Tech giants are spending nearly $400 billion on AI infrastructure in 2025. Amazon alone plans to invest over $100 billion. Microsoft is allocating $120 billion. Alphabet has committed $85 billion. Meta raised its forecast to between $66 billion and $72 billion.

These numbers are staggering. For comparison, the European Union allocated less for its entire defense budget in 2023 than what these four companies are spending on AI this year.

But here is what none of these executives will tell you. There is a fundamental contradiction at the heart of this spending spree. And understanding this contradiction is crucial for anyone working in tech who wants to grasp what is really happening to their industry and their job security.

The Contradiction Capital Cannot Escape

We need to talk about something that capitalists desperately want to hide. Only living labor creates new value. This is not ideology. This is economic reality.

When you automate a job with AI, you are not creating a magic money machine. You are replacing a source of value creation with a machine that can only transfer value. The AI system embodies past labor. The workers who built it. The workers who created the training data. The workers who maintain the servers. But it does not create new surplus value.

This matters more than almost anything else in understanding the economics of AI automation.

Capital invests in AI to escape its dependence on labor. The $400 billion being spent on AI infrastructure represents a massive shift from variable capital to constant capital. From wages paid to workers to investment in technology. From people who create value to machines that transfer it.

Amazon cutting 30,000 jobs while spending $100 billion on AI perfectly illustrates this dynamic. They are trying to replace value-creating labor with value-transferring machines. In the short term, this looks brilliant. Labor costs drop. Profit margins soar. CEO compensation jumps 37%.

But this creates a problem that gets worse the more successful automation becomes.

Where Does Profit Actually Come From?

Let’s be precise about the Marxist theory here. Profit comes from surplus value. Surplus value comes from exploiting living labor. A worker produces more value in their working day than they receive in wages. The difference is surplus value, which the capitalist claims as profit.

A machine cannot create surplus value. It can only transfer its own value to the product. When you buy a robot or an AI system for $1 million, that $1 million worth of value gets transferred bit by bit into each product the machine helps produce. The machine depreciates. Its value spreads across all the commodities it touches. But it creates no new value.

This is why Marx called it constant capital. The value is constant. It does not grow. It just transfers.

Workers are different. Workers are variable capital. Pay a worker $50,000 per year and they might create $100,000 worth of value. The $50,000 difference is surplus value. This is where profit comes from. This is the only place profit comes from under capitalism.

Now imagine what happens when you start replacing workers with AI on a massive scale. You are systematically removing the source of surplus value from production. You are cutting off the very thing that generates profit.

Individual capitalists do not care about this. Amazon’s executives are not worried about the aggregate rate of profit in the entire economy. They only care about Amazon’s profit. And in the short term, cutting 30,000 workers while their AI systems transfer value looks like pure profit increase.

But what happens when Microsoft does the same thing? When Google does it? When Meta does it? When every major corporation follows the same logic?

The Tendency of the Rate of Profit to Fall

Marx identified this as the tendency of the rate of profit to fall. As the organic composition of capital increases (more machines, fewer workers), the rate of profit faces downward pressure.

This is not just theory. We can see this pattern play out historically. The rate of profit has declined over the long term in advanced capitalist economies, punctuated by periods of restructuring that temporarily restore profitability by destroying capital and reasserting control over labor.

The mechanism is straightforward. As capital investment in machinery grows relative to investment in labor, less surplus value is produced per dollar of total capital invested. The rate of profit (surplus value divided by total capital) falls.

Capitalists try to counteract this through various means. They intensify exploitation of remaining workers. They expand into new markets. They cheapen the elements of constant capital through technological improvements. They export capital to regions with lower wages. But these are countervailing tendencies, not permanent solutions.

The $400 billion AI spending spree is capital’s latest attempt to escape this contradiction. But the contradiction is structural. It cannot be escaped through better technology. It can only be displaced and deferred.

The Realization Problem Gets Worse

There is another dimension to this crisis that makes it even more severe. Even if AI systems could somehow create value (they cannot), there is still the problem of realizing that value through sales.

Who buys the products if the workers have been fired?

This is sometimes called the underconsumption problem or the realization crisis. Marx was careful about this. He did not reduce capitalist crises to underconsumption alone. The contradictions of production are primary. But the inability of workers to buy back the full product of their labor is a real feature of capitalism.

