Knowledge Commons
HomeTopicsSeriesGuidesPopularAboutContact

Why Does Labour's Share Shrink as Artificial Intelligence Grows?

The Fed's Warning, the Productivity Paradox and the Class Anatomy of the Distribution of Income

37 min readAuthor: Oğuz Demirkapı
Why Does Labour's Share Shrink as Artificial Intelligence Grows?

"The machine is new, the chain is old."


First, Briefly: Ten Theses

This will be a long essay. For anyone in a hurry, the summary first, then the detail.

  1. Who made the warning? On 12 November 2026, Michael Barr, a member of the Board of Governors of the US Federal Reserve, said in a speech at Yale University that artificial intelligence could create "serious short-term disruptions" in the labour market, that the gain could flow mainly to capital, and that as a result "labour's share of income would fall, market concentration would rise, and inequality would grow."
  2. What do the figures say? In the US nonfarm business sector, labour's share fell to 52.8% in the second quarter of 2026; after the revision of 5 November, the third quarter, at 53.0%, is the lowest level measured since 1947. In the same period, corporate profit margins broke a record at 14.9% of national income.
  3. A paradox? Productivity is rising, labour's share is falling. For bourgeois economics this is a paradox; for Marx it is the ordinary working of the accumulation of capital. A rise in productivity is not, of itself, a rise in wages. It is the instrument of the production of relative surplus value.
  4. It is not new. What we are living today is the continuation of a break that has gone on since 1979: in the US, between 1979 and 2025, net productivity rose 90.2%, and the typical worker's wage rose 33.0%. Artificial intelligence did not create this tendency. It arrives as a new weight on the pan of a scale that was already tipped.
  5. Technology does not determine distribution. The relation of power does. If labour's share stayed high between 1945 and 1975, the reason was not the machines of that period. It was the organised working class, the existence of the socialist bloc, and the fear that forced capital to make concessions.
  6. What does "capital-biased innovation" mean? This concept, which Barr uses, is, in Marxist language, the rise of the organic composition of capital, the replacement of living labour by dead labour (the machine, the model, the data centre). Artificial intelligence is doing this for the first time at a mass scale in the field of mental labour.
  7. The monopoly problem is at the centre. According to Barr, the advantage of the leading artificial-intelligence companies may be "unprecedented". This is an official admission that we are speaking not of competitive capitalism, but of monopoly capitalism.
  8. The bottom rung of the ladder is being pulled away. The first signs of replacement are seen among young and inexperienced workers. The reserve army of industry is now fed by university graduates too.
  9. Turkey is living a harsher form of the same tendency. According to TÜİK figures, the share of labour payments in value added fell to 38.1% in the second quarter of 2026. In 1974 the minimum wage was 80.6% of national income per person; in 2025 it was 43.6%.
  10. The solution is not "education" and "competition". It is property and organisation. Artificial intelligence is a means of production distilled from humanity's common knowledge, from what Marx called the "general intellect". The rise in productivity born of it is social too. The growth of the share depends on the organising of the power that will take the share.

Now let us go into the detail.


Why Does a Central Banker Speak of Labour's Share?

Let us note this first: this warning was not made by a union confederation, a left economics institute, or a workers' party. The one who made the warning is a member of the board of the most powerful monetary institution of world capitalism.

While speaking at Yale on 12 November, Michael Barr called on society to prepare for the "serious short-term disruptions" that artificial intelligence could create in the labour market, and said that rapid change could make it harder for workers to adapt (TradingView/Stocktwits). The core of the speech is in these sentences: artificial intelligence may be of the kind some researchers call "capital-biased innovation", that is, its return goes essentially not to labour but to capital. In Barr's words, in that case "labour's share of income falls, market concentration rises, and inequality increases."

Barr's real worry is concentration. According to him, the scale of the advantage held by the leading artificial-intelligence companies may be "unprecedented"; while large companies meet the cost of artificial-intelligence investment and go on growing, the market share of small firms will recede. If the ownership and the investments of the leading companies stay concentrated, the return will flow, out of proportion, to the owners of the capital that displaces labour (BigGo Finance).

