The 2028 Global Intelligence Crisis is a fictional macro memo that Citrini Research and Alap Shah published on Substack on February 22, 2026. It describes AI succeeding so well that it wipes out white-collar income and drags the economy down with it.
The authors call it a scenario, not a prediction. Markets still reacted, and economists argued about it for weeks.
This article explains what the memo claims and how its feedback loop works. It then covers the strongest criticism and checks the scenario against October 2026 data.
The short answer: the data do not show the memo’s crisis, but they do show early signs of weaker hiring in AI-exposed jobs.
What is the 2028 Global Intelligence Crisis?
The 2028 Global Intelligence Crisis is a scenario essay by Citrini Research and co-author Alap Shah, written as a memo dated June 30, 2028. In that memo, unemployment reaches 10.2% and the S&P 500 sits 38% below its October 2026 high.
The authors call the piece a thought exercise. Shah posed the question, Citrini wrote this part, and Shah wrote two companion pieces. Wikipedia credits James van Geelen and Alap Shah as the authors and notes the report was viewed millions of times.
The memo ends by reminding readers that it is February 2026, the S&P 500 is near its highs, and “the canary is still alive.”
How does the “intelligence displacement spiral” work?
In the memo, AI improves, firms cut white-collar staff to protect margins, and they spend the savings on more AI. Laid-off workers spend less, so firms cut again. The authors say the loop has no natural brake because its cause is a technology, not a business cycle.
Why does falling demand not slow AI spending?
The memo treats AI spending as a swap for payroll, not as new capital investment. A firm that spent $100M on staff and $5M on AI moves to $70M and $20M, so AI spending quadruples while total costs fall.
That answers the obvious objection. Weak demand would normally slow an AI buildout, but here the buildout comes out of the wage bill.
The memo also leans on concentration. It asserts that white-collar workers hold about half of US jobs but drive roughly 75% of discretionary spending. The text gives no source for that figure, so treat it as the authors’ claim.
How does job loss turn into financial stress?
The memo’s chain runs through private credit, insurers, and mortgages. Software-linked loans default first, including an imagined $5B Zendesk facility marked down to 58 cents.
Insurers that hold that credit then face tighter capital rules. The memo’s last step is prime mortgages, in a market of roughly $13 trillion, where borrowers with 780+ credit scores lose their incomes.
Its name for the result is “Ghost GDP”: output that appears in national accounts but never circulates through households.
How does the loop reach government finances?
The memo says income and payroll tax receipts fall as labor income weakens, while benefit spending rises, because automatic stabilizers assume displaced workers get reabsorbed.
No source reviewed here shows tax receipts weakening because of AI. The fiscal loop is a proposed mechanism, not an observed trend.
Why did the Citrini memo move stocks in February 2026?
US stocks fell on Monday, February 23, 2026, and Bloomberg and The Wall Street Journal said the memo played a part. The S&P 500 dropped 1.04% and the Nasdaq 1.13%. The memo was one factor among several.
A new tariff announcement landed the same day. IBM fell sharply as markets reacted to an Anthropic tool for modernizing COBOL code.
Software stocks were also sliding before the memo. One tally has the S&P 500 software and services index shedding about $1 trillion in the weeks after January 28, amid fears about new Anthropic agent tools.
One strategist quoted by Finviz said the sell-off built on AI fears that already existed. The memo gave a fragile market a vivid story.
What do the strongest critics say?
The strongest rebuttal came from Citadel Securities, which argued that real-time data show no AI-driven labor collapse. It also said adoption follows an S-curve and that energy and compute limits cap the speed of displacement.
Frank Flight’s Citadel note, titled “The 2026 Global Intelligence Crisis,” cited 11% year-over-year growth in software job postings and stable workplace AI use. His central line was that technology can improve recursively without adoption doing the same.
