Microsoft recently offered voluntary redundancy to 60,000 employees, around 27% of its global workforce. This was of course presented as an issue of flexibility and delivered as a courteous invitation to those who might already be contemplating a better future elsewhere. Yet, the vocabulary of corporate hygiene cannot hide a confession about its own insane objectives. The company effectively announced, in the blandest possible corporate language, that human labour is an adjustable liability in the face of a far more urgent object of devotion: compute.
The Wage as Liability
As it extends exit packages across much of its workforce, Microsoft is preparing to commit roughly $170 billion in 2026 to AI infrastructure (down from an earlier $175 billion forecast) for data centres, chips, cooling, and energy – the vast material apparatus required to keep wealth accumulating at warp speed. The wage relation – that old social compact in which people sell their time and capacities in exchange for a share of the surplus they help to produce – is being progressively converted into silicon, land, electricity, and capex projections. It is therefore not enough to say that workers are being replaced by AI, which of course they are. They are also being financially dispossessed by the very infrastructure that promises to make them unnecessary.
This is why the idiom of “voluntary exit”, now deployed across much of the Western economy, deserves particular critical attention. Nothing about it is voluntary in any meaningful sense when the alternative is to remain inside an organisation reorganising itself (“restructuring”) around the assumption that fewer people will be needed. Offers of this kind launder a strategic decision into an individual choice: leave with a package, the company suggests, and you may be doing yourself a favour. Meanwhile, the real decision has already been taken – a decision concerning the imperatives of an AI build-out whose costs are socialised through labour insecurity while its projected returns are privatised as titles to wealth by the shareholders.
What we are watching is certainly not an unfortunate side effect of technological progress, let alone a sacrifice we must endure for the greater good to come, as the techno-optimists would have it. It is, rather, an agonising work society turning on itself, dismantling the wage relation to finance the machinery intended to supersede it. If the work society, founded on the mass exploitation of human labour, should not become an object of nostalgia, its automated, hyper-financialised replacement cannot fail to appear as its ghastly asocial double. As such, it is already delivering growing numbers of “workers without work” into socio-economic immiseration – a cut-throat competition for survival without precedent in modern history.
Here, we should be wary of any philosophical temptation: while this condition of self-emptying might be thought necessary to trigger radical change, there is no guarantee that it will. The implosion of the work society may well drag on for decades, even centuries, as a hybrid “neo-techno-feudal capitalism” becomes normalised. Microsoft supplies the blueprint for this supposed “brighter future”: it cannibalises the social substance of work – jobs, wages, security, collective capacity – in order to build the automated future ahead of schedule. This is the material inverse of the techno-optimist delusional fantasy. Marc Andreesen’s manifesto assures us that technology ‘drives wages up, not down’ and that the ‘techno-capital machine’ of markets and innovation ‘spirals continuously upward’ for everyone. What Microsoft demonstrates is the opposite: technology is the instrument through which capital liquidates the wage relation, and the upward spiral is financed by the workers it expels. The worker is no longer treated primarily as a source of value; she is becoming the temporary financing mechanism for her own programmed obsolescence.
A quick look at the figures exposes the perversity of this arrangement. Microsoft employed roughly 223,000 full-time workers as of June. Its current buyout programme covers vice-presidents and below across the United States, Canada, the United Kingdom, and India, spanning Windows, Copilot, Xbox, finance, and HR. Datacentre, marketing, and cybersecurity teams are exempt, since the productive core of the AI build-out must be protected. The rest are offered a “dignified exit.” Meanwhile, the company’s capital expenditure reached $35.8 billion in the June quarter, more than double the previous year’s figure, and it returned $10.2 billion to shareholders. Again, the message couldn’t be clearer: capital increasingly treats the wage as a cost to be minimised, even where it remains largely affordable. More precisely, it buys out the wage-earner to subsidise the machine.
The Template Spreads
What I’ve described above is the template of today’s corporate hygiene. The evidenceis quickly accumulating and can no longer be dismissed. Oracle’s workforce fell by 13% (approximately 21,000 employees) in fiscal year 2026, with the company citing AI adoption among the reasons for such reduction. Amazon announced 16,000 corporate cuts in January, while Meta shed roughly 8,000 employees (10% of its workforce), partly to offset heavy AI spending. Across the tech sector, 134,917 jobs were eliminated at 328 companies in 2026 as of late October, already surpassing the 122,606 recorded for all of 2025, averaging 459 losses per day.
In short, the horse has already bolted, and well beyond Silicon Valley. Barclays is cutting 5,000 jobs, 60% in support functions including developers, as it presents AI as a source of new efficiency. American Express is eliminating 5% of its workforce (roughly 3,500 positions) while raising its profit forecast. Its Chief Financial Officer described the goal as becoming ‘leaner, more agile, and also to prepare for AI.’ We are not talking about distressed firms liquidating under duress; these are profitable enterprises converting payroll into capital expenditure for the apparatus of automation. They may not consciously grasp that they are accelerating the fall in the rate of profit across the economy as a whole, but one thing is clear: they will keep doing it.
