The developed world is simultaneously suffocating from a shortage of workers and preparing to make millions of people economically unnecessary. Most disturbing of all, both catastrophes describe the same future.
Tim O'Harrow is 79 years old. For most of his life, he voted Republican and donated money to Republican candidates. His family has owned a farm in northeastern Wisconsin for four generations. Today, it has 1,600 cows. Immigrants from Mexico and Honduras milk them, assist with calving, perform artificial insemination, and manage the herd's computerized records. One of these workers arrived in America as a teenager, mastered virtually every aspect of the operation over twenty years, and is now teaching the trade to the owner's 19-year-old grandson. In March 2026, watching the immigration raids unleashed across the country by the Donald Trump administration, the elderly farmer uttered words almost unimaginable from someone with his political history: "I don't know if I'm a Republican anymore."
For twenty years, O'Harrow tried to hire local workers. He recruited college students and single mothers. He even tried hiring people facing criminal charges. Nearly all of them eventually called just minutes before the morning milking to say they weren't coming.
His farm is a remarkably accurate model of the developed economy in the second quarter of the twenty-first century. The jobs exist. The workers don't.
Meanwhile, in the same America, precisely the opposite story is unfolding. Amazon, the second-largest private employer in the United States, expects to avoid hiring more than 600,000 people it would otherwise need by 2033, according to internal documents leaked in the fall of 2025. Employers are increasingly attributing layoffs to artificial intelligence. According to consulting firm Challenger, Gray & Christmas, companies linked approximately 40% of the 97,000 job cuts announced in the United States in May 2026 directly to AI. An economy that cannot find someone to milk cows is simultaneously showing an accountant from the office next door the exit.
Conventional wisdom treats these as two separate problems, one demographic and the other technological. I am convinced they are the same problem viewed from opposite ends. The aging world is running short of workers today, and that is precisely why it will run short of customers tomorrow.
America's Milk Is Produced by Workers Washington Wants to Deport
Wisconsin is known as America's Dairyland, and the industry contributes more than $50 billion annually to the state's economy. According to estimates from the University of Wisconsin–Madison, immigrants, primarily from Mexico and Central America, perform approximately 70% of all labor on the state's dairy farms, and nearly all lack legal immigration status.
The reason is bureaucratically straightforward. The H-2A agricultural visa is available only for seasonal work, while cows must be milked two or three times a day, 365 days a year. Congress has never created a year-round agricultural work visa for livestock operations, despite two decades of requests from industry associations.
Farmers pay $14 to $20 an hour and cannot afford much more. Profit margins in the dairy business are thin, wholesale milk prices are determined by the market, and a herd owner has about as much control over those prices as over the weather.
For an American worker, the job is a bad bargain: getting up before dawn, no days off, summer heat, and Wisconsin's bitter winter cold. For someone from a Mexican village, those same wages can finance the construction of a house back home.
Economists at Texas A&M University calculated back in 2015 that immigrants accounted for more than half of America's dairy workforce and that losing them would drive retail milk prices up by approximately 90%.
American cinema explored this scenario in the 2004 satire A Day Without a Mexican. Washington, apparently, mistook the comedy for an instruction manual.
The obvious answer is robots.
A robotic milking system costs several hundred thousand dollars, plus another $5,000 to $10,000 annually in maintenance. One milking unit can serve roughly sixty cows, meaning O'Harrow's herd would require about twenty-five such machines. That translates into a multimillion-dollar loan for a business whose profit margins barely cover payroll.
Little wonder that, according to industry estimates, only about 10% of America's approximately 25,000 dairy farms use robots. In Europe, where farms tend to be smaller, the proportion is roughly one-third lower.
This is the first version of the future: a world in which human beings are still needed as workers because machines remain more expensive than people or cannot yet perform certain tasks.
Prosperity in such a world depends on a single condition: productivity must grow faster than the workforce shrinks.
Let's see whether that condition is being met.
The Canceled Miracle: Productivity Has Stopped Growing
One of America's most widely cited geopolitical forecasters has long reassured his audience that technological progress will cushion population decline. Productivity, he argues, will continue to increase, allowing aging societies to enjoy rising GDP per capita even without immigration.
The statistics do not support that optimism.
According to the U.S. Bureau of Labor Statistics, labor productivity in American manufacturing has been essentially stagnant since around 2012 and has actually declined in some segments since 2014.
Neither the tariff wars of Trump's first term nor the subsidies introduced under Joe Biden made a decisive difference. The CHIPS Act alone allocated approximately $52.7 billion to the semiconductor industry.
Almost all recent productivity gains in the American economy have come from the service sector.