When Amazon cuts 30,000 jobs, those are 30,000 people with reduced purchasing power. When the trend spreads across the economy, purchasing power contracts. Demand falls. Inventories pile up. Factories sit idle. The crisis appears as a crisis of demand, but it originates in the contradictions of production.

Here is the cruel irony. Capital needs workers both to exploit them in production and to realize surplus value through consumption. Fire too many workers, and you destroy both sources of value. You eliminate the labor that creates surplus value, and you eliminate the purchasing power that realizes it through market sales.

The AI bulls will tell you that new jobs will be created. History shows us that automation creates new roles. This is true to an extent. But there are critical differences this time.

First, the pace of change is unprecedented. Goldman Sachs estimates that 6% to 7% of US workers could lose their jobs to AI adoption in the near term. Some researchers project that 30% of current US jobs could be fully automated by 2030, while 60% will see significant task-level changes.

Second, the new jobs being created require dramatically different skills. Recent analysis shows that 77% of new AI-related positions require master’s degrees. The workers being displaced from customer service, data entry, and administrative roles cannot simply transition to becoming AI ethics officers or machine learning engineers.

Third, and most importantly, even if workers retrain and find new jobs, the fundamental problem remains. Those new jobs still exist within the same capitalist relations. They still involve the same contradiction between value creation through living labor and capital’s drive to replace that labor with machines.

Tech Workers Need to Understand This

If you work in tech, this matters to you directly. Right now, you might feel safe. You might be building the AI systems that replace other people’s jobs. You might think your skills make you indispensable.

But look at what is already happening. Unemployment among 20- to 30-year-olds in tech-exposed occupations has risen by almost 3 percentage points since the start of 2025. The Society for Human Resource Management reports that 32% of computer and math-related professions have already been automated by 50% or more.

The logic of capital does not stop at the factory floor or the customer service center. It extends to software engineers, to data scientists, to product managers. AI that can write code will eventually write the code that writes code. The contradiction plays out at every level.

We are already seeing this. 14% of all workers have already been displaced by AI, with the rate higher among younger and mid-career workers in tech and creative fields. Your technical skills do not exempt you from the logic of capital accumulation. They just delay your displacement by a few years.

The $400 Billion Question

So what is really happening with this $400 billion in AI infrastructure spending? We are watching capital try to resolve its fundamental contradiction through technological means. We are watching the largest companies in the world invest unprecedented sums to replace human labor with machines.

Meta, Amazon, Alphabet and Microsoft intend to spend as much as $320 billion combined on AI technologies and datacenter buildouts in 2025, a figure that has since been revised upward. Amazon plans to shell out over $100 billion, up from $83 billion in 2024, with CEO Andy Jassy calling it “a once-in-a-lifetime type of business opportunity.”

But is it really an opportunity? Or is it a trap that capital is laying for itself?

The more successful this automation project becomes, the more severe the contradictions will grow. Each wave of layoffs reduces the aggregate purchasing power of society. Each replacement of workers with AI systems reduces the total surplus value being created. Each billion dollars shifted from variable capital to constant capital pushes the rate of profit downward.

In the short term, individual firms can profit enormously. Amazon’s profit margins prove this. But in the aggregate, the system faces deepening crisis. The $400 billion being invested in AI infrastructure today is setting up the conditions for a massive realization crisis tomorrow.

What History Teaches Us

This is not the first time capital has tried to resolve its contradictions through automation. Every major wave of technological change has promised to liberate capital from its dependence on labor. Every wave has failed to resolve the fundamental contradiction.

The industrial revolution replaced artisan labor with machine production. Capital celebrated. But the result was not smooth sailing. It was boom and bust cycles, overproduction crises, and eventually the Great Depression.

The computer revolution promised to boost productivity and create prosperity. It did boost productivity. But wages stagnated while profits soared. The inequality gap widened. And we got the 2008 financial crisis.

Now we have AI. The promises are the same. The contradictions are the same. The outcome will be the same, only more severe.

The Demand Crisis That’s Coming

Research shows that while 85 million jobs will be displaced by 2025, 97 million new roles will simultaneously emerge, suggesting a net positive job creation. But this aggregate number hides crucial details.

First, the jobs being destroyed are immediate and concentrated. Customer service, data entry, administrative work. These disappear quickly. The new jobs being created are diffuse, speculative, and require entirely different skills.

