In contemporary digital capitalism, Artificial Intelligence, a broad and underdefined umbrella term in computer science referring to machines or programs designed to perform human-like cognitive tasks, attracts unprecedented volumes of investment while being discursively framed as the horizon of full automation. Yet rather than abolishing labor, the expansion of AI reorganizes and multiplies specific forms of human work, because, in effect, AI binds collective knowledge labor in digital infrastructures and model architectures. According to this perspective, the current prominence of AI is historically situated in the neoliberal restructuring of capitalism since the 1980s to account for pervasive over-accumulation and continuously sinking productivity levels. Crucially, this machinery does not operate independently. Irrespective of its different forms, its functionality depends on continuous inputs of labor that render data usable as a raw material. While platform users generate abundant data traces, this activity does not constitute productive labor in the Marxist sense, as it does not occur within a wage relation that directly produces surplus value but rather forms part of a service relation whose profits derive from the redistribution of already produced value . Data only becomes valorizable through data work. This labor is indispensable for all domains of AI development, meaning training, maintenance, and supervision. The contradictory character of AI development, namely its promise of automation coupled with its structural dependence on human labor, manifests in the specific conditions of data work. Because AI is presented as self-operating, the labor that sustains it is rendered invisible, externalized, and persistently devalued. Investment flows disproportionally into computational infrastructure and model scaling, while the workers who transform heterogeneous data points into trainable inputs remain precarized, rendered interchangeable, and difficult to statistically capture.

The current turn towards generative AI, understood as technologies that aim at producing new, statistically probable data points that structurally resemble their training data, unfolds under conditions of stagnating growth and surplus capital, appearing less as the expression of new productive capacities than as an attempt to absorb overaccumulated capital in search of profitable outlets. Hundreds of billions invested in data centers, GPUs, and model scaling contrast sharply with the modest and uneven productivity gains empirically observable across sectors. The bulk of realized profits accrues to hardware producers and infrastructure providers, while many AI applications circulate capital within a tight oligopoly through mutual investments and long-term cloud contracts. Especially important here are the so-called ‘Magnificent 7’, who monopolize hardware production, platform infrastructure, and hosting space. Even where generative AI is widely implemented, measurable returns remain limited, often confined to narrowly formalized service tasks. Rather than inaugurating a new wave of broad-based value creation, generative AI thus appears predominantly as a technology for redistributing existing value and stabilizing expectations under conditions of structural stagnation. In this sense, the prominent ‘productivity paradox’ of contemporary capitalism extends into AI development. As investment in fixed capital expands, the relative weight of living labor within total capital declines, accelerating crisis-prone patterns of accumulation associated with a rising organic composition of capital. Historically, such tendencies are counteracted by capitalist strategies that raise the rate of surplus value or reduce production costs. These counter-tendencies include wage suppression, the global expansion of a reserve army of labor, intensified exploitation, and the cheapening of elements of constant capital. Under conditions of AI-driven restructuring, precarization functions precisely as such a counter tendency, as a structurally enforced response to mounting pressures on profitability. The fragmentation of tasks, global outsourcing of data work, algorithmic control, and the constant expansion of a digitally mediated labor pool, i.e., the creation of a global reserve army, depress wages and increase competition among workers. In this way, generative AI simultaneously expresses and exacerbates tensions within knowledge-based service sectors, which originally emerged from the 1970s onwards as a means to absorb and valorize excess capital. Generative models promise to automate linguistic and symbolic production, yet in practice, they primarily substitute already highly formalized tasks such as writing e-mails or summarizing documents, while expanding monitoring, correction, and coordination labor, as stated by consulting Firm McKinsey & Company. Knowledge labor is thus not abolished but objectified, meaning that workers’ linguistic, cultural, and organizational competencies are extracted, formalized, and re-embedded in model architectures as scalable outputs. In this sense, the rise of generative AI marks less a technological breakthrough than a symptomatic attempt to manage the structural limits of digital capitalism through enforced labor exploitation.

As the material capacity of AI to generate broad-based productivity gains remains limited, its continued expansion depends increasingly on speculative expectations and discursive constructions of technological inevitability. To legitimize massive investments despite stagnating profitability, the already established AI hype is actively intensified. Whether framed as ‘AI-boosterism’ or as apocalyptic ‘AI-doomerism’ , both variants fabricate a sense of inevitability to which only further technological innovation appears as a response. In this way, AI is conjured as a “technological fix” for capitalist crisis tendencies. A central site of this constructed inevitability is the debate on Artificial General Intelligence (AGI), the fictional prospect of a system with human-like cognitive capacities across domains. Claims of its imminence rest largely on assumptions about exponential scaling of data, compute, and model size. Yet current generative systems operate through stochastic token prediction, reproducing statistical regularities without semantic understanding, intentionality, or a coherent world model. Longstanding philosophical critiques, such as John Searle’s distinction between weak and strong AI, as well as persistent technical limits including bias, opacity and hallucinations further undermine claims of imminent general intelligence. Despite these limitations, industry-affiliated segments of science reinforce narratives of disruption and automation. As an illustrative example, researchers at OpenAI frame large language models as alleged ‘general purpose technologies’ comparable to electricity or steam power, even though empirical evidence fails to demonstrate comparable productivity effects. Scientific authority thus lends credibility to speculative claims, reinforcing a technological determinism that presents AI-driven restructuring as both inevitable and economically transformative. In this configuration, the ‘general intellect’ does not function as a collectively shared knowledge-standard that facilitates the production of societally necessary goods but rather is estranged along two dimensions. Not only to continuously improve the technologies employed to control the labor force, exemplified in the historical lineage from the scientific management proposed by Francis Taylor through to contemporary forms of algorithmic management , but simultaneously to naturalize and universalize the contradictions of capitalist exploitation. Scientific knowledge, increasingly embedded in corporate funding structures and competitive innovation regimes, contributes to the reproduction of a technological determinism that presents AI-driven restructuring as both inevitable and socially beneficial. The AI hype thus appears as a direct expression of the ‘ruling ideas’ of the capitalist class, in that it ideologically conceals the discrepancy between mobilized investment volumes and realized productivity gains, while legitimizing continued capital concentration and labor exploitation.