Why Wall Street panicked over a sci-fi blog post

Last year, investors worried that AI would crash the economy by making too little money.

Now, they fear it will do so by making too much.

On Sunday, a little-known financial analysis firm called Citrini Research published a piece of science fiction: A memo dated June 2028, in which its researchers sketch a pocket history of “the global intelligence crisis” — an AI-triggered meltdown of the world’s financial, economic, and political systems.

In this account, the problem isn’t that AI proves unprofitable — and America’s data centers become rusted-out memorials to a 21st century Tulip Mania.

In Citrini’s telling, AI does exactly what its boosters promised (at first, anyway). The technology fuels rates of productivity growth unseen since the 1950s, generates mind-boggling profits for its owners, and massive GDP gains.

But it also irrevocably devalues white-collar labor and rapidly destroys a wide array of major businesses. Over time, the AI boom eats the rest of the economy. Growth and the S&P 500 both collapse, unemployment tops 10 percent, the mortgage market wobbles, the Occupy Silicon Valley movement blocks the entrance to OpenAI’s offices — all while the big labs keep raking in cash.

Such counterintuitive soothsaying might seem unremarkable. Bloggers sketch dystopian AI scenarios every day. Yet the Citrini memo appeared to do what few — if any — works of science fiction have done before: reduce the value of US stocks by more than $200 billion.

AI and the white-collar doom loop

To understand why the memo made such an impression, it’s worth examining its vision in more detail.

Citrini tells two distinct — but overlapping — stories. The first is about how AI could trigger a doom loop that destroys consumer demand. The narrative goes like this:

• AI advances render a steadily growing number of white-collar workers obsolete. By the end of 2026, Claude agents can do the work of “a $180,000 product manager for $200/month.” And the same is true of myriad other roles in consulting, software, real estate, financial advice, legal services, and more.
• Companies respond by cutting headcount and reinvesting their savings in AI.
• Higher investment in AI leads to more capable agents, devaluing the skills of even more white-collar workers.
• Displaced professionals slash their spending and drag down wages in the working-class economy: As laid-off McKinsey consultants start driving Ubers, rates for existing drivers fall amid heightened competition. And the same dynamic plays out in other sectors.
• AI’s productivity gains are generating massive wealth. But most of the returns flow to an extremely narrow elite. And when the super rich get richer, they don’t necessarily spend more money. Sam Altman needs only so many cars and TVs. So much of the AI industry’s profits don’t circulate back into the economy.
• Meanwhile, upper middle-class Americans are slashing their spending — either because they’re jobless or afraid they will be soon — and blue-collar workers aren’t seeing much wage growth. Thus, consumer demand collapses.
• As falling demand eats into companies’ profits, they scramble to find cost-savings. More and more discover that the easiest way to shore up their margins is to invest in AI and lay off workers.
• Higher investment in AI yields even more capable agents.
• More white-collar workers become obsolete.
• Companies respond by cutting headcount and reinvesting their savings in AI.

The cycle perpetuates itself with no natural brake.

How AI eats DoorDash

Citrini’s second story is a micro one, focused on how AI will disrupt certain businesses and industries. The core idea is that AI agents will turbo-charge competition — and shrink rents — throughout the white-collar economy.

Here’s a summary of the memo’s basic reasoning:

• Humans have a limited tolerance for comparison shopping. We don’t have the time or patience to exhaustively research every purchase we make. Instead, we default to familiar brands. Even corporate leaders do this when choosing which enterprise software to buy.
• This has enabled incumbent businesses to charge higher prices than perfectly competitive markets would allow. In total, trillions of dollars of enterprise value rests on this kind of rent extraction.
• AI agents don’t get impatient. And they can rapidly compare prices from across the entire internet.
• By 2028, people with no tech savvy will be using AI agents on a daily basis. They’ll simply click open an app and ask it to find them the cheapest flight, best apartment listing, or lowest-fee delivery app.
• Meanwhile, AI agents will massively lower the bar to entry in the markets for software, travel booking, real estate, food delivery, and much else. Using Claude Code, a single person — let’s call him Bob — can build a new delivery platform in an afternoon.
• On that platform, Bob offers lower fees than DoorDash or Seamless to consumers, restaurants, and drivers.
• In our world, Bob’s startup probably wouldn’t get anywhere; at first, it would have few participating drivers and restaurants. Consumers would stick with the brands they knew out of habit and convenience.
• But in the world where everyone is constantly using AI agents, hungry households don’t log into DoorDash to order pad thai — they ask ChatGPT to order them pad thai through whichever delivery service is charging the lowest fees. Likewise, restaurants and drivers don’t default to working with DoorDash but rather, ask their agents to sign them up for the least extractive platform. Bob’s app can therefore replicate DoorDash’s network in a matter of days.
• Thanks to people like Bob, rents in the food intermediary economy collapse.
• Similar dynamics play out in insurance (people and firms don’t automatically renew their coverage but engage in exhaustive comparison shopping), enterprise software (corporations can build their own in-house or choose from a cornucopia of agent-built startups, forcing down rates), real estate (traditional brokerages become unnecessary as AI agents eliminate information asymmetries between buyers and sellers), and elsewhere.