Now that sector is also under threat. Hiring in highly skilled industries has fallen to recessionary levels. Employee turnover, which helped boost productivity during the pandemic, has nearly stopped. Investment in certain categories of automation equipment has declined.
The blow could arrive at precisely the moment companies begin deploying AI tools on a massive scale, the very technologies on which hopes for the next productivity breakthrough depend.
The global picture is no more encouraging.
According to one calculation, between 2000 and 2025, hourly output per worker among approximately twenty leading countries shifted from a range of $21.40–$85.60 to $22.00–$91.60. The authors interpret this as average annual growth of approximately 0.5%.
I am skeptical of that arithmetic, and in an unexpected direction: measured by the endpoints of those ranges, growth appears even weaker, around 0.1% to 0.3% annually.
Stagnation, not progress.
Ireland's leadership in such rankings has nothing to do with extraordinary engineering brilliance.
In 2015, Irish GDP surged 26% in a single year after multinational corporations transferred intellectual property rights to Dublin and booked the associated profits there.
Economists dubbed the phenomenon "leprechaun economics."
Ireland's productivity miracle exists largely on paper. No major country can reproduce that trick.
Depopulation adds another mechanism to this stagnation, one optimists prefer not to discuss.
When fewer people are available in the labor market, employers hire whoever they can find rather than whoever is best qualified. Competition for jobs weakens, incentives to acquire skills diminish, and output falls even when machinery remains unchanged.
A depopulating world risks declining productivity without any deterioration in technology.
Five Years to Master a Diesel Engine: Why America Can't Find a Mechanic for $120,000
In November 2025, Ford CEO Jim Farley revealed on a university podcast that the company's dealership network could not fill approximately 5,000 mechanic positions paying up to $120,000 annually, nearly twice the country's median salary.
According to Farley, learning to rebuild the diesel engine of a Super Duty pickup truck requires at least five years.
He broadened the point: "We have over a million openings in critical jobs: emergency services, trucking, factories, plumbers, electricians, and other skilled trades."
Ford and its dealers increased their scholarship program for aspiring automotive technicians from $1 million to $4 million.
Against a backdrop of one million unfilled positions, that is a gesture, not a policy.
Industry estimates suggest the United States could face a shortage of approximately 550,000 plumbers by 2027.
The electrician shortage was estimated at roughly 80,000 in 2025 and could reach 224,000 by 2030.
In the summer of 2023, Taiwan's TSMC postponed the opening of its first Arizona plant from 2024 to 2025, explicitly citing a shortage of qualified specialists capable of installing the equipment.
A company that ultimately pledged to invest $165 billion in the state ran into a shortage of people, not money.
Farley blames inadequate vocational training.
His diagnosis is only half right.
Training facilities are expanding slowly primarily because applicants are not lining up. Young Americans are reluctant to enter physically demanding trades even when substantial earnings are virtually guaranteed.
The U.S. labor force participation rate has fallen to approximately 62%, compared with 66% twenty years ago.
An important qualification is necessary here.
International comparisons placing the United States at 61.9%, below Japan's 63.6% and far behind Germany's 80.8% and Switzerland's 84.4%, often compare incompatible measures.
European figures are generally calculated for people aged 15 to 64, while the American figure covers everyone aged 16 and older, including millions of retirees.
Among Americans in their prime working years, ages 25 to 54, labor force participation is near historic highs.
This correction does not invalidate the concern. It sharpens it.
The overall figure is being dragged down by aging: baby boomers are retiring faster than younger workers are entering the workforce.
Hisayuki Idekoba, CEO of Japan's Recruit Holdings, which owns Indeed, the world's largest job-search platform, puts the issue more bluntly: society is afraid of the wrong thing.
According to Indeed's calculations, the American workforce will shrink by approximately 20 million people over the next fifteen years. Aging will account for roughly 80% of that decline, with artificial intelligence responsible for only 20%.
Japan has partially offset its labor shortage by bringing large numbers of women and retirees into the workforce.
The United States and Europe have almost exhausted that reserve.
The Japanese executive's complaint is striking precisely because it is so ordinary: "I don't want to wait two months to get a water leak fixed."
According to him, that is already the reality in America.
Why AI Will Reach Lawyers Before Plumbers
Back in the 1980s, robotics researchers identified a phenomenon that became known as Moravec's paradox: tasks that humans find intellectually difficult are often easy for machines, while tasks a child performs without thinking can be extraordinarily difficult for robots.
A computer defeated the world chess champion in 1997.
Yet a robot capable of reliably folding laundry or replacing a sink trap in an unfamiliar apartment remains largely a demonstration prototype.
This explains the defining asymmetry of our era.
