The Flying Pigs 360
Horses that can drive cars will replace horses that can't. Pigs will fly, in a full circle, and Uncle Sam's flies highest. And the six-figure safety net is already thin. A tour of the AI story, and of the last time machines rewired society.
In 1915, America had about 26 million horses and mules. They pulled the plows, the wagons and the streetcars. They were the economy’s engine.
Picture a consultant standing in a stable that year, giving the horses a pep talk:
“Cars won’t replace you. Horses that can drive cars will replace horses that can’t. So learn to drive!”
By 1960 there were about 3 million left. Not one of them had learned to drive.
Learn to drive, Dobbin
Today’s version of the pep talk is everywhere: “AI won’t replace you. A person using AI will.”
It sounds like advice. It is really a description of a short stage in between. A car needed a driver, but it never needed a horse. An AI tool needs a prompter today. The whole point of the industry’s spending is to need one less tomorrow.
Nearly 90% gone in 45 years. Skills didn't save them. Ownership might have.
The horse didn’t lose to a better horse. It lost to something that didn’t need one.
The last time machines rewired society
This has happened before, and it was not gentle.
The industrial revolution made the world rich. It also pulled millions of people off farms and into factory towns, where a job could vanish with the next new machine. The shoe factories and clothing factories that built the boom employed whole towns, until they didn’t. Within a few decades the world went through two world wars and a Great Depression.
After the 1929 crash, the jobs disappeared and almost nothing caught the people who lost them:
One worker in four, or worse. Most had no unemployment insurance and no savings to fall back on.
What came next was the cushion. In 1935 the US created unemployment insurance and Social Security, so that the next shock would land on something softer than the street.
The technology was never the villain. The speed of the change, and the lack of a safety net, were.
The machines were never the problem. The missing cushion was.
Even the $500K crowd has no cushion
The comforting story says the people at risk are the low-skilled. The skilled, well-paid ones will ride the AI wave.
Goldman Sachs asked more than 5,000 Americans aged 45 to 75 whether they were “primarily living paycheck to paycheck and struggling to make progress on long-term goals.” Here is the share who said yes:
It falls as income rises, then climbs again above $300K. Big mortgages, private schools and lifestyle creep eat the difference.
More than a third of households earning over half a million dollars say they live paycheck to paycheck. These are the lawyers, coders, bankers and managers that “learn AI” is supposed to protect. And these are exactly the jobs AI is aimed at first.
A horse that earns a lot and saves nothing is still a horse.
High income is not a cushion. Savings are.
Flying pigs, full circle
So who is paying for all of this? Here the story gets stranger.
We call it The Flying Pigs 360. The pigs are the promises: revenue that arrives when pigs fly. The 360 is the path the money takes, in a full circle.
It works like this. A big tech company invests in an AI lab. The lab spends that money renting computing power from the same company. The company books it as sales. The pig flies, and lands right back where it took off.
Michael Burry, the investor who called the 2008 housing crash, has been charting this loop for a year:
- November 2025: on a chart of the AI deal web, he wrote that “true end demand is ridiculously small.”
- The same month: he estimated that by stretching the assumed life of their chips, the largest cloud companies would understate costs by about $176B from 2026 to 2028.
- August 2026: with Bloomberg counting about $46B of equity stakes and $879B of multi-year purchase promises between the players, his verdict was “the money is circling, not multiplying.”
- September 28, 2026: he moved his bets against Nvidia, Micron, Nebius, Palantir and a chip ETF into put options with earlier dates. “The bubble in AI may burst sooner than later.”
Be fair to the pigs. Burry’s point is not that these companies have no revenue. Nvidia sells real chips, and the big AI labs report real sales. His point is that too little of it comes from customers spending their own money. He has also been early before, and early looks a lot like wrong for a while.
When the customer is funded by the seller, the sale is a promise, not a fact.
Uncle Sam, the biggest pig of all
The flying pig with the widest wings belongs to Uncle Sam.
It took until 1996 to borrow the first $5 trillion. The last $20 trillion took ten years:
Doubled in a decade. Treasury counts, each October.
And this pig flies in a circle too. The Congressional Budget Office expects the government to borrow about $1.8 trillion this year, and to pay about $1 trillion in interest. In effect, more than half of the new borrowing goes to pay interest on the old. By 2036 the yearly interest bill reaches $2.1 trillion.