Barr also tries to keep the balance: he says that so far there is "considerable evidence" that artificial intelligence has displaced relatively few workers, and that in return it has raised the productivity of some workers, grown demand in some sectors, and created new occupations (Bhaskar Live/IANS). But he underlines one detail: artificial intelligence may have begun to take the place of labour in some sectors, especially among young and inexperienced workers (Rallies/Reuters summary).

The timing is meaningful too. The speech came immediately after the US midterm elections, in which worries about the rapid expansion of artificial-intelligence infrastructure turned into a reaction, and before the forum in San Francisco of the task force that Fed Chair Kevin Warsh set up to examine the economic effects of artificial intelligence (Tekedia). This is not only Barr's personal worry. Inside the US ruling class there is a fear that the social cost of the wave of artificial-intelligence investment could turn into a political crisis. We discussed the relation of techno-capital to the election in our essay on the US midterms, and to whom this wave of investment is billed in our essays on the artificial-intelligence bubble.

Barr is not new to this subject. In his speech of February 2026 he drew three scenarios for the future of artificial intelligence, and described one of them, in which artificial-intelligence agents displace professional and service jobs en masse and a large part of the population becomes "effectively unemployable", as a "jobless boom" (HR Executive). In July he made this question the title outright: "Will artificial intelligence raise living standards for broad sections, or will it grow income and wealth inequality?" (Federal Reserve).

Why, then, does a central banker say these things? Let us not miss the irony: the Fed itself played the leading role in labour's share turning downward from the 1980s. Paul Volcker's raising of interest rates to historic levels in 1979, and his conscious creation of a wave of unemployment, was one of the basic moves that broke the bargaining power of the American working class. That a member of the same institution today speaks with worry of labour's share shows that it senses the system is approaching the limit of the imbalance it created itself. The worry is not for the workers. It is for insufficient demand, political instability, and a crisis of legitimacy.


The Figures: Productivity Up, the Share Down

The latest picture in the US

The US Bureau of Labor Statistics (BLS) announced the third-quarter figures on 5 November 2026. In the nonfarm business sector, labour productivity rose 3.6% annualised; output grew 3.7%, while hours worked barely moved. In the same report labour's share was given as 53.0%, and this is the lowest level of the series that began in the first quarter of 1947 (BLS).

The detail of the figure is even more instructive. Unit labour cost fell 0.7% in the third quarter; the reason was that hourly compensation rose 2.9% while productivity rose 3.6% (DOL). That is, the worker produced more, and was paid less for each unit produced. Inflation-adjusted hourly compensation rose only 1.0% in the last four quarters (FourWeekMBA). In the second quarter, real hourly compensation had fallen 3.3% (BLS).

Let us set down a note on method: the 52.8% that Barr mentioned in his speech is the figure the BLS published for the second quarter and called "the lowest level recorded" (BLS). On 5 November the BLS revised the 2021–2026 figures upward, and in the new series the lowest point became the 53.0% of the third quarter. The detail differs. The direction is the same.

What is on the other pan? According to the account of Gregory Daco, chief economist at EY-Parthenon, corporate profit margins broke a record at 14.9% of national income while labour's share fell to 52.8% (Fortune). Daco's observation is this: "The rise in productivity is protecting margins, not income." He also adds that he does not see a floor for labour's share: "I don't think there's a floor."

Let us set down the long-term picture: from the Second World War until the beginning of this century, labour's share had run roughly between 60% and 65% (Rallies/Reuters summary). Today we are at 53%. This means that about a tenth of national income has shifted from the worker to capital.

The Economic Policy Institute (EPI) calculates the barest form of this shift: between 1979 and 2025, net productivity rose 90.2% while the typical worker's wage rose 33.0%. If wages had followed productivity, the typical worker would be earning 16.40 dollars more an hour today (EPI). The EPI stresses that this break is not "natural", and that it is the product of conscious policy choices such as raising the minimum wage infrequently, weakening the unions, tolerating unemployment, and cutting taxes for the rich.

Barr's own figure also shows the scale of the inequality: in the US the richest 20% of households take 52% of all income (Crypto Briefing).