Second, the wage levels matter. If you fire a $50,000 per year customer service worker and eventually that person becomes a $30,000 per year gig economy worker, aggregate demand has fallen even if employment statistics look similar.

Third, and most crucially, there is a time lag. Workers need to retrain. They need to relocate. They need to find the new positions. During this transition, they are not earning. They are not spending. Demand collapses.

This is not speculation. We have seen this pattern before. Research estimates that net job losses from rising Chinese import competition over 1999-2011 were in the range of 2.0-2.4 million. The job losses from generative AI could be comparable, but concentrated in different regions and different industries.

The result was not smooth adjustment. It was devastated communities, political upheaval, and social crisis. We are about to do this again, but faster and on a larger scale.

Why Silicon Valley Won’t Admit This

The executives spending $400 billion on AI infrastructure are not stupid. Many of them understand economics. So why do they persist in a strategy that leads to systemic crisis?

Because individual capitalists are locked in competition. If Amazon does not automate, Microsoft will gain market share. If Google holds back on AI investment, competitors will eat their lunch. No individual firm can refuse to automate without being destroyed by firms that do.

This is the anarchy of capitalist production. Each capitalist makes individually rational decisions that collectively lead to crisis. Each CEO maximizes their own firm’s profit without regard for the system as a whole. And the system has no mechanism to coordinate investment decisions for collective benefit.

Capital is trapped by its own logic. The drive to maximize profit forces investment in automation. Investment in automation reduces the source of profit. The contradiction intensifies. Crisis looms.

But executives cannot admit this publicly. To admit it would be to admit that the entire system is irrational. That the $400 billion being spent on AI is not a path to prosperity but a road to crisis. That capitalism itself is the problem, not the solution.

So instead we get narratives about innovation and progress. We get promises of a better future through technology. We get assurances that the jobs will be fine, people just need to adapt. We get everything except honesty about the contradiction at the heart of capital accumulation.

The Empirical Evidence

Let’s look at what is actually happening right now, not theoretical possibilities.

Big Tech firms Amazon, Alphabet, Microsoft, and Meta reported they were set to spend as much as a cumulative $364 billion in their respective 2025 fiscal years, up from prior estimates. This represents an increase from $325 billion earlier in the year.

At the same time, job losses are mounting. Amazon announced 30,000 cuts. Microsoft, Google, and Meta have all conducted major layoffs while claiming record profits. Recent estimates from Goldman Sachs suggest that 6% to 7% of US workers could lose their jobs because of AI adoption.

The pattern is clear. Massive capital investment in automation. Soaring profit margins for the companies leading this charge. And systematic destruction of the jobs that created that value in the first place.

This is not progress. This is the contradiction playing out in real time.

The Technical Composition and the Value Composition

There is an important distinction to make here between technical efficiency and value creation. This is where many defenders of automation get confused.

Yes, AI makes processes more efficient. Yes, it can do certain tasks faster and with fewer errors than humans. Yes, productivity increases in a technical sense. But productivity in the technical sense is not the same as productivity in the value sense.

A firm that replaces 10 workers with an AI system can produce the same output with less labor time. The technical efficiency has increased. But the value created has not increased proportionally. The AI system transfers its own value to the product, but it does not create new value the way living labor does.

This is why firms can see their profit margins rise in the short term while the rate of profit in the economy falls in the long term. The individual firm captures a larger share of total surplus value by reducing its labor costs. But the total surplus value being produced in the economy declines as workers are systematically replaced by machines.

The aggregate effect is crisis. But by the time the crisis manifests, the CEOs who made these decisions have already collected their bonuses and moved on.

What About the Counter-Arguments?

Defenders of AI automation make several arguments. We need to address them directly.

“AI will create new jobs we cannot imagine yet.” This is technically true but misses the point. Yes, new jobs will emerge. But the fundamental relationship remains the same. Those jobs will still involve the exploitation of labor for surplus value. They will still be subject to automation in the next wave. The contradiction does not disappear. It just shifts to new sectors.

“Productivity gains will benefit everyone through lower prices.” This assumes a level of market competition that does not exist. In reality, productivity gains flow to capital in the form of higher profits. Amazon’s soaring margins while cutting 30,000 jobs prove this. The prices do not fall enough to compensate workers for their lost wages and job security.

