Why Capitalism Can't Survive AI, Part 2: The Shape of the Collapse

Part 1: This Time It's Different covered why AI will permanently replace most high-paid jobs. Parts 3 and 4 cover why UBI can't fix it and what actually can.

---

In Part 1, I laid out why AI is going to make tens of millions of the most highly paid workers permanently unemployable. Not just knowledge workers, but drivers, certain kinds of manufacturing workers, and many others. Not gradually, not over a generation, but in waves of corporate decisions over the next three to five years.

This article walks through what happens to the economy when those layoffs start. What are the realistic scenarios, and why will the standard government responses fail?

The human wreckage

Before the economics, think about what the layoffs do to people.

Millions of professionals who spent years and hundreds of thousands of dollars training for careers in law, accounting, engineering, marketing, consulting, and software development discover that their fields no longer employ humans. They're sitting in the houses and condos they bought when they were making big salaries, unable to make their mortgage payments, realizing that they are never going to work as knowledge professionals again. Not in a year. Not when the economy recovers. Never.

A whole generation of young people watches this happen and draws the obvious conclusion: there's no point going to college to train for a profession that won't exist by graduation. University enrollment craters. Colleges that were already struggling to fill seats face an existential crisis. The entire pipeline of professional education, from undergraduate programs to law schools to MBA programs, starts to collapse, because the jobs those degrees led to are gone.

And the displaced professionals don't just sit at home. They need income. So they start competing for the jobs that haven't been automated yet: manual labor, physical trades, service work. Lawyers applying for construction jobs. Software engineers competing with warehouse workers. Marketing directors trying to get hired as electricians' assistants. This flood of desperate professionals into the low-wage labor market drives wages down for the people who were already doing that work. The skilled factory worker who was making $60,000 a year is now competing with former accountants willing to work for far less because anything is better than nothing.

The result is evictions and foreclosures on a massive scale, across every income level. The million-dollar townhouse and the $1,200-a-month apartment. The professional class and the working class, hit by different mechanisms but drowning in the same crisis.

That's just the direct impact of the AI layoffs themselves. What makes it catastrophically worse is one of the best-understood dynamics in economics, and one that most people have never given much thought.

The demand spiral

Every wave of AI layoffs will drive disruption and displacement on a far greater scale than the layoffs alone. The reason is the demand spiral.

The mechanism is simple. When a large number of workers lose their jobs, they stop spending money. When they stop spending money, the businesses they used to spend at lose revenue. Restaurants lose regulars. Shops lose customers. Contractors lose projects. Those businesses start cutting staff. The newly laid-off restaurant workers and shop clerks stop spending. The businesses they used to patronize lose revenue. More layoffs follow. Each round of job losses causes more spending to disappear, which causes more job losses. It's a self-driving dynamic.

This is what happened in the Great Depression. Factories had more capacity than ever. They could produce more goods than at any point in human history. But nobody could afford to buy them, because the workers who would have bought them had been laid off. So factories shut down, which put more people out of work, which meant even fewer customers. The economy spiraled downward for years.

A demand spiral doesn't stop on its own. It keeps going until something external breaks the cycle, or until the economy has been so thoroughly destroyed that it can't fall any further.

If AI displaces 5% of the workforce directly, the demand spiral can push total job losses to 15% or 20% or higher, because every displaced worker's lost spending destroys other people's jobs. And unlike a one-time shock like a financial crisis, AI displacement will come in waves that could reach 50% or more of all jobs. Before the demand spiral from the first round of layoffs has time to stabilize, the next wave hits. Each wave triggers its own cascade. The demand spiral compounds wave after wave, because the technology keeps getting better and the layoffs keep accelerating.

Why confidence won't save us this time

After the Great Depression, governments figured out a trick. At the first sign of a downward spiral, they flood the economy with spending: print money, bail out banks, send checks, run deficits. Whatever it takes to prop up demand and give businesses confidence that customers will keep buying.

The key word is confidence. The trick works because it convinces businesses and consumers that the downturn is temporary. If you believe customers will come back, you don't shut your factory. You don't lay off your remaining staff. You hold on, because you believe things will go back to normal.

COVID was the clearest example in our lifetimes. At the beginning of the pandemic, most companies got ready to cut production, investment and employment, expecting the worst economic crash in a generation. But the government stepped in almost immediately with guarantees: trillions of dollars of stimulus, paying workers' wages directly, sending checks to households, and propping up businesses that would have collapsed. It worked, in part because most people bet that COVID would end in the not-too-distant future and that government would therefore be able to keep up the support for the duration of the crisis. As a result, companies didn't just hold on, they made massive investments.

Before COVID, the same thing happened with the 2008 financial crisis. And the dot-com bust before that. And the savings-and-loan collapse before that. In every case, the government jumped in, propping things up until confidence returned.

