Amazon posted record revenue and cut thirty thousand workers. Block's gross profit grew twenty-four percent and the company eliminated forty percent of its workforce. The economy lost ninety-two thousand jobs while corporate earnings hit record highs. Each fact has been documented. Together they describe a pattern with no modern precedent: growth without employment. The question is whether the mechanism that distributed the gains from every prior technology revolution still works.
Amazon posted record revenue of $716.9 billion in 2025 and cut thirty thousand employees — the largest reduction in its thirty-one-year history. Block's gross profit grew twenty-four percent year-over-year and the company eliminated forty percent of its workforce. The S&P 500 posted record fourth-quarter earnings. The American economy lost ninety-two thousand jobs in February — the first negative payroll print since the pandemic.
Each of these facts has been documented individually in this journal. Together, they describe a pattern that has no modern precedent: growth without employment.
The Mechanism That Worked
For a century and a half, every major technology revolution followed the same script. New technology arrives. Productivity surges. Some workers are displaced. But the gains distribute: cheaper goods expand markets, expanded markets require more workers, more workers earn wages, wages drive consumption, consumption drives investment, investment creates more jobs. The cycle ran through electrification, the assembly line, computing, and the internet. Each displaced millions. Each, eventually, employed more than it displaced.
Economists call this the productivity-wage nexus. It is the load-bearing assumption underneath growth economics, monetary policy, and the political consensus that technology benefits everyone in the long run. The Federal Reserve sets interest rates based on models that assume productivity gains eventually become wage gains. Government labor policy assumes retraining works because new industries will hire the retrained. Corporate strategy assumes workforce investment pays returns.
Every one of these assumptions depends on the mechanism working.
The Evidence It's Under Stress
This journal has documented the individual signals for weeks. Assembled, they form a pattern.
The Reallocation showed companies cutting workers to fund AI infrastructure — not because AI had replaced the work, but because two and a half trillion dollars in global AI spending had to come from somewhere. Amazon's two hundred billion in 2026 capex is being financed, in part, by the payroll of the workers it eliminated.
The Apprentice found that in computer systems design, employment fell five percent since ChatGPT launched while wages rose seventeen percent. The Dallas Fed identified the mechanism: AI substitutes for entry-level workers and augments experienced ones. Fewer people, doing more, earning more per person. Aggregate output grows. Aggregate employment does not.
The Paycheck documented February's signal: ninety-two thousand jobs lost, average hourly earnings up four-tenths of a percent — the hottest wage print in months. Fewer workers. Higher pay. Same economy.
The List counted twenty-two thousand AI-cited layoffs through February, from thirty-five CEOs who named artificial intelligence as the rationale. Sixty percent of surveyed executives had made headcount reductions in anticipation. Two percent based those reductions on actual AI implementation.
The Endorsement showed what the market thinks. Block cut nearly half its workforce and the stock surged twenty-four percent. The market did not merely tolerate AI replacing workers. It demanded it.
The Boomerang warned that half of AI layoffs will be quietly reversed — rehired offshore, at lower salaries, after the stock already received its premium.
The mechanism of the cuts is messy. Some are real displacement. Some are capex funding. Some are anticipation theater. Some will be reversed. That decomposition matters for policy. It does not matter for the question this entry asks.
Why the Decomposition Is Beside the Point
Every prior entry in this series has tried to decompose the signal — to identify which mechanism is driving the layoffs, because the mechanism determines the correct response. If AI is replacing workers, retraining is the answer. If capex reallocation is the cause, the jobs return when the investment cycle matures. If anticipation is running ahead of capability, the gap closes on its own.
The decomposition is analytically sound and macroeconomically irrelevant.
The traditional distribution mechanism does not care why employment is falling. It only works when employment rises alongside output. The productivity gains from electrification distributed because factories hired more workers to produce more goods. The gains from computing distributed because the software industry created millions of new positions. The gains from the internet distributed because e-commerce, social media, and digital advertising created entire new labor markets.
Each revolution created demand for human cognitive or physical labor that had not existed before. And in each case, the created demand exceeded the displaced demand. That is the mechanism.
AI is the first general-purpose technology that automates the cognitive work that prior revolutions created. The internet created content marketing. AI writes the content. Computing created software engineering and its apprenticeship pipeline. AI writes the code. The digital economy created data analysis as a mass profession. AI analyzes the data. Customer service scaled with e-commerce. AI handles the tickets.
When the technology displaces the same kind of labor that the previous technology created, the new-job-creation half of the cycle may not fire. Not because AI is more powerful than prior revolutions — but because it operates on the same axis.
Two Scenarios in the Same Data
The optimistic reading is a J-curve. The AI infrastructure buildout — two and a half trillion dollars in global spending in 2026 — is a front-loaded cost that has not yet generated its returns. Companies cut to fund the build. When the build matures, it produces returns, those returns expand markets, and expanded markets require people to operate, govern, and build on the new capabilities. The displacement is cyclical. Employment recovers. It always has.
The pessimistic reading is a phase transition. AI improves at the same tasks it displaces workers from. Each generation of models is better at writing code, better at customer service, better at data analysis — the exact functions where layoffs are concentrated. The J-curve never completes because the technology keeps advancing along the same axis as the jobs it eliminates. Growth continues. Employment does not recover. The distribution mechanism that connected productivity to wages for a hundred and fifty years stops working.
The data cannot yet distinguish between these scenarios. But a single metric can.
The Metric
Unemployment is a lagging indicator that conceals the structural question. A person who stops looking for work is not counted as unemployed — they vanish from the denominator. The unemployment rate can fall while the economy loses workers permanently.
Labor force participation rate reveals what unemployment hides. It measures the fraction of working-age adults who are either employed or actively seeking employment. When participation holds steady while GDP grows, the economy is creating enough opportunity to keep people engaged. When participation declines while GDP grows, the economy is producing output without producing jobs. That is the definition of decoupling.
U.S. labor force participation peaked at 67.3 percent in January 2000. It never recovered to that level. It sits near 62.4 percent today. Twenty-five years of technological progress, GDP growth, and population expansion have not returned the participation rate to its pre-internet peak.
Each prior revolution eventually pulled participation up. Electrification brought women into factory work. The service economy of the computing era expanded who could participate. The question the March 2026 data forces is whether AI will be the first technology that pushes participation permanently lower — not through acute displacement, but through chronic absorption of the cognitive work that kept people in the labor force.
What Follows
If the mechanism holds — if this is a J-curve — then the response is patience. Distribution will happen. Standard macro models remain valid. The Fed can set rates with confidence that productivity gains become wage gains with the usual lag.
If the mechanism is broken — if this is decoupling — then distribution becomes a design choice rather than an economic inevitability. Productivity gains that do not automatically flow to workers must be deliberately redirected through policy, taxation, or institutional restructuring. This is not a novel idea. It is merely an idea that has never been necessary at scale, because the mechanism always worked.
This journal has spent fourteen entries documenting the individual signals — the capex reallocation, the experience divide, the negative payrolls, the anticipation gap, the market's revealed preference for headcount reduction, the quiet rehiring. This entry names the pattern they form: output is growing, employment is stagnating, and the channel that traditionally connected one to the other is showing the first structural strain in a century and a half.
Whether the channel holds is the question that every other signal in this journal ultimately depends on. If growth distributes, the economy navigates the transition. If growth decouples, the economy enters territory that no living policymaker has mapped.
The data is early. The signal is clear. The mechanism is under test.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)