Labor shortages are concentrated in occupations requiring dexterity, balance, and improvisation in messy, unpredictable physical environments.
Labor surpluses are emerging in occupations where work can be reduced to text, spreadsheets, and rules.
In 2002, an American urban theorist popularized the concept of the "creative class," grouping approximately 38 million Americans, roughly 30% of the workforce, under that label.
The category included scientists, engineers, architects, and designers, along with financiers, lawyers, physicians, and managers.
A quarter-century later, generative AI has forced a reassessment of this entire construct.
It turns out that most supposedly creative professionals spend much of their time performing routine intellectual tasks governed by rules they did not invent.
That is precisely why software agents can take over their jobs so readily.
By some estimates, only around 30 million people worldwide, approximately 1% of all workers, perform genuinely creative work that cannot be delegated to an algorithm.
Since January 2024, International Monetary Fund Managing Director Kristalina Georgieva has repeatedly cited the fund's estimate that AI will affect approximately 60% of jobs in advanced economies and 40% in emerging markets.
The poorest countries, with an exposure rate of 26%, are the most insulated from this wave.
Their protection comes from their own poverty: they simply have fewer office jobs that can be handed over to algorithms.
An Israeli historian and bestselling author of books about humanity's future once compared two centuries of warnings about technological unemployment to the fable of the boy who cried wolf.
He reminded audiences that, in the original story, the wolf eventually arrives.
The Industrial Revolution deprived humans of their monopoly on physical strength but left their cognitive abilities largely untouched.
Now machines are challenging that monopoly as well.
And they do not need to be smarter than humans.
They only need to perform particular tasks well enough.
Recent corporate practice reinforces the point.
Since late 2025, Amazon has eliminated approximately 30,000 corporate positions, roughly one-tenth of its office workforce, partly attributing those cuts to AI-driven efficiency.
CEO Andy Jassy had already warned publicly that AI-powered automation would reduce the number of corporate jobs.
China Turns Out the Lights on Factory Floors
In the early 1980s, Japan's FANUC opened a factory near the foot of Mount Fuji where robots assembled other robots without human involvement.
The concept of the "lights-out factory" was born: a production facility that requires no lighting because there is no one there who needs to see.
For forty years, such factories remained an engineering curiosity.
China is turning them into industrial policy.
Xiaomi invested approximately $330 million in its factory in Beijing's Changping district.
Company founder Lei Jun claims the facility can produce one smartphone per second around the clock without human intervention.
According to the International Federation of Robotics, 542,000 industrial robots were installed worldwide in 2024, twice as many as a decade earlier.
The global installed base reached 4.66 million units.
China accounted for 295,000 installations, representing 54% of the global market and almost nine times the number installed in the United States.
China's installed base surpassed two million robots, five times the American figure. For the first time, 57% of robots installed in China were manufactured domestically.
Popular articles about robotics still routinely cite a worldwide total of 2.25 million robots.
That figure is approximately half the actual number, and the error itself is revealing: people are debating the future using data from the past.
There is also exaggeration in the opposite direction.
A 2019 forecast by the British firm Oxford Economics projected 20 million robots worldwide by 2030, including 14 million in China.
At current industrial robot installation rates, those figures appear inflated.
Another estimate from the same report, however, remains compelling.
Since 2004, each additional industrial robot has displaced an average of 1.6 manufacturing workers, and robots could eliminate as many as 20 million factory jobs by 2030.
Economists at the Massachusetts Institute of Technology reached a comparable conclusion using American data in 2020: each additional robot per thousand workers reduced the employment-to-population ratio by 0.2 percentage points and wages by 0.42%.
Beijing's strategy is driven by demographics, not merely cost savings.
Approximately 120 million people work in Chinese factories.
China's population has been declining since 2022, while the generation born after 2000 is increasingly rejecting the "996" work culture: 9 a.m. to 9 p.m., six days a week.
For factory owners, robots are no longer simply a way to save money.
They have become a condition for survival.
The most revealing detail of China's transition is the emergence of a new occupation: robot trainer.
Young people from rural areas and vocational colleges, often without university degrees, earn between 6,000 and 10,000 yuan a month.
That is roughly comparable to delivery couriers' pay, but they work in clean offices, and the occupation is considered prestigious.
During an eight-hour shift, they perform as many as fifteen different tasks, repeating each movement ten times so the machine can learn it.
The authorities have officially recognized teleoperation as a new vocational training pathway for young people without advanced qualifications.
An entire social class is diligently teaching machines that will eventually take its place.
The 600,000 People Amazon Will Never Hire
Amazon's leaked documents require careful interpretation. Popular accounts frequently distort their meaning.