Before governments figured out the trick, demand spirals would eventually end on their own when things had gotten so bad that there was nowhere else to go but up. For centuries, the world economy had periodic crashes that were understood to be an unavoidable feature of the industrial economy. Without government intervention, the economy would hit rock bottom, and then some business owners would finally bet that demand could not fall any further. They'd start hiring. That hiring would put money in workers' pockets, which would stimulate spending, which would give other businesses confidence to hire. The spiral would reverse.

The AI layoffs will break both of these mechanisms.

When the government floods the economy with money during an AI displacement crisis, nobody is fooled. Everyone can see that these jobs are not coming back. No company is going to uninstall the AI that replaced its marketing department. Law firms are not rehiring associates. Consulting firms are not bringing back analysts. This isn't a recession where workers are waiting to be called back. This is a permanent structural change, and everyone can see it.

So the confidence trick doesn't work. Businesses that survive the initial wave don't invest in hiring, because they know AI is cheaper. Workers who still have jobs don't spend freely, because they can see what's coming for them next. The government can send all the checks it wants, but when everyone can see that the displacement is permanent, people hunker down. And that hunkering down accelerates the spiral instead of slowing it.

What if the government does nothing?

Consider the possibility that the government simply doesn't respond. This is not far-fetched. Political gridlock is currently the norm. Congress can barely pass a budget in good times. Imagine trying to pass a multi-trillion-dollar emergency spending bill while half the legislature is arguing about whether AI displacement is even real and the national debt is already at levels that dwarf those which helped trigger the French Revolution.

Maybe Congress passes something, but it's too small and too slow. Perhaps the political will for bailouts has been exhausted after decades of growing debt. Or the party in power believes the market will sort itself out.

In that scenario, the demand spiral runs unchecked. It's difficult to guess how bad things could get. Think mass evictions and foreclosures, millions of families literally going hungry, businesses closing down en masse, stock markets crashing and life savings vanishing. It's hard to imagine government not taking action at some point, even under political leadership that is staunchly anti-bailout. But if relief comes too late, the consequences will be disastrous.

When the government tries the standard playbook

The more likely scenario is that the government does respond. Stimulus checks. Extended unemployment benefits. Maybe it pays workers' wages directly, the way it did during COVID. Maybe it puts a moratorium on evictions and forces banks to accept late payments.

This is the standard playbook. It worked for COVID and for 2008 and, more or less, for every crisis since the Depression.

It will not work this time.

Imagine if COVID never stopped. Imagine if, instead of lasting two years, the lockdowns went on forever. Would the government have been able to just keep paying everyone's wages indefinitely? Obviously not. Even the short-term COVID response, covering a fraction of the workforce for a limited period, massively increased the national debt and triggered inflation by putting money in people's hands while the supply of goods was constrained. Multiply that by a much larger share of the workforce, including all of the highest-paid workers, and make it permanent, and the numbers become absurd.

The deeper problem isn't the money. It's that the playbook was designed for temporary crises. The government sends checks to bridge a gap. Workers wait out the downturn. Businesses hold on because they expect customers to return. Then the crisis passes, workers go back to work, tax revenue resumes, and the government starts paying down the debt it accumulated.

With AI, workers never go back to work. Tax revenue never resumes. The debt never gets paid down. The bridge goes on forever, which means it isn't a bridge. It's a permanent subsidy of a permanently broken economy. And every month that passes, more AI displacement hits, requiring more spending, pushing the debt higher, while the tax base keeps shrinking because the workers who used to pay taxes keep losing their jobs.

The moratorium on evictions can't last forever. Banks can't absorb unpaid mortgages indefinitely. Landlords can't survive without rent. At some point, the temporary measures expire, and the underlying reality reasserts itself: millions of people have no income and no prospect of income.

The key to the standard playbook is that it keeps confidence high because everyone sees the government is doing what's needed to defeat a temporary, irrational crisis of confidence. But in the case of the AI layoffs, the crisis of confidence is not irrational. Everyone can see that the jobs are gone forever, and no amount of cheerleading or stimulus spending by the government can change that.

What comes next

The standard playbook is exhausted. So the conversation moves toward something more radical: a permanent Universal Basic Income, or maybe a job guarantee in which the government pays workers to work even though no private firm wants to hire them. If AI is going to permanently displace workers, the government should permanently replace their incomes. Tax the AI companies, send everyone a check, or pay them for make-work. It's the obvious next step, and it's what most smart people writing about AI assume is at least part of the solution.

In Part 3, we'll examine whether that can actually work. Can the government sustainably tax AI-driven companies enough to fund permanent income replacement for tens of millions of displaced workers? The answer, when you run the numbers, is not what most people expect.