Automation Pressures or Cost Cuts?

Mainstream explanations attribute these cuts to the rise of artificial intelligence (AI). Generative AI, a branch of machine learning that draws associations across massive data to generate images, text, or predictions without preset instruction, has been cast as a disruptive force capable of reshaping society. The CEO of Anthropic, one of the leading generative AI firms, says we need to stop “sugarcoating”: AI could “wipe out half of all entry-level white-collar jobs and spike unemployment to 10-20 percent in the next five years.”

There are reasons to be skeptical of this alarmism. A recent Yale Budget Lab study finds that “while the anxiety over the effects of generative AI on today’s labor market are widespread, our data suggests it remains largely speculative.” Similarly, Nobel laureate economist Daron Acemoğlu sees no evidence that the new technology will have “revolutionary effects” on the economy. Countering claims that AI will double US GDP growth, he estimates a modest gain of barely 1.5 percent over the next decade, with minimal impact on productivity. Other economists have found that even in settings where AI has been deployed, the economic impact has been minimal.

If this is true, then what explains the layoffs? The second most common story shifts the focus from Silicon Valley to Washington. In the wake of the Great Recession, central banks set interest rates extremely low, initiating what came to be known as the zero-interest rate period, or ZIRP. This era lasted through 2021. Between 2008 and 2021, cheap credit allowed firms to borrow, expand, and fund higher-risk investment. When rates started to rise in 2022, overleveraged firms faced higher debt-servicing costs and were forced to cut spending.

There is some truth to this narrative. Oracle, for instance, entered the AI race late and now faces $95 billion in long-term debt. It would need to increase its annual revenue to more than $300 billion by 2030 to justify its investments. Amid mounting pressure to meet financial targets, the company laid off 3,000 workers in September. Yet while some debt-financed firms like Oracle struggle, they are not representative of the sector as a whole. At best they explain only part of the story.

According to Goldman Sachs, AI-related companies issued $141 billion in new corporate debt in 2025, and Bloomberg reports that US tech companies raised another $157 billion in bond markets as of late September, up 70 percent from last year. These are impressive figures, but still only a small fraction of the roughly $1.5 trillion in total projected AI spending this year.

Too Big to Fail

Most AI expenditures are not by debt-ridden start-ups, but cash-rich incumbents such as Meta, Microsoft, and Amazon. These firms have low leverage, deep reserves, and easy access to cheap, high-grade credit. For example, founder of Good AI Capital Darwin Ling writes that Meta, which has already invested $14.3 billion in AI infrastructure and hiring, recently announced it would build two new data centers, which will deliver six gigawatts (GW) of power in total to run large-scale models. For perspective, one GW is equivalent to the energy of consumption of a small US state.

To fund this, Meta will combine $26 billion in debt and $3 billion in equity, the largest private equity transaction in history. This debt will not appear directly on Meta’s books, but rather through a special purpose vehicle (SPV), while Meta will lease back the infrastructure. This is intended to serve as a financial roadmap for other companies looking to invest in massive infrastructure projects for hyperscalers.

This reflects the broader circular financing logic core to the AI economy in which firms are simultaneously investors, clients, and creditors. The same small set of companies fund, supply, and sell to one another in an increasingly closed loop of mutually dependent monopolies. For example, Oracle, Nvidia, CoreWeave, and SoftBank trade $1 trillion worth of AI deals among themselves. Returns derive not from productivity gains but from a form of enclosure of access rights, or owning the platforms, data, and infrastructures on which others depend. As such, the system is structured so that those investing are increasingly self-reinforcing and too big to fail.

Or as the Economist explains plainly, “For a start, a lot of today’s spending could prove worthless. [But] the good news is that today’s financial system could probably absorb the blow.” The reason for this is that those providing credit are private-market funds funded by rich individuals and institutions rather than depositors, and AI start-ups are financed by well-capitalized ventures and sovereign-wealth funds that can withstand loss.

Layoffs as a Strategic Choice: Pitting Capital Against Living Labor

In spite of its strong financial position, Meta announced it would lay off 5 percent of its employees this week to trim “nonessential teams” and focus on “advancing AI.” The firings, like many other tech layoffs, are not driven by financial strain or genuine automation-related pressure but by a strategic choice to restructure contracts and weaken the position of labor. Meta can make these cuts because of its outsize economic and political power, which insulates it from market and social discipline and allows it to impose a new regime of accumulation on its workforce.

Under the guise of technological inevitability, companies are using the AI boom to rewrite the social contract — laying off employees, rehiring them at lower wages, intensifying workloads, and normalizing precarity. In short, these are political choices masquerading as technical necessities, AI is not the cause of the layoffs but their justification.

The growing share of white-collar workers rendered precarious or redundant by capital’s technological drive form a new surplus population or a pool of disposable and downwardly mobile workers used to depress wages and normalize insecurity. They are not external to capitalism but internal to its reproduction.

The commonly voiced injunction that workers ought to “reskill” or be left behind provides a moral and ideological justification for attacks on labor. But the effects of AI-driven reorganization will also be felt by the majority of workers who will not lose their jobs thanks to the new technology. This is because managerial reorganization under the guise of technological improvement puts pressure on workers to accomplish tasks in less time, take on new tasks, or absorb the tasks of those fired, all for the same pay.

In response to this push to extract more value from workers, unions must be willing to challenge these forms of reorganization. This means organizing around compensation, job security, data rights, and working conditions. Struggles of this kind are already taking place, led by unions looking to collectively bargain for workers affected by AI adoption.

Yet labor alone cannot win this struggle. The enclosure of the AI economy is sustained not only by private monopolies but by the state itself, which supports these companies through subsidies, permissive regulation, and political capture. Countering Big Tech will require a political project to reclaim democratic control over the use to which the products of human intelligence are being put.

In Karl Marx’s words, “Capital is dead labour, that, vampire-like, only lives by sucking living labour, and lives the more, the more labour it sucks.”