The company is not planning to "replace" 600,000 existing employees.
Its robotics division expects to double sales by 2033 while keeping its U.S. workforce roughly at its current level.
In other words, the goal is to avoid hiring more than 600,000 additional workers who would otherwise be needed, including 160,000 by 2027.
The team's objective is to automate 75% of operations, saving an estimated 30 cents per item processed and $12.6 billion between 2025 and 2027.
Amazon's warehouse in Shreveport, Louisiana, operates with approximately 1,000 robots and roughly 25% fewer employees than it would require without automation.
Amazon has stated that the documents reflect the perspective of one team rather than the company's overall strategy.
Politically, the distinction between "firing" and "not hiring" is fundamental.
A laid-off worker joins a picket line.
A job that never existed has no one to mourn it.
This new form of unemployment is invisible. It has no identifiable victims.
Tellingly, the same documents reportedly advised employees to avoid words such as "automation" and "AI," replacing them with "advanced technology" and "cobots," or collaborative robots.
Amazon denies this.
Nevertheless, the recommendation itself reveals a greater awareness of the risks than any corporate press release.
The economics are unforgiving.
Morgan Stanley analysts estimate that a humanoid robot with operating costs of approximately $5 an hour could replace two employees earning $25 an hour each and generate approximately $200,000 in lifetime economic benefits.
I place this estimate firmly in the category of investment enthusiasm. Humanoid robots have yet to achieve mass deployment on assembly lines.
But the underlying economic logic is sound.
Average hourly earnings in the U.S. private nonfarm sector reached $37.41 in March 2026.
The more expensive human labor becomes, the faster a machine pays for itself.
Who Will Buy What the Robots Produce?
Here begins the second version of the future, a world in which machines do all the work.
And this is where the central trap lies.
A robot does not eat, sleep, take vacations, or demand that the factory be heated.
It also buys nothing.
An economy resembles a balance scale.
As the productive capacity of lights-out factories grows heavier on one side, the income of the workers displaced by those factories diminishes on the other.
Warehouses fill with goods manufactured at unprecedented efficiency, while purchasing power evaporates along with wages.
Polish science fiction writer Stanisław Lem described this mechanism in the 1950s, in Ijon Tichy's twenty-fourth voyage from The Star Diaries.
On the planet of the Indiots, New Machines are invented that can operate without supervision.
Workers lose their livelihoods and starve to death while remaining perfectly free.
Mountains of unsold goods spill over factory fences, and the starving population is eventually subdued by force.
When the protagonist asks where the factories' income went, a local calmly explains that it flowed to the worthy people who legally owned them.
My generation read this parable in our youth as a satire of capitalism, which was precisely why Soviet censorship allowed it to circulate.
Rereading it in 2026 is an unsettling experience.
We remember something else as well.
In the mid-1980s, the Soviet press celebrated "unmanned technologies" and flexible automated manufacturing systems that were supposed to accelerate economic development.
Nobody asked what would happen to the workers.
In a planned economy, unemployment did not exist by definition.
A market economy enjoys no such luxury.
American labor-union folklore preserves a story, most likely apocryphal, about a Ford executive showing United Auto Workers leader Walter Reuther an automated production line in the 1950s.
The executive supposedly asked with a smirk how Reuther intended to make the machines pay union dues.
Reuther allegedly answered with a question of his own: How are you going to make them buy Fords?
There is no documentary evidence that this exchange actually occurred.
But it captures the economic problem more precisely than many formal reports.
The Luddites who destroyed textile machinery in England between 1811 and 1816 lost their battle.
In 1812, Parliament made machine-breaking a capital offense, and the Industrial Revolution ultimately created more jobs than it destroyed.
Optimists routinely cite this precedent.
Nobel laureate Wassily Leontief offered a different historical analogy in 1983: the horse.
The internal combustion engine did not retrain horses for new occupations.
The horse population simply collapsed.
In 1930, John Maynard Keynes predicted that his grandchildren would enjoy a fifteen-hour workweek by 2030.
Four years remain before that deadline.
No mechanism currently exists on a mass scale that would allow societies to produce goods without human labor while simultaneously sustaining the monetary demand needed to purchase them.
If displaced workers receive benefits, the money must come from somewhere, potentially through taxes on automation.
Set those taxes too high, and the economic advantages of robots disappear.
Set them too low, and the benefits will not support the people who have lost their jobs.
Production then runs into the wall of shrinking demand.
In February 2017, Bill Gates proposed taxing robots in much the same way governments tax the income of the workers they replace.
That same year, South Korea reduced tax incentives for investments in automation, a move described as the world's first indirect robot tax.