The picture in the world

According to the International Labour Organization's (ILO) report Employment and Social Trends 2026, published in January 2026, the global labour share in 2025 was 52.6%, below the 2019 level; this also shows that the rise in real wages stayed behind the rise in productivity (CrackitToday summary). The ILO's 2024 account is more concrete: the global labour share has receded 1.6 points since 2004; this means a gap of 2.4 trillion dollars in 2024 alone, compared with what workers would have earned if the labour share had stayed fixed. The same report had also warned that, if historical patterns continue, the leaps in generative artificial intelligence could create an additional downward pressure on the labour share (Dawn).

The ILO report also puts another fact on record: artificial intelligence and automation can block the way of educated young people looking for their first job in occupations that require high skill, especially in high-income countries (Dawn). This is the confirmation, by an international institution, of the phenomenon we named "the bottom rung of the ladder" in The Poverty of the White Collar.

The picture in Turkey

In Turkey the situation is harsher. According to TÜİK's second-quarter figures for 2026, the economy grew 2.3%, but the share of labour payments in gross value added fell from 38.3% a year earlier to 38.1% (Evrensel). Set beside the 53% in the US, it is plain how low labour's share is in Turkey. (The methods of measurement differ; the US figure covers the nonfarm business sector, the TÜİK figure covers all value added. But the gulf between them is too large to be explained by a difference of method.)

According to the calculations of DİSK-AR, the minimum wage was 80.6% of national income per person in 1974, and this ratio fell to 43.6% in 2025; in 2022 labour's share of national income hit bottom at 26.5% while capital's share peaked at 54.5% (Evrensel). According to DİSK-AR's September 2026 Wage Losses Monitoring Report, the total loss of more than 17 million insured workers in the first eight months of the year, because of inflation and tax deductions, reached 1 trillion 863 billion lira. In the words of Arzu Çerkezoğlu, general president of DİSK: "Today in Turkey the crisis of distribution has got out of control." (Evrensel)

DİSK-AR also sets down one of the basic reasons for the spread of low wages in Turkey: only about 10 percent of workers are covered by a collective agreement (Turkish Minute). In Two Thermometers we discussed how inflation works as an instrument of distribution, and in our latest essay on the industrial crisis we discussed the closing of factories. In Turkey the mechanism that melts labour's share is, in the main, not artificial intelligence. It is inflation, tax, and the lack of organisation. But when artificial intelligence is added to this picture, a bargaining power that is already weak will be worn down further.


The Historical Context: For Whom Did the Machine Work, Each Time?

If we try to understand the fall of labour's share only with today's figures, we see it as a "technology shock". Yet the relation between the machine and distribution has a history of two hundred years, and this history teaches us one clear thing: technology does not determine distribution; whose hands hold the technology does.

Ricardo's admission and the Luddites (1811–1821)

David Ricardo, one of the greatest names of classical economics, added a new chapter titled "On Machinery" to the third edition, dated 1821, of his work On the Principles of Political Economy and Taxation. In this chapter he wrote openly that he had previously thought machines would benefit everyone, but that he was now convinced mechanisation could harm the interests of the working class. This admission by one of the founders of bourgeois economics showed that the reaction of the Luddite movement (1811–1816), which was breaking weaving machines in England in those same years, was not "ignorance" but the expression of a real conflict of interest.

The Luddites were not against the machine. They were against the machine being used against them. The problem was that they chose, historically, the wrong target: breaking the machine does not change the ownership of the machine.

"Engels' pause" (1800–1840)

In a study he published in 2009, the economic historian Robert Allen showed that, in the first decades of the Industrial Revolution, output per worker rose rapidly while real wages barely moved for decades. He named this period "Engels' pause", with reference to the observations in The Condition of the Working Class in England, which Engels wrote in 1845. Productivity was rising, the profit share was swelling, and the workers were living in the misery of the factory towns.

Wages began to rise only after the 1840s. Why? The Chartist movement, the struggle to legalise the unions, the ten-hour working-day law, the entry of the working class onto the stage as a political power. That is, what raised wages was not the machine. It was organisation. The question of whether what we are living today is an "artificial-intelligence Engels pause" is not asked for nothing.

