More productive, not richer: the math of skill vs ownership

For the first time in data going back to 1947, US employees got less than half of gross domestic income, and AI lands on top of that split.
In the second quarter of 2026, employees in the United States received 49.99% of the country's gross domestic income. It is the first quarter below half in official data going back to 1947.
In the same quarter, output per hour was 2.2% higher than a year earlier. Real hourly compensation, adjusted for inflation, was 0.1% lower. One in five American businesses already uses AI.
The economy produces more per hour. The people who work those hours get a smaller slice.
Let me be blunt. In 18 years in business, as a founder and as an investor, I haven't seen a more expensive belief than this one: if I get more productive, I'll get richer. Productivity is not a paycheck. Ownership is.
I went through the official US data on income, productivity, pay and wealth. Here is what it shows.
A 25-year story, not an 80-year one
You often hear that labor has been losing to capital ever since the war. The data says otherwise.
From 1947 to 1970, employees' share of US income rose from 53.6% to 58.4%. Over the whole half-century from 1947 to 2000, the official labor share index for the US nonfarm business sector slipped just 3%.
Then it broke. From 2000 to mid-2026 the same index fell 16%, and its last four readings are the four lowest on record. In 2025 employees received 50.9% of gross domestic income, the lowest annual share since 1935.
It isn't only an American story. Official international estimates put labor's share of world GDP at 52.7% in 2025, down from 54.0% in 2004.
Where the money went
From 2000 to 2025, employees' share of US income fell 5.7 percentage points. Corporate profits after tax gained 5.0 points, from 4.3% to 9.3%. In the second quarter of 2026 they reached 10.2%, the highest in data going back to 1947.
Let's put a dollar figure on it. From 1947 to 2000, employees received 55.9% of gross domestic income on average. At that share they would have earned $1.56 trillion more in 2025. That is $10,700 for every full-time-equivalent job in the country, a 10% raise on average compensation of $108,000.
That raise never arrived. Not because people worked less. Because the extra output went to the other side of the income statement.
Workers did get richer, just much more slowly
To be fair to the data, real pay grew. From 2000 to 2025, output per hour in the US nonfarm business sector rose 63%. Real hourly compensation rose 25%.
It wasn't always like this. From 1947 to 1973 the two moved almost together: productivity up 104%, real hourly compensation up 93%.
The gap that opened after that has two layers. From 1973 to 2000 it was almost entirely about prices: what workers buy got more expensive faster than what their employers sell, while labor's share of the pie barely moved. After 2000 a second layer appeared, the split itself. Since 2000, prices explain 46% of the gap and the smaller share for labor explains 54%. The first layer is inflation. The second is a question of who owns what.
AI is landing on top of this split
Productivity has picked up. Over 2023-2025, output per hour grew 2.4% a year, against 1.5% a year from 2005 to 2019. Real hourly compensation grew 1.4% a year. In the first half of 2026 productivity was up 2.2-2.9% from a year earlier, while real pay stood still. In the third quarter, the median full-time worker's weekly earnings, adjusted for inflation, were 2.1% lower than a year before.
And adoption has barely started. Official survey data counted 19.8% of US businesses using AI in May 2026, and 39.7% in the information sector. Central bank research puts work use of generative AI at 41% of the workforce in November 2025.
Here is the mechanism as I see it from the investor's side of the table, where Fanaura backs AI-native founders. A salary is a price, and prices follow supply. When a model can do part of your task list, the supply of what you sell goes up: more people can do your job with the same model, and some of the work needs no person at all. The gain lands with whoever owns the model, the product built on it and the relationship with the client who pays. AI doesn't invent this split. It arrives in an economy where the split is already the most lopsided on record.
Who owns the other half
Official data show that in mid-2026 the top 10% of American households owned 88.1% of all stocks and fund shares held outside pension plans. The top 1% alone owned 50.9%. The bottom half owned 0.6%. In 2000 the top 10% held 81.9%.
Private businesses look the same: 84.8% belongs to the top 10%.
So when the profit side of the economy grows, it grows for a small group of owners. Most people sit on the side that keeps shrinking.
Owning your own job is not ownership
One more line in the national accounts deserves attention. The income of sole proprietors and partnerships, the self-employed and small firms, fell from 7.3% of gross domestic income in 2000 to 6.5% in 2025. In 1947 it was 14.0%.
Being your own boss didn't protect anyone. A freelancer who bills by the hour still sells hours, just without a payroll department. The share that grew went to the owners of corporations.
What I'd do with this
Measure your personal labor share. Split last year's income into what you earned for your time and what you earned from things you own. For a salaried professional without equity, it's 100% time. That is full exposure to the line on the chart that keeps falling.
Get paid in what is gaining share. Equity, options, profit share, revenue share. If AI doubles revenue per employee at your company and your pay is a pure salary, the gain goes to the owners by default. Negotiate for a piece of the result, not just a price for your hours.
Turn part of every paycheck into ownership. The bottom half of American households owns 0.6% of the stocks and fund shares that households hold outside pension plans. A fixed share of every paycheck that buys a stake in the profit side is the simplest way to stand on both sides of the chart.
Use AI to build an asset, not just to finish the same job faster. Time saved on the same job is a gift to your employer. Time saved and put into a product, a client base or a body of work you own compounds for you.
Skills still matter. They get you a seat at the table. Ownership decides how much of what the table produces you take home, and for the first time in the data, the people doing the work take home less than half.
What share of your income last year came from something you own, rather than hours you sold?
Building an AI-native company? Apply in five minutes. →
More from Fanaura
Building an AI-native company that sells results instead of hours? Send us the project.
Fanaura invests its own capital and the capital of its limited partners (LPs), usually through a separate SPV for each deal. If we see a fit, we get in touch.
Get the news by email
One short note each weekday. No marketing, and you can unsubscribe at any time.