Here, the rapid expansion of surveillance threatens to close the circle, since the very infrastructure that makes labour disposable is also making populations legible and sortable. Palantir – the company of self-appointed billionaire-philosopher-benefactor Alexander Karp – has developed ELITE for ICE, which populates a map with deportation targets, assembles dossiers, and assigns “confidence scores” to people’s addresses by drawing on Medicaid data and other sources. ICE’s data stored on Microsoft’s Azure infrastructure reportedly more than tripled as the agency used Azure AI tools to analyse video and translate text. While Microsoft denies that its technology enables mass surveillance, the magnitude of the data increase suggests otherwise.
Despite marketing itself as a progressive tech giant, Amazon Web Services (AWS) serves as the lucrative digital backbone for ICE’s deportation machine, which has sparked fierce protests and petition campaigns from labour unions and immigrant advocacy groups. Activists and investigative reports have long exposed how Amazon is directly complicit in the surveillance state by hosting the Department of Homeland Security’s massive biometric databases on its AWS GovCloud servers. These include facial records, fingerprints, and iris scans for over 230 million people. Amazon also provides much of the cloud infrastructure behind Palantir’s tracking software. This gives the company a direct role in the processing of personal information, including mobile phone records and driving licence data, which federal agencies can use to identify, locate and pursue individuals targeted for deportation.
And the surveillance also turns inward. Meta’s Model Capability Initiative logs employees’ mouse movements, clicks, and keystrokes on workplace computers in order to train AI agents. Over 1,500 employees signed a petition calling Meta a ‘factory for extracting employee data.’ The company then offered 30-minute ‘privacy breaks’ but did not abandon the programme. The bottom line is that labour’s residual activity is harvested as raw material for the machine that will execute the final displacement.
The Interregnum of the Machine
As mainstream media have recently begun to acknowledge, the financial architecture sustaining this dystopian constellation is now showing “signs of fragility”. None other than the Bank for International Settlements (BIS) – the “central bank of central banks” – warns that the five largest hyperscalers are committing over $1 trillion to AI investment, with ‘circular financing’ arrangements (chipmakers and cloud providers holding stakes in AI labs that then contract to purchase their chips) creating opaque exposure where the same asset may be pledged multiple times. The BIS now explicitly compares the logic of the massive shock in the making to that of recent speculative manias: capital cannot find profitable investment in labour, so it bets on automation. Apollo’s Chief Economist Torsten Slok spells out the implicit fragility of the bet: ‘Despite massive investments and widespread hype around artificial intelligence, the technology has delivered zero financial payoff to corporate America outside of tech.’ One does not need to settle the old question of technology’s role in value creation to recognise the present contradiction: there is no empirical evidence whatsoever that AI is lifting productivity growth in the broader economy.
We should therefore be precise when invoking Antonio Gramsci’s interregnum. The old work society is dying (Microsoft’s 60,000 buyouts is among its latest death certificates) but we should not fool ourselves: what cannot be born yet is not the new as “radical emancipation”, but the automated afterlife of the old order. Why? Ultimately, because its financing depends on liquidating the very demand (wage labour) that would make it profitable. It is for this reason that the interregnum is populated by a chain of morbid symptoms: bubbles, surveillance, manipulations, disavowals.
In a recent essay, Alenka Župančič asks whether the interregnum might not be the regnum itself – whether there is anything ‘on the other side’ – and she is right that the injunction to ‘adapt to the new reality’ is an ideology that wants us to accept the worst of the old order under a new name. But she does not articulate what is actually imploding: the self-valorising movement of capital itself, which can now persist only by feigning undeadness – by simulating the growth it can no longer generate. Her critique remains at the level of the subject’s disavowal, without engaging the historical circumstances that produce it. The nightmare Župančič describes is not a transhistorical condition of the subject. It is the subjective form of a specific objective crisis: the exhaustion of a mode of production that is now liquidating the wage relation, and is therefore unable to reproduce the elementary social bond upon which its own profitability depends.
Župančič, in other words, does not ask why the status quo has become ‘too costly for the system itself’. My answer is the one I have been trying to articulate for years: in its necessary alliance with technology, capital increasingly treats the wage as unaffordable – not because it cannot pay, but because capital serves only its headless self-expansion, indifferent to the social destruction it wreaks. This is why it is making the wage relation unworkable, while also enjoining us to embrace the delusional fantasy that the same technology will deliver us from the trouble it creates.
Given this objective impasse, the impossibility around which a “we” might form is not merely symbolic. It is the material impossibility of a system that is hollowing out its own social basis while building the apparatus of control required to contain the fallout. Microsoft asks workers to leave so that it can build the machines that will make their departure permanent. The workers are not merely being replaced; they are being asked to fund the conditions of their own technological redundancy. That is a ransom note written in depreciation schedules. In the meantime, ICE uses Microsoft’s cloud to track deportable bodies, while Meta records its employees’ keystrokes to train the agents intended to replace them. The circuit closes: labour’s social substance is converted into financial capital, and part of that financial capital is converted into an architecture of mass control and manipulation. As I argued in a recent note, the same AI machinery that identifies and tracks populations also provides the automated apparatus for their elimination – deportation is just one modality among others.
The work society is being abandoned while also being replaced by something colder, financed by the same capital that once depended on the wage. That is our interregnum. Our only chance lies in the failure of this apparatus – not because failure automatically delivers “the new”, but because it keeps the new from being foreclosed in advance.