“People have always feared automation and been wrong.” This assumes that the future will be like the past. But the pace and scale of AI automation is different. Previous waves of automation took decades to unfold and primarily affected physical labor. AI can automate cognitive work much faster. The adjustment period may not exist.

“This is just Luddite thinking.” The Luddites were not anti-technology. They were anti-exploitation. They understood that the question was not whether machines should exist, but who controls them and who benefits. We should learn from the Luddites, not dismiss them.

The Way Forward

So what does all this mean for those of us watching this unfold? What do we do with this knowledge?

First, we need to stop pretending that individual solutions will solve systemic problems. No amount of personal upskilling will protect you from the logic of capital accumulation. You cannot outrun the contradiction by learning to code or mastering the latest AI tools. The contradiction will catch you eventually.

Second, we need to organize collectively. The only force that can challenge capital’s drive toward automation is organized labor. Workers across all sectors need to demand control over how technology is deployed. Not just better severance packages or retraining programs, but actual power over the decision to automate.

Third, we need to question the fundamental premise that productivity gains should flow to capital. When AI makes a process more efficient, why should the benefit go entirely to shareholders? Why not shorter working hours for the same pay? Why not worker ownership of the AI systems themselves?

Fourth, we need to recognize that this is ultimately a fight over the means of production. Who owns the AI systems? Who controls the data they are trained on? Who decides how they are deployed? These are questions of power, not just questions of economics.

The $400 billion being spent on AI infrastructure represents a massive consolidation of productive power in the hands of a tiny number of corporations. We are watching the creation of the most centralized means of production in human history. And it is all privately owned, operated for profit, with no democratic accountability whatsoever.

The Crisis is Not a Bug, It’s a Feature

Here is what we need to understand. The coming crisis is not an accident. It is not a policy mistake that better regulation could prevent. It is not a failure of foresight by executives who should have known better.

The crisis is built into the structure of capitalism itself. The drive to maximize profit forces investment in automation. Investment in automation reduces the source of profit. The system cannot help but create the conditions of its own crisis.

Marx called this the “internal contradictions” of capitalism. The system works according to its own logic, following its own laws, and that logic leads inexorably to crisis. Not because anyone wants the crisis, but because the structure of the system makes crisis inevitable.

The $400 billion AI spending spree is not solving the problems of capitalism. It is intensifying them. Every dollar shifted from variable capital to constant capital, every worker replaced by a machine, every percentage point of labor share that shifts to capital share pushes us closer to a major crisis.

And when that crisis comes, capital will do what it always does. It will demand that workers pay the cost. It will insist on austerity, on wage cuts, on the destruction of social programs. It will use the crisis it created to reassert control over labor and extract even more surplus value from those who remain employed.

Unless we fight back. Unless we organize. Unless we question the fundamental logic of a system that treats human beings as costs to be minimized rather than the creators of all value.

Conclusion: The Contradiction Cannot Be Resolved

We began by talking about a single paragraph. The contradiction at the heart of automation. But understanding this contradiction means understanding capitalism itself.

Capital cannot escape its dependence on labor, no matter how much it invests in technology. Profit comes from exploiting living labor. Machines do not create new value. The more successful automation becomes, the more severe the contradictions grow.

The $400 billion being spent on AI infrastructure in 2025 is not evidence of capital’s strength. It is evidence of capital’s desperation. It represents a massive attempt to resolve through technological means what can only be resolved through social transformation.

The contradiction will not be resolved by better AI. It will not be resolved by more investment. It will not be resolved by workers learning new skills. The contradiction is structural. It is built into the system.

And that means the only way forward is to change the system itself. To move beyond a mode of production organized around profit and exploitation. To create a society where technological progress actually benefits those who create the value. Where AI systems are social resources, not private property. Where productivity gains mean less work and better lives for everyone, not unemployment and misery for some while a handful of executives collect billions.

This is not utopian dreaming. This is hard economic analysis. The contradictions are real. The crisis is coming. And the only question is whether we will allow capital to impose its solution through crisis and destruction, or whether workers will impose our own solution through organization and struggle.

The machines are not the problem. The problem is who owns them. The problem is a system that treats the elimination of jobs as a success story. The problem is capitalism.

And understanding the contradiction at the heart of automation is the first step toward understanding what we must do to get past capitalism and into something better.

Sources