Marx: productivity is the instrument of relative surplus value

In the first volume of Capital (especially in the chapter "Machinery and Modern Industry"), Marx solved this question conceptually. The logic is this:

  • The value the worker produces in a working day divides in two: the necessary labour time in which they meet the value of their own labour power (the wage), and the surplus labour time they leave to capital unpaid.
  • Capital enlarges surplus value in two ways: by lengthening the working day (absolute surplus value), or by shortening necessary labour time (relative surplus value).
  • A rise in productivity shortens necessary labour time to the extent that it cheapens the goods the worker consumes. Even if the working day stays the same, surplus labour time grows.

The conclusion that comes out of this is: in capitalism, the natural tendency of a rise in productivity is not to increase labour's share, but to reduce it. For labour's share to stay fixed, or to rise, the working class has to take back a part of the rise in productivity by struggle.

The statistic we call "labour's share" corresponds, roughly, to the ratio in Marx between variable capital (v) and newly created value (v + s). The fall of labour's share means the rise of the rate of surplus value (s/v), that is, of the rate of exploitation. Where the BLS says "unit labour cost fell", Marx would say "the rate of exploitation rose".

Kaldor's "constant share" and the golden age (1945–1975)

In 1961 the English economist Nicholas Kaldor counted, among the "stylised facts" of capitalist growth, that labour's share stayed constant in the long run. Mainstream economics taught this for decades as if it were a law of nature. A writer at Haver Analytics recalls that, when learning Solow growth accounting, the constancy of the labour share was an accepted "stylised fact" (Haver Analytics).

But this "constancy" was not a law of nature. It was the product of a particular balance of power between classes. The distinguishing features of the postwar period:

  • Strong unions and a collective-bargaining order in the West,
  • The existence of the socialist bloc and the "fear of October" it created (the concessions we told in detail in our essay on the October Revolution),
  • The worker being seen, in Fordist production, as necessary also as a mass consumer,
  • The Bretton Woods order, in which the movements of capital were limited.

The EPI's figures confirm this too: inflation-adjusted wages and productivity grew together from the end of the Second World War until the 1970s; since 1979 productivity has run ahead of the wage (EPI). The EPI's concise sentence: "When unions were at their strongest, the wage rose together with productivity."

The neoliberal counter-revolution (1979–today)

1979 is a breaking point: the Volcker shock, Reagan's firing in 1981 of more than eleven thousand striking air-traffic controllers, Thatcher's war against the miners. In Turkey the name of the same period is the 24 January decisions and 12 September. As DİSK-AR also stresses, the melting of the ratio of the minimum wage to national income began with the 12 September coup and the neoliberal policies that followed it (sol.org.tr).

In this period the mechanisms that lowered labour's share fed one another: the shifting of production to low-wage countries, deunionisation, subcontracting and precarity, financialisation, and the doubling of the global reserve army of labour with the dissolution of the socialist bloc. And, of course, computerisation and automation.

Artificial intelligence is coming on top of this forty-year tendency. As Daco also notes, the productivity gains that explain today's divergence rest, to a large extent, on the time before the artificial-intelligence boom; a decade of automation, the pullback in hiring after the pandemic, and capital expenditure are the source of it (Shopifreaks). This is why Fortune's headline is very apt: the workers' share is at a record low "before the AI boom even begins".

A table of periods
PeriodTechnologyBalance of powerEffect on labour's share
1800–1840Steam, the weaving machineUnorganised workers, banned unionsProductivity rises, the wage stands still ("Engels' pause")
1840–1914The railway, electricity, steelUnions and workers' parties riseWages rise slowly
1945–1975The Fordist assembly lineStrong unions, the socialist bloc, the "fear of October"Labour's share is high and relatively constant (60–65% in the US)
1979–2022The computer, the internet, the global supply chainDeunionisation, financialisation, coupsThe gap between productivity and the wage opens
2022–…Generative artificial intelligence, agentsMonopolised platforms, a fragmented working classLabour's share is at a record low; even the Fed is warning

The lesson of the table is clear: the same technology gives different results of distribution under different balances of power.