That matters for the AI story. The cushion that caught people after the 1930s, unemployment insurance and Social Security, is paid for by a borrower already this stretched. If machines shrink the number of paychecks, tax receipts fall just as the calls on the cushion rise.
The cushion that caught the last disruption now runs on credit.
Optimus: the final nail
There was always a fallback in the comforting story. If AI takes the desk jobs, people can still do the work AI can’t: lifting, carrying, fixing, building.
Tesla is now turning its old Model S and X lines in Fremont into a factory for Optimus humanoid robots. According to The Information, Tesla is aiming to build 1,000 a week by the end of 2026, ten times its pace in the second quarter. Elon Musk has said a robot could one day cost $20,000–30,000, and that customers may be able to buy one by late 2027.
Do the arithmetic on a robot that works around the clock:
At Musk’s price ($30K)
At analysts’ high end ($100K)
Human: the US federal minimum wage. Robot: its price spread over five years of round-the-clock work, before power and repairs.
Even at more than three times Musk’s price, the robot undercuts the minimum wage by more than two-thirds. AI comes for the thinking. Optimus comes for the lifting. That leaves the comforting story with nowhere to send people.
AI takes the desk. Optimus takes the dock. The pep talk runs out of rooms.
What we are watching
- Cloud giants cutting their spending plans.
- AI labs raising new money mainly to pay their computing bills.
- More companies shortening the assumed life of their chips.
- Robot output hitting its targets while hiring falls.
- AI sales growing from customers who pay with their own money.
- Older chips still earning their keep.
- New kinds of work growing as fast as old kinds disappear.
The Kuberin view
In 1915, the smart move was never to teach a horse to drive. The money went to whoever owned the cars, the oil and the roads.
That is the lesson for the AI age, for workers and investors alike:
- Be an owner, not a horse. Owning a share of great businesses beats being the input they are trying to replace.
- Own the real thing, not the flying pig. Look for companies whose customers pay with their own money, not ones paid by their own suppliers.
- Build the cushion first. A high salary is a flow. A cushion is a stock. When the machines arrive, only one of them helps, and the public cushion is already borrowed.
History says these transitions end well for society in the long run, and badly for many people along the way. The difference between the two is usually a cushion.
When pigs fly in circles, don’t stand underneath.
Calculation note. Robot hourly cost = price ÷ (5 years × 8,760 hours a year): $30,000 ÷ 43,800 ≈ $0.68, and $100,000 ÷ 43,800 ≈ $2.28. It ignores electricity, repairs, software and downtime, and assumes a robot can work every hour of the year, which none can yet. The $7.25 is the US federal minimum wage; most workers earn more, which only widens the gap. Debt per day = ($40.34T on October 13, 2026 − $37.92T on October 14, 2025) ÷ 364 days ≈ $6.7B. The interest-to-borrowing comparison uses CBO’s February 2026 projections for fiscal 2026 (about $1.0T of net interest against a $1.8T deficit); money is fungible, so “goes to pay interest” is a share, not a ledger entry. Horse figures are US Department of Agriculture counts of horses and mules, rounded. Survey figures are from the Goldman Sachs Asset Management 2026 Retirement Survey of 5,106 people aged 45 to 75, and reflect how respondents described themselves.
Disclaimer. Educational content only. Not investment, legal, tax, or financial advice. The analysis provided is for illustrative purposes and does not constitute a recommendation to buy or sell any specific security. Views attributed to Michael Burry are as reported in the sources below; Kuberin is not affiliated with him or his funds. All investments involve risk, including possible loss of principal.
Sources. US horse population (Wikipedia) · Horse and mule statistics (Yale Energy History) · Great Depression unemployment (St. Louis Fed) · Great Depression in Germany (Wikipedia) · Social Security Act of 1935 (SSA) · High earners living paycheck to paycheck (Fortune) · Burry on the AI deal web (X) · Burry on depreciation (CNBC) · AI circular deals (Bloomberg) · Burry on circular financing (24/7 Wall St.) · Burry moves up his timeline (CNBC) · Burry’s new put positions (Motley Fool) · Debt to the Penny (US Treasury) · Budget and Economic Outlook 2026–2036 (CBO) · CBO outlook summary (CRFB) · Optimus production ramp (Yahoo Tech, citing The Information) · Fremont line conversion (Electrek)