Gerard MacDonell of Front Harbor Macro called the memo an “allegorical essay” and said production generates income. Michael Bloch published a mirror scenario, “The 2028 Global Intelligence Boom,” arguing that falling prices from cheaper production are a gain in living standards, not a crisis.
The memo’s reply is that earlier technologies always created work that needed humans, while AI improves at the very tasks displaced workers would move into.
Does “production creates income” settle the debate?
No. Output pays someone, but the memo’s claim is that the income stops reaching households who spend it.
The dispute turns on four possible offsets:
- Lower prices keep purchasing power intact.
- Capital owners’ spending recycles profits into the economy.
- New work replaces displaced jobs.
- Government transfers redirect gains to households.
The memo’s rejoinder is that these offsets arrive too slowly, reach too few people, or fall short of the lost labor income. Its fictional Congress debates direct transfers and a tax on AI compute, and the memo argues that rate cuts cannot fix a problem rooted in the value of human intelligence.
Critics answer that 10% unemployment would trigger a far larger fiscal and monetary response than the memo assumes. Both sides are reasoning about a future that has not happened.
What does 2026 data say about the scenario?
The data do not show the memo’s crisis. Unemployment was 4.2% in September 2026 against the memo’s 10.2% endpoint, but Dallas Fed research finds AI exposure already trimming job postings in affected occupations.
| Indicator | Memo’s scenario | Latest observation | Reading |
|---|---|---|---|
| US unemployment rate | 10.2% (June 2028) | 4.2%, September 2026 (released Oct 2) | Lagging indicator; far from the endpoint |
| Monthly payroll growth | White-collar openings collapse | +29,000 in September 2026, about 90,000 expected | Soft hiring; claims remain historically low |
| S&P 500 | 38% below Oct 2026 highs | Record close of 7,818.93 on October 6, 2026; 7,811.54 on October 9 | Memo’s pre-crack peak (near 8,000) sits within about 2.5% |
| Labor share, nonfarm business | Memo’s own 56% to 46% path | 52.8% in Q2 2026, the lowest on record | Long decline; not attributable to AI |
| Job postings in AI-exposed roles | Collapsing white-collar openings | Texas postings down about 2.6% in 2025 from AI exposure (Dallas Fed, Sept 1, 2026) | Early demand effect; one state |
Why is unemployment a weak test on its own?
The memo’s 10.2% is a 2028 endpoint. Early effects would more likely show up as fewer entry-level openings, slower hiring, and flatter wages than as mass layoffs.
Analysts can sort indicators by timing. Job postings and hires lead. Employment and wages move with the economy. Delinquencies and defaults lag.
What did the Dallas Fed find?
The Dallas Fed’s September 2026 analysis uses a task-based exposure measure developed by Anthropic and links it to Lightcast job postings. By early 2025, more-exposed occupations posted roughly 8% fewer openings than less-exposed ones.
Postings are not layoffs. A separate Dallas Fed study found no rise in moves from employment into unemployment among young workers in highly exposed occupations. Fewer postings do not mean more layoffs: hiring demand is shifting, while job loss among young workers in exposed roles has not risen.
Texas graduates from more-exposed majors also had a relative decline in finding work within a year, and first-year earnings about 5% lower than peers between 2021 and 2024.
The strongest early signal is still state-level. A matching pattern in national job-openings data or in Indeed’s occupation-level series would be the next confirmation step.
Does the record-low labor share support the memo?
The direction matches, but the cause is unproven. The BLS series begins in 1947, ran mostly between 60% and 66% for decades, and started a sustained decline after 2000.
The memo’s 64%, 56%, and 46% figures do not match this series. Readers should not compare them directly.
Do financial-sector job losses prove AI displacement?
Not by themselves. Financial activities have shed 129,000 jobs since a May 2025 peak, mostly in insurance. The sources reviewed do not tie that loss to AI.
Attribution is hard in general. Two trackers counted the same 2026 layoffs: one found 56% named AI somewhere, the other labeled only 7.8% as directly AI-driven. Analysts call that gap AI-washing.