Class Analysis: The Mechanism Behind the Paradox

What does "capital-biased innovation" mean in Marxist language?

The concept of "capital-biased innovation" that Barr uses is mainstream economics rediscovering an old truth in its own language. The explanation of the writer at Haver Analytics gives the technical skeleton of this: because the elasticity of substitution between artificial-intelligence agents and some workers is high, the rapid fall in the relative cost of artificial-intelligence agents raises the capital/labour ratio by more than the fall in the relative cost of capital; as a result, capital's share of income rises, and labour's falls (Haver Analytics).

Marx names this "the rise of the organic composition of capital": in production, living labour (variable capital) is replaced more and more by dead labour (constant capital: the machine, the model, the data centre, the chip). The difference is this: in earlier waves the replacement was, in the main, in the field of manual labour. Today, for the first time, mental labour, that is analysis, writing, translation, coding, customer service, design, is being replaced at a mass scale. What we called mental Taylorism is exactly this: the application, to the labour of the mind, of Taylor's breaking up of manual labour, measuring it, and mechanising it.

Whose labour is dead labour?

Here we come to the most vital point of Marxist analysis. What is an artificial-intelligence model? It is the distilled form of the texts billions of people wrote, the pictures they drew, the code they wrote, the answers they gave, the labels they corrected. It is the form, turned into constant capital, of humanity's common knowledge, of what Marx called the "general intellect" in the "Fragment on Machines" in the Grundrisse, from the labelling worker in Kenya to the open-source developer in Turkey, from the Wikipedia volunteer to the engineer answering a question on a forum.

In that fragment Marx writes that, with the development of modern industry, the production of wealth will depend less and less on direct labour time, and more on the application of science and of social knowledge to production. This is a deep contradiction for capitalism: while wealth is born more and more from social knowledge, an attempt is still made to measure it by individual labour time and to gather it under private property.

Artificial intelligence is the barest form of this contradiction. The rise in productivity is born of a social accumulation; but the return flows to the shareholders of a handful of companies. In our essays The Expropriation of the General Intellect and distillation we named this "digital enclosure": the enclosing of common knowledge and its turning into private property. The fall of labour's share is the reflection of this enclosure onto the table of distribution.

Monopoly: the economy not of competition, but of concentration

Barr's warning of an "unprecedented advantage" is in fact the official recognition of the present form of the monopoly capitalism that Lenin analysed a hundred years ago in Imperialism. The computing power, the data, the energy and the capital required to train the leading models have reached such a scale that only a few companies and states can join this race.

The effect of monopoly on labour's share works through two channels:

  1. A margin above cost (the markup): Companies protected from competition can hold prices far above cost. This extra profit swells capital's share in national income.
  2. Monopsony: When few employers remain in a sector or a region, the workers' bargaining power falls, and wages are pressed down.

Artificial intelligence strengthens both at once. On the one hand, the large companies that own artificial intelligence gain share in their sectors; on the other, artificial intelligence sits at every worker's bargaining table as "an alternative that can take the worker's place". The boss now holds in their hand not only the threat of unemployment, but the threat "the model can do this job too". In The Human in the Loop and Forty Percent of the Code we discussed the effect of this threat on software workers.

The reserve army of industry: new layers

According to Marx, capitalism constantly produces an army of the unemployed and the half-employed in order to keep wages under pressure: the reserve army of industry. Mechanisation is the basic source that feeds this army.

What Barr and the ILO draw attention to is that this army has begun to be fed by a new layer: young, educated, white-collar workers. In February, Barr relayed a study based on ADP payroll data, said that the employment of workers at the start of their careers, in occupations highly exposed to artificial intelligence such as software developers and customer-service representatives, had fallen relative to the others, and, adding that entering a weak labour market has lasting negative effects on earnings, said "for these workers the short term can have long-term consequences" (HR Executive).

This is especially important for Turkey. As we showed in our unemployment file, broadly defined unemployment and youth unemployment are already very high. If the first rung of the ladder in front of young people who graduate from university and look for an "entry-level" position is pulled away by artificial intelligence, Turkey's reserve army of labour will grow still more crowded with diploma-holding young people.

