Automation, surplus value and class struggle when algorithms manage the means of production.

Introduction

When Karl Marx published Das Kapital in 1867, the cutting‑edge technology was the steam engine. Today’s equivalent is the cloud‑based GPU cluster powering large language models. Yet the questions Marx posed—who owns the machines, who benefits from their output, and how does technology reshape class relations—are more relevant than ever. AI firms claim that algorithms can replace knowledge workers, but Marx reminds us that value is ultimately rooted in human labour, even if that labour is hidden in training‑data sweatshops or in the mining of lithium for server‑farm batteries.

From Steam Engines to GPUs: Constant Capital Upgraded

In Capital Volume I Marx defined machinery as constant capital—an advance that transfers its existing value to commodities while allowing capitalists to squeeze more surplus from fewer workers. GPUs and transformer models fit the definition perfectly: they embody huge sunk costs, depreciate rapidly and are deployed to cut payroll or extract rents from software subscriptions. Nvidia’s meteoric profits echo Marx’s observation that the firm selling the machine can reap abnormal gains before competition equalises returns.

Surplus Value without Workers?

AI evangelists often boast of ‘labour‑free value creation’. Marx would counter that the training of models on billions of human‑generated tokens already embodies labour; inference then draws down that pre‑stored pool of value. Moreover, the Beta testers, prompt engineers and Kenyan annotation crews who label toxic content remain part of the global working class, receiving a fraction of the surplus they help unlock.

The Profit Squeeze: Rising Organic Composition of Capital

Marx predicted that as capital deepens its reliance on machinery (today, AI), the organic composition of capital rises, living labour falls, and the rate of profit tends to decline. Federal Reserve Governor Michael Barr recently echoed this concern, warning that AI’s high capex and energy costs could outpace its productivity pay‑offs.

A Reserve Army Reprogrammed

Generative AI doesn’t abolish the reserve army of labour—it globalises it. Language models draw on an oversupply of precarious click‑workers who micro‑task on platforms like Upwork and Amazon Mechanical Turk, keeping wages low and labour abundant. The International Labour Organization notes that clerical roles, dominated by women, face the highest automation risk, deepening gender inequalities. citeturn0news19

General Intellect vs. Private Code

In the Grundrisse Marx foresaw ‘general social knowledge’ becoming a direct force of production. AI operationalises that prophecy, but in proprietary form—gigantic models guarded by non‑disclosure agreements. The contradiction between socially produced data and privately appropriated AI rent mirrors the enclosure of commons in the 18th century.

Contradictions and Counter‑Moves

Policy ideas now range from a data‑dividend tax on Big AI to universal basic income funded by automation rents. Tech unions push for collective bargaining over algorithmic management, while EU lawmakers debate copyright levies on model training. Whether these reforms succeed depends on class power, not on the silicon itself.

Conclusion

AI magnifies capitalism’s core dynamics rather than transcending them. Far from rendering Marx obsolete, the age of algorithms vindicates his insight that technology, in the hands of capital, is both a productivity miracle and a weapon of domination. The struggle over who controls the code will decide whether AI deepens exploitation or lays groundwork for a post‑capitalist horizon.