If the AI Bulls Are Right, the S&P 500 Crashes 38%

There’s a scenario circulating in macro circles right now that stops you cold. The argument: if the AI bulls turn out to be exactly right, the S&P 500 falls 38%.

It comes from Citrini Research, framed as a “June 2028 Macro Memo,” a dispatch written from a hypothetical future. I sat with it for a while after reading. It’s not a prediction. The authors are careful about that. But the internal logic is airtight in a way that’s genuinely unsettling. Most bearish AI takes focus on hype and failure. This one is about success, and why success might be the real danger.

The Structure: Growth That Doesn’t Circulate

In the scenario, October 2026. S&P 500 hits 8,000. Corporate earnings are at all-time highs. Productivity growth is running at its best pace since the 1950s. And real wages are collapsing.

This is what the authors call “Ghost GDP” output that shows up in national statistics but doesn’t flow through the real economy. It’s measured wealth that no one spends.

• A single GPU cluster replaces the output of 10,000 Manhattan white-collar workers
• The consumer economy (70% of U.S. GDP) starts losing its foundation
• Machines generate zero dollars of discretionary spending
• The authors’ framing: not “an economic cure-all” but “an economic pandemic”

The paradox is structural. Productivity gains have always eventually raised living standards, but that transmission mechanism runs through wages. Strip out the wage component, and you get growth statistics divorced from consumer reality.

When Friction Goes to Zero, the Middlemen Disappear

By early 2027, AI agents are handling consumer decision-making autonomously. This is where Citrini’s sharpest line lands: “Much of what we called relationships was just friction wearing a friendly face.”

Decades of habit-based brand loyalty evaporate when an agent optimizes every purchase. The moat protecting consumer businesses wasn’t preference; it was inertia. And agents don’t have inertia.

• DoorDash’s competitive advantage was “I’m hungry and I don’t feel like thinking, I’ll tap the app on my home screen.” Agents don’t have home screens.
• Buyer-side real estate commissions compress from 2.5–3% to under 1%
• Mastercard’s payment volume growth rate drops from 5.9% to 3.4% in Q1 2027
• Agent-to-agent transactions route around the 2–3% card interchange fee via stablecoin settlement

The businesses that looked like platforms were actually toll booths. When the road changes, the toll booth becomes irrelevant.

White-Collar Unemployment Hits the $13 Trillion Mortgage Market

This is the scenario’s most viscerally unsettling section. The top 10% of U.S. earners account for more than 50% of consumer spending. When their incomes get cut in half or disappear entirely, the economic shock is orders of magnitude larger than the headline unemployment number suggests.

The 2008 comparison is instructive, but the mechanism here is different. In 2008, the loans were bad from the start. Here, the loans were sound: the world changed after the ink dried.

• $13 trillion in residential mortgages were underwritten on the assumption that borrowers would maintain their current income
• Delinquency rates start rising in zip codes with FICO scores above 780
• San Francisco home prices fall 11% year-over-year in the scenario
• The authors write that people “borrowed against a future that humans can no longer reliably inhabit”

A 780 FICO score doesn’t mean much if the career that justified it no longer exists.

The Private Credit Daisy Chain

Zendesk becomes the scenario’s symbolic case study. Acquired for $10.2 billion in 2022, with $5 billion in direct lending structured around the assumption that its annual recurring revenue would keep recurring.

When AI agents handle customer service autonomously, that ARR becomes what the authors call “revenue that just hasn’t left yet.” And when you trace the “permanent capital” backing these deals to its source (insurance company annuity holders), the systemic exposure snaps into focus.

• Moody’s downgrades $18 billion across 14 PE-backed software debt issuers in a single action (April 2027)
• Apollo shares fall 22% in two days after Athene receives a negative financial stability outlook
• India’s $200 billion IT services export base fractures; the rupee falls 18% in four months
• The IMF begins preliminary consultations with New Delhi in Q1 2028

The private credit market was built on recurring revenue projections. Those projections were built on human labor being the scarce input. That assumption is being stress-tested now.

A Feedback Loop Without a Brake

Citrini calls the governing dynamic of this entire scenario the “Intelligence Displacement Spiral.”

AI capability improves → layoffs → consumption contracts → margin pressure → more AI investment → AI capability improves.

There’s no natural circuit breaker. The authors make the point explicitly: cutting rates to zero and buying every MBS on the market doesn’t change the fact that a Claude agent can do an $180,000-a-year PM’s job for $200 a month.

• The labor share of income falls from 56% in 2024 to 46% by 2028 in the scenario
• Federal tax revenues run 12% below the CBO baseline
• Corporate AI budgets grow while total operating costs shrink: OpEx substitution, not expansion
• Policy response speed fails to keep pace with the rate of technological change

This is what makes it a spiral rather than a correction. Traditional monetary and fiscal tools operate on demand. They can’t address a structural shift in what labor is worth.

The Canary Is Still Singing

The memo ends with a line that stays with you: “The canary is still alive.”

As of February 2026, the S&P is near all-time highs and the negative feedback loop hasn’t started. This is a scenario, not a forecast. The distinction matters. But the central paradox, that the more precisely correct the AI optimists turn out to be, the more dangerous things get, is worth sitting with.

“Are you standing on an assumption that won’t survive this decade?” It’s the cheapest question you can ask right now.