The contradiction of demand: the system's own limit

The fall of labour's share is a problem for capitalism too, and this is the real source of Barr's worry. Workers do the large part of total consumption. As wages stay behind productivity, the effective demand that will buy the goods produced narrows. One analysis writes this openly: artificial intelligence, by encouraging investment, can raise the natural rate of interest, but the erosion of labour's share at the same time can weaken consumer demand and offset growth (Pulse/Advisor Perspectives).

Marx resolves this as the "crisis of overproduction": while capital tries to expand production without limit, it keeps the worker's power of consumption limited. This contradiction can be postponed for a while by debt, by financial bubbles, by asset prices. In an economy in the US where stock-market wealth approaches multiples of disposable income, asset performance has become the basic determinant of consumption, especially in high-income households (Fair Observer). That is, consumption is fed less and less by the wage, and more by the share portfolio. This means tying the future of the economy to the value of artificial-intelligence shares, and it explains why the argument over the AI bubble is so intertwined with distribution.

The labour aristocracy and the wage paid in shares

One more detail: highly skilled engineers and specialists in the artificial-intelligence sector often receive share-based pay; this also ties their income to the performance of capital (Fair Observer). This is a contemporary form of Lenin's concept of the "labour aristocracy". The workers on the upper floor of the technology companies, to the extent that they take a share of capital's gain, tend to break away from the rest of their class. In Beyond the Frontier we described this layer as a section swinging between the "mental proletariat" and the "labour aristocracy". Binding this layer to the rest of the class is one of the most important tasks of the organising of computing labour.


Contested Points: Let Us Be Honest

Class analysis is not hiding the complexity of the data. It is reading it correctly. So let us set down the objections too.

Objection 1: The cause of the fall is not artificial intelligence. True, at least for now. Daco and others say that the source of today's fall rests on the time before artificial intelligence. Morgan Stanley economists also note that tying the latest leap in productivity to artificial intelligence or to automation is "an open question" (Fortune). Our thesis is already this: artificial intelligence is not creating the tendency. It carries the potential to take it over and speed it up. The problem is not in the technology. It is in the property relation into which the technology is set.

Objection 2: If it is measured correctly, there is no fall. The writer at Haver Analytics argues that, when production and import taxes are taken out of the denominator, no fall in labour's share is seen even three and a half years after the release of ChatGPT (Haver Analytics). This is an objection of method that has to be taken seriously. But it does not change two things. First, the long-term break that has gone on since 1979 is seen in every measurement. Second, the same writer also foresees that the spread of artificial-intelligence agents will show itself in a fall in labour's share; the argument is not about "whether it will happen", but about "whether it has begun".

Objection 3: Productivity and the wage have drawn close again in the recent period. Some studies show that the annual gap between productivity and the real wage narrowed in the 2023–2026 period (GE Review). This is the effect of the tight labour market after the pandemic and of the 2023 strike wave (UAW, SAG-AFTRA, Kaiser). It confirms exactly what we say: when workers are organised and bargaining power is high, the share can be taken back.

Objection 4: If artificial intelligence gets cheap, everyone gains. Barr mentions this possibility too: competition can cheapen artificial intelligence and spread it, and many businesses can meet their need with smaller language models (Stocktwits). Cheap goods can cheapen the worker's consumption basket. But Marx's analysis of relative surplus value tells exactly this: the cheapening of consumer goods is the condition of capital paying the worker less. Cheap artificial intelligence, on its own, does not raise labour's share; it can make it easier for small capital to compete with big capital, and that is all.


What Does the Fed Say, What Does the Class See?

