhere. Taken seriously, Harvey’s framework predicts not the disappearance of labor but an intensified struggle over where labor is performed, who controls technology, and who captures the value labor creates.

Harvey made the prediction at the beginning of a recent conversation with Joshua Citarella for Doomscroll. He argued that technological innovation has historically been used to weaken labor and that artificial intelligence will continue this process until labor becomes nearly irrelevant. The first half of that claim is difficult to dispute. Capitalists have repeatedly used machinery, automation, surveillance, and new forms of workplace organization to reduce workers’ bargaining power. But the second half does not follow from the first. The fact that capital uses technology against labor does not mean that technology can eliminate capital’s dependence on labor.

David Harvey in conversation with Joshua Citarella, Doomscroll, recorded as part of the Whitney Museum program. The claim about labor becoming inconsequential opens the episode.

Harvey’s own historical account makes that distinction clear. In the same conversation, he describes the United States of the 1960s as technologically sluggish because monopoly had reduced competitive pressure. A few major automobile companies could coordinate prices and enjoy stable profits without continually transforming production. Neoliberalism later restored more aggressive competition, weakened organized labor, and forced firms to pursue cost-cutting and technological change more intensely. On this account, technology is not an autonomous force that arrives from outside society. Its development and deployment depend on the organization of markets, the power of workers, the policies of states, and the strategies of firms.

The history of precarity works the same way. Secure union jobs did not disappear because machines spontaneously demanded a more flexible workforce. Employers outsourced production, converted employees into contractors, relocated factories, attacked unions, and changed labor law. Governments tolerated or encouraged these changes. Technological development assisted that project, but it did not independently determine it. Harvey cannot treat technology as the product of political-economic institutions when explaining the past and then treat it as an autonomous historical force when predicting the future.

His own value theory poses an even more fundamental problem for the claim that labor will become inconsequential. During the interview, Harvey is asked how a fully automated company could earn profits if labor is the source of value. He answers by distinguishing value from profit. A firm does not necessarily retain only the value produced by its own employees. Through market prices and the equalization of profit rates, value can be redistributed from labor-intensive sectors toward more capital-intensive ones. A company employing relatively few workers may therefore receive a large share of the total profits generated across the economy.

This point can sound abstract, but its practical meaning is straightforward. A highly automated software company may directly employ fewer people than a factory, hospital, warehouse, or logistics network. It can nevertheless capture enormous returns because it operates within an economy sustained by labor elsewhere. Workers manufacture its chips, build its data centers, generate its electricity, maintain its networks, construct its buildings, produce its training data, and purchase the goods and services through which its revenues are realized. The company may reduce the amount of labor appearing on its own payroll, but it has not become independent of labor. It is living on labor performed somewhere else.

What is possible for one company also cannot automatically be generalized to the entire economy. A single firm can automate much of its production, dismiss workers, and continue earning the average rate of profit. Every firm cannot eliminate labor simultaneously, because the pool of value being redistributed would disappear. Harvey himself describes the broader contradiction: individual capitalists increase productivity for their own advantage even when their collective actions undermine the conditions of accumulation. Each firm may rationally seek to reduce its labor costs, while the system as a whole still depends on labor to produce value and on workers’ incomes to sustain demand.

That is a theory of crisis, not a theory of labor becoming inconsequential. Capitalism may repeatedly weaken the workers on whom it depends. It may generate unemployment, suppress wages, and produce severe contradictions between productive capacity and effective demand. Those contradictions may require political intervention, as Harvey argues in his discussions of Roosevelt and Keynes saving capitalism from the destructive actions of individual capitalists. But none of this suggests that labor quietly fades from economic importance. It suggests that capitalism continually destabilizes the conditions of its own reproduction.

The claim about artificial intelligence also rests on an assumption that deserves more scrutiny: that AI is inherently capital intensive and will therefore remain controlled by a small number of enormously powerful firms. That assumption increasingly looks less like a technical fact than a description of the way the American AI industry has chosen to organize itself. The dominant companies have built proprietary models behind closed interfaces, surrounded them with vast data centers, and presented ever-growing computational requirements as an unavoidable feature of progress.

Yet the cost of obtaining a fixed level of model performance has fallen rapidly. Smaller models, quantization, expert routing, improved hardware, and competition from open-weight systems have all reduced the amount of computation required for many tasks (Cottier et al. 2025; Stanford HAI 2025). Not every organization needs the largest and most expensive model. For many applications, a cheaper system that can be downloaded, customized, and operated locally may be more useful than a frontier model rented through a proprietary interface.

AI may remain capital intensive, but if it does, that outcome will have to be produced and defended. Scarcity can be maintained through intellectual-property rules, export controls, government procurement, security requirements, liability standards that only the largest firms can satisfy, and restrictions on open models. Each measure may have a narrow justification, but together they can construct a protected market in which a handful of companies retain control over the technology (Acemoglu and Johnson 2023). The monopoly is not a natural consequence of machine intelligence. It is a political and institutional accomplishment.

This is where Harvey’s prediction begins to resemble the claims made by Silicon Valley itself. The technology companies insist that their systems are uniquely powerful, extraordinarily expensive, and destined to replace enormous amounts of human labor. Harvey accepts the same premises and changes only the moral tone. The companies describe the future as liberation or abundance; Harvey describes it as the destruction of labor power. But both accounts grant the firms precisely what they want most: the belief that their technology is historically irresistible.

Harvey is right that capitalists will attempt to use artificial intelligence against workers. They will use it to intensify labor, monitor performance, cut staffing, weaken occupational boundaries, and threaten employees with replacement. Some jobs will disappear, while others will be reorganized or degraded. But “capital will use AI to weaken labor” is not the same proposition as “AI will make labor unnecessary.” The first is a theory of class power. The second is technological determinism.

Harvey’s own history supports the first proposition and rejects the second. His value theory shows that apparently automated firms remain dependent on labor elsewhere. His account of capitalist competition shows that technology is developed and deployed through institutional struggle. His crisis theory shows that efforts to eliminate labor costs can undermine the conditions of profit itself. None of these arguments lead to a world in which labor becomes inconsequential.

Artificial intelligence will change the organization of work, but its effects will be determined by ownership, law, labor power, public investment, and the institutions through which it is deployed. The political question is not whether a machine can think. It is who controls the machine, whose work it reorganizes, and who captures the benefits.

Sources

Views expressed in these essays are my own.