With margins collapsing, these rent-extracting firms accelerate the “do layoffs, invest in AI, see lower demand because no one has jobs, do layoffs” cycle.

And then there’s a financial crisis

In Citrini’s narrative, all this puts strains on the financial system. Traders and businesses made a lot of highly leveraged bets on the then-reasonable assumptions that 1) competition would not suddenly skyrocket throughout the consumer economy and 2) highly skilled professionals would almost always be able to pay off their mortgages.

AI explodes these premises, along with some financial institutions’ balance sheets. Credit conditions tighten. The recession deepens.

There are some problems with these stories

It can be difficult to know precisely why stocks moved up or down at any given time. But on Monday, it sure looked like Citrini’s memo weighed on markets, as shares of several companies it mentioned — including DoorDash — fell unexpectedly. Many financial publications attributed these declines to the Substack post.

This is all very weird.

For one thing, Citrini said it was merely exploring one under-discussed hypothetical, not claiming that its scenario was likely to happen.

For another, there are many reasons to think Citrini’s narrative is implausible — at least, in its full details.

Here are a few prominent objections to its reasoning:

AI won’t necessarily cause mass white-collar unemployment. Generative AI has been with us for a while now, yet US unemployment remains near historic lows. Even the most AI-exposed professions have been holding up well: Job openings for software developers actually increased over the past year and radiology employment has been rising.

Every previous general purpose technology has eliminated some jobs but also created new ones. The constraint on employment has historically been fiscal and monetary policy, rather than the capabilities of machines. Human wants are infinite. And companies have found countless ways to employ human labor in service of those wants.

There are reasons to think this time will be different — but also, reasons to think it will not. And our experience thus far provides cause for taking the latter seriously.

All that money invested in AI goes somewhere. That said, the memo’s core premise — that AI will displace a wide swath of white-collar workers — isn’t implausible. Its attempt to work through the implications, though, isn’t entirely convincing

In Citrini’s scenario, AI companies are reaping world-historic profits off the largest productivity gains in nearly a century — and plowing them into new infrastructure, at a rate of $200 billion per quarter. The sector’s boom continues, even as consumer demand collapses.

But it’s not clear that these two things could actually persist simultaneously.

When AI labs pour hundreds of billions into data centers, the money does not vanish — it flows to construction laborers, electricians, plumbers, HVAC technicians, steel workers, power plant supervisors, turbine technicians, engineers, and lawyers. And those people turn around and spend a portion of their earnings on goods and services in their local areas.

An economy in which AI monopolizes investment might not be ideal for national welfare. But it isn’t obviously inimical to growth-sustaining demand. Instead of addressing this point, Citrini simply asserts that the money spent on AI doesn’t circulate through the broader economy.

DoorDash exists for a reason. On a micro level, Citrini almost certainly overestimates how easily entrepreneurs can undercut existing firms with the aid of agentic AI.

Sure, Bob can vibecode “DoorSprint” overnight and offer lower fees. But providing competitive customer service, logistics optimization, insurance, or recourse for when a driver steals a pizza isn’t easy. And coding agents can’t instantly persuade restaurants, drivers, and consumers that DoorSprint can be trusted to faithfully mediate financial transactions. Which is a big problem since — in the world Citrini sketches — agentic AI would almost certainly be minting scam apps at industrial scale every day.

Collapsing rents would increase consumer demand. But okay, let’s say Citrini is right that AI will force down prices across a wide array of industries. That would effectively redistribute income away from business owners and toward consumers: When DoorDash is forced to charge lower fees, it makes less money and its customers’ dollars go further.

This sort of redistribution increases consumer demand. Working-class Americans spend a higher share of their incomes than wealthy shareholders do. So taking a dollar from the latter — and giving it to the former — tends to increase total consumer spending in the economy.

This dynamic wouldn’t necessarily outweigh the demand-destroying factors in Citrini’s scenario. But the memo fails to even acknowledge this tension between its two stories.

The government would probably do something. In Citrini’s narrative, America’s productive capacity skyrockets: Thanks to AI, the nation can generate drastically more economic value per worker-hour than it can today.

At the same time, millions of America’s most politically and socially influential citizens are ruined.

The first development would give the US government the capacity to restore growth: It could collect massive revenues from the beneficiaries of all that new production, and give the money to Americans who’d spend it.

The second development, meanwhile, would seemingly give Congress an impetus to enact such redistribution. When high-paid consultants, lawyers, financial analysts, and software engineers are all laid off at once, they are unlikely to suffer quietly. Privileged strata abruptly losing their expected status and living standards is the stuff from which revolutions are made. If their dispossession coincided with a collapse of the broader economy, politicians would likely scramble to redirect dollars in their general direction.

The future will be weird

All this said, Citrini’s note is still a fascinating and useful thought experiment. No one can be certain where AI is taking us. And the technology’s consequences could very well be destabilizing.

The fact that Citrini’s memo (apparently) rattled global markets is itself an indication of this moment’s radical uncertainty: Even Wall Street traders are struggling to distinguish science fiction from reality.