The Fed's frame (Barr)The class reading
Artificial intelligence may be "capital-biased innovation"The organic composition of capital is rising; living labour is being replaced by dead labour
The problem is a "short-term disruption"The problem is structural: the speeding up of the break in distribution that has gone on since 1979
The advantage of the leading companies may be "unprecedented"Monopoly capitalism; the enclosure of the general intellect as private property
The solution: job creation, education, matching workers with new jobsEducation does not create bargaining power; organisation does
The solution: competition, small models, broad accessAccess is not property; to use is not to own
Inequality "may increase"Inequality is already structural; the richest 20% take 52% of income
The Fed can intervene only in cyclical problemsIn 1979 the Fed consciously changed distribution in favour of capital
Society "should prepare"Which society? The working class should prepare with its own agenda

What stands out in Barr's proposals is that the question of property is never asked. Education, job matching, competition policy: all of these foresee the worker adapting themselves better to the market, or the market working more "fairly". No one asks whose property artificial intelligence ought to be. Yet labour's share depends exactly on the answer to this question.

Even Barr himself had accepted in February that the Fed could intervene only in cyclical, not structural, changes in employment (American Banker). That is, the central bank sees the problem, but it also knows that the solution is not in its own toolbox. The solution is political and a matter of class.


Three Questions: Whose Hands, Whose Control, Whose Benefit?

Let us apply to the question of artificial intelligence and distribution the three questions we at Knowledge Commons ask of every technology.

Whose hands? The leading models, the data centres that train them, the chips and the energy infrastructure are in the hands of a handful of US and Chinese companies. The shares of these companies are also, in the main, in the hands of the richest households and the big funds. The knowledge that is the raw material of the model is humanity's common accumulation. The means of production is social, its property is private.

Under whose control? It is not the workers who decide where, how, and in place of which worker artificial intelligence will be used. It is managers and shareholders. The use of artificial intelligence in the workplace is not a subject of collective agreements. It is counted as a part of the right of management. As we showed in The Card Around Our Necks, algorithmic management and surveillance are also part of the same apparatus of control.

For whose benefit? The figures are in the open: while productivity rises 3.6%, unit labour cost falls, profit margins break a record, and labour's share is at a historic bottom. In today's order, the productivity return of artificial intelligence is for the benefit of the shareholders. The share that is reflected in wages is only as much as the workers can take back by force.


A Special Note for Turkey

Artificial intelligence is not at the head of the mechanisms that lower labour's share in Turkey. Here distribution is changed in capital's favour, much more crudely, through inflation, the tax system, the minimum wage turning into the general wage, deunionisation and subcontracting. The concluding sentence of DİSK-AR's report sums up this mechanism well: inflation is "a suction pump" that takes from the workers and transfers resources to the rich (Evrensel).

But artificial intelligence will affect Turkey by two roads:

  1. As a peripheral country: Turkey is not the producer of artificial-intelligence technology. It is its user. This means that the monopoly rent to be born of artificial intelligence flows to companies abroad, while the companies in Turkey lay off workers in order to lower costs. Value flows out, and unemployment stays inside.
  2. In the service and white-collar sectors: Fields such as call centres, banking, accounting, translation, software subcontracting and content production employ a wide mass of white-collar workers in Turkey, and this mass is to a large extent unorganised. This will be the first place that artificial-intelligence replacement hits.

That the rate of unionisation in Turkey is low, and that collective-agreement coverage stays at about 10 percent, means that the working class is defenceless in the face of the artificial-intelligence wave. Without this picture changing, the rise in productivity that artificial intelligence will bring cannot be expected to be reflected onto the worker.


What Is to Be Done? Concrete Tasks

Labour's share is not a law of nature. It is the result of a struggle. The question, then, is not "what will artificial intelligence do", but "what will we do".