Which signals would confirm or weaken the scenario?
Seven signals follow the memo’s causal chain. Weak evidence is hiring softness in AI-exposed roles while income and spending hold. Stronger evidence is labor displacement, falling household income, weaker spending, and rising credit stress appearing together, with a traceable link to AI adoption.
| Signal | What to measure | Primary source | Strengthens the thesis | Weakens it |
|---|---|---|---|---|
| Hiring | Openings and hires in exposed occupations | BLS JOLTS; posting studies | Persistent gap versus unexposed roles | Postings recover or rise |
| Employment | Layoffs, unemployment duration, participation | BLS jobs report | Displacement spreads beyond entry level | Duration and claims stay low |
| Compensation | Real wages, labor share | BLS Productivity and Costs | Falls while productivity rises | Stabilizes alongside output |
| Consumer demand | Real discretionary spending | BEA | Spending falls with affected incomes | Spending holds |
| Private credit | Software-borrower defaults and recoveries | Fund and rating-agency reports | Losses linked to AI disruption | Defaults stay isolated |
| Housing | Prime-borrower delinquencies by metro | Official mortgage-performance data | Deterioration tied to job losses | Delinquencies stay near norms |
| Financial amplification | Lending standards, spreads, insurer exposure | Fed Senior Loan Officer Survey | Credit tightens alongside labor stress | Credit stays open |
These are monitoring categories, not findings that each signal is deteriorating. They reflect this article’s reading of the memo’s chain, not a list from Citrini. The same logic shows why white-collar jobs sit at the start of the chain: the spiral only works if lost income there stops circulating.
Three results together would weaken the scenario. Those are a sustained recovery in postings for exposed roles, a stable or rising labor share, and no material credit stress in software lending. A weak result on one signal alone would prove little either way.
Is the Citrini scenario a forecast?
No. It is a stress test, and the 2026 evidence supports only its earliest link: softer hiring in AI-exposed roles.
Readers who want to judge it should watch whether hiring weakness spreads into income, spending, and credit. Until several of those move together, the 2028 Global Intelligence Crisis remains a risk to monitor, not a result.
Frequently asked questions about the Citrini report
Q. Is the 2028 Global Intelligence Crisis a real prediction?
No. Citrini Research published it on February 22, 2026, as a fictional economic scenario exploring how AI-driven job losses could trigger a financial crisis by 2028.
Q. Who wrote the Citrini Research report?
James van Geelen of Citrini Research and Alap Shah collaborated on the project, with Shah contributing companion analyses.
Q. What is Ghost GDP?
Ghost GDP describes AI-generated economic output that increases measured production without delivering equivalent purchasing power to households.
Q. What is the intelligence displacement spiral?
It is a proposed cycle in which AI replaces workers, reduces consumer spending, and encourages businesses to invest in further automation.
Q. Why did the Citrini report affect the stock market?
The report amplified existing AI disruption fears during the February 23, 2026, sell-off. Tariff developments and software-sector concerns also contributed.
Q. What did Citadel Securities say about the report?
Citadel disputed the rapid-displacement scenario, citing stronger software hiring, gradual AI adoption, and infrastructure constraints.
Q. Does 2026 economic data support Citrini’s AI crisis scenario?
Not yet. September 2026 unemployment was 4.2%, although Dallas Fed research found weaker job postings in AI-exposed Texas occupations.
Q. Where can I read the original 2028 Global Intelligence Crisis report?
Read the free report on the official Citrini Research website
Related: How Swiss Data Protection Affects AI Companies vs EU GDPR
| Disclaimer: This article is for informational and educational purposes only and does not constitute financial or investment advice. The 2028 Global Intelligence Crisis is a hypothetical scenario, not a confirmed forecast. Economic data and market conditions may change, and interpretations of AI’s economic impact remain uncertain |