In 2017 and 2018, Finland paid 2,000 unemployed people €560 a month with no conditions attached.
Participants reported improved well-being, but employment changed little, and the experiment was not extended.
One of the founders of America's best-known technology magazines, a futurist with four decades of experience, paints an appealing picture.
Humanity, having largely stopped reproducing because raising children is inconvenient, invites trillions of synthetic agents into the world to serve the relatively few people who remain.
Human beings, meanwhile, devote themselves to art, discovery, and leisurely meals with friends.
He candidly admits that he has no idea what such an economy would look like and mentions universal basic income as one possible innovation.
That honesty is to his credit.
It is not a foundation for public policy.
The Paris Billionaire's Mistake
On June 17, 2026, at the VivaTech conference in Paris, Jeff Bezos offered a prediction that immediately spread around the world:
"I strongly disagree with that point of view. I think AI will actually create a labor shortage."
The Amazon founder argued that people have an essentially infinite number of things they would like to accomplish but cannot because of existing constraints.
AI, he suggested, will remove those constraints.
From the same stage, he discussed Prometheus, an AI startup he co-founded in November 2025 to accelerate physical manufacturing.
Bezos is right about one thing, and I concede it without hesitation.
Over the next ten to twenty years, the working-age population of developed countries will physically shrink.
People, rather than capital, will become the binding constraint on business.
Idekoba is making essentially the same argument, only without the cosmic optimism.
Bezos's mistake lies elsewhere.
He is thinking in terms of the mass-consumption era: production expands, economies of scale lower prices, and buyers inevitably emerge.
That formula worked while populations were growing.
In an aging world, it begins to break down.
The number of working people, and therefore wage earners, declines, while the number of older people, who generally consume less, increases.
Bezos did not explain what demand businesses would be scaling up to meet.
Nor should we overlook the speaker's own interests.
A man whose company reportedly plans to avoid hiring 600,000 workers and has already eliminated 30,000 office positions is assuring the world that workers will be in short supply.
The most uncomfortable possibility is that both statements may prove true simultaneously.
There will be too few people for the jobs machines cannot yet perform.
And there will be too many people for the jobs machines have already taken.
Twenty-Nine Robots Instead of One Hundred Forty-Five
For readers in the former Soviet Union, this debate is closer to home than it might appear.
Russia is experiencing the first version of the future in almost chemically pure form.
Unemployment is hovering around 2%, a record low.
The economy is constrained by labor shortages.
For decades, much of the country's physically demanding work was performed by migrants from Central Asia, whose lives became substantially more difficult after the March 2024 terrorist attack at Crocus City Hall.
Vladimir Putin has set a goal of placing Russia among the world's top twenty-five countries in robot density by 2030.
Meeting that target would require approximately 145 industrial robots per 10,000 manufacturing workers.
Estimates of the current level vary, and that discrepancy is revealing in itself.
In 2026, Minister of Industry and Trade Anton Alikhanov cited a figure of 29 robots per 10,000 workers and acknowledged that it had not changed during 2025.
Consulting firm Kept, working with the Industrial Robotics Center at Innopolis University, calculated a density of 40 robots and reported annual growth of 36%.
Even that more optimistic estimate is 3.6 times below the target and less than one-quarter of the global average of 177 robots.
Russia risks becoming trapped between the two versions of the future.
It already lacks enough people for the first.
It does not yet have enough machines for the second.
The Farmer's Grandson and the Quarter-Million-Dollar Machine
The two versions of the future, usually presented as opposites, are actually successive stages of the same process.
First, an aging economy runs short of people who are both willing and able to work, and productivity growth stalls.
Then comprehensive automation becomes the only apparent solution, and the economy encounters a different shortage: people who can afford to pay for what is produced.
Technology solves the first problem and creates the second.
Solving the second requires institutions capable of redistributing income, and nobody has yet invented an adequate system.
I am prepared to make two testable predictions.
By the end of 2028, the U.S. Congress will not have created a year-round work visa for livestock farming. America's dairy industry will continue to depend on undocumented workers until machines begin displacing them.
By 2030, Russia will not have reached even half its target of 145 robots per 10,000 workers.
I leave the central question unanswered because neither economists nor governments currently have an honest answer.
Who will support the people rendered economically unnecessary, where will the money come from, and will those people be able to buy what the robots produce?
Back on O'Harrow's farm, the owner's 19-year-old grandson is learning from a man who arrived from Mexico as a teenager and has spent twenty years carrying much of the operation on his shoulders.
A decade from now, that grandson may well be managing robotic milking stations from a tablet.
The machines will milk the cows without weekends off and without visas.
But they will never buy the milk.