On the union plane
  • Carry artificial intelligence to the collective-agreement table. The use of artificial intelligence in the workplace, decisions of replacement, the change in workload, and surveillance should be a subject of bargaining, not a right of management. The artificial-intelligence clauses that the Hollywood screenwriters (the WGA) won in the 2023 strike showed that this is possible.
  • The demand for a productivity share. The demand for a wage rise should be tied not only to inflation, but also to the rise in productivity. The formula "inflation + productivity" is an old demand that the union movement has forgotten.
  • The shortening of working hours. The most direct form of socialising the rise in productivity is to work less for the same wage. The demand for a 32-hour week is the eight-hour-day demand of the age of artificial intelligence.
  • Organise the white collar. Organising the workers of computing, finance, the call centre, the media and design is the first line of defence against artificial-intelligence replacement. The Computing Worker's Handbook was prepared for this.
On the political plane
  • The taxation of the artificial-intelligence rent. The extra profit of the monopolist artificial-intelligence companies should be transferred to public benefit.
  • A public artificial-intelligence infrastructure. Computing power, data and models should be under public and social control. This is the only realistic road to breaking the monopoly of a handful of companies.
  • The commons of data and knowledge. Against the enclosure of the social knowledge that is the raw material of the models, open data, open models and cooperative structures should be defended. The value produced from common knowledge should stay common. The theme of the Odak 2026 conference of the Free Software Association (ÖYD) sums this up well: "The People's Data, the People's Model." (An Invitation to Odak 2026)
  • Tying the minimum wage to national income. As DİSK proposes, the minimum wage should be at least 60% of national income per person (BirGün).
On the intellectual plane
  • Stand against the discourse of "technological fate". The discourse "artificial intelligence is coming, adapt" presents distribution as if it were a natural event. Distribution is a class relation, and it can be changed.
  • Claim the data. Translate indicators such as labour's share, unit labour cost, and the productivity-wage gap into union and political language. DİSK-AR's reports are very valuable in this respect; they should be spread.
  • Bring common knowledge together with common property. As we said in We Will Bury Capitalism: a common product is not common without common property.

Dear Young Comrades

If a central banker stands up and speaks of "the gain going to capital, labour's share falling, inequality growing", we can read this in two ways.

First, there is a worry inside the ruling class. The trillions of dollars flowing into artificial-intelligence investments are being tried to be won back by being cut from the workers' wages and from their future; they are afraid that at some point this could lead to a social explosion. The reaction rising against data centres in the US midterms is the first sign of this fear.

Second, and more important: this admission is the confirmation of what we have been saying for two hundred years. The machine does not save the worker on its own. The steam engine did not, the assembly line did not, the computer did not, and artificial intelligence will not either. But the machine does not have to enslave the worker either. Whose hands the machine is in, under whose control it works, and for whose benefit it is used, depends on the balance of power between the classes.

In the 1840s the English workers raised their wages by organising. After 1945 the Western workers won a share of productivity with their unions and with the pressure created by the fear of socialism. In the Spring Actions of 1989 the Turkish working class thawed, in the squares, the wages that 12 September had frozen. Each time, the share went to those who were strong enough to sit at the table.

Today artificial intelligence is born of humanity's common knowledge. This puts in front of us a truth seen, for the first time in history, this openly: the source of wealth is social. The distribution of wealth, then, must be social too.

Labour's share is shrinking, because the knife that cuts the cake is still in their hands. To take the knife, one first has to be able to sit at the table. And the only way to be able to sit at the table is to be organised.

The machine is new, the chain is old. But those who recognise the links of the chain also find the way to break them together.

For better days, all of us together.


Sources

The Fed and official figures
News on Barr's 12 November speech
Analyses
The ILO
Turkey
Classic texts
  • David Ricardo, On the Principles of Political Economy and Taxation, 3rd edition (1821), the chapter "On Machinery".
  • Friedrich Engels, The Condition of the Working Class in England (1845).
  • Karl Marx, Grundrisse (1857–58), the "Fragment on Machines".
  • Karl Marx, Capital, Volume I (1867), Part Four: "The Production of Relative Surplus Value", Chapter 15 "Machinery and Modern Industry"; Chapter 25 "The General Law of Capitalist Accumulation".
  • V. I. Lenin, Imperialism, the Highest Stage of Capitalism (1916).
  • Robert C. Allen, "Engels' Pause: Technical Change, Capital Accumulation, and Inequality in the British Industrial Revolution", Explorations in Economic History, 2009.
  • Nicholas Kaldor, "Capital Accumulation and Economic Growth" (1961).
Related essays from Knowledge Commons

This essay is shared under the CC BY-SA 4.0 licence. It is a living document; it will be updated as new figures arrive. The BLS will announce the revised figures for the third quarter on 8 December 2026.

Related Posts

Newsletter