Stocks for Thoughts · The Wage Bill

Six rungs, one price.

Every rung is a different answer to "what is AI actually selling?" Each one is turned into a company value the same way: annual revenue, a 30% net margin, a 20× multiple. The market has already climbed to rung four — three times what the historical rate of return on invention would support, and two-fifths of what Nvidia's chief executive says out loud.

observed — money changing hands, or the historical record hypothetical today's price
What AI is selling Implied company value  → Value
1What's actually collected
Every dollar an end customer paid for AI last year, added up — and it double-counts.
happening nowrev ~$0.11T · 0.2% of world wages
$0.7T
2The historical rate
Inventors have kept about 2.2% of the value they created. Apply that.
1948–2001 recordrev ~$1.4T · 2.2% of world wages
$8.2T
3All the world's software
AI wins the entire global software market. Every seat, everywhere.
rev ~$1.5T · 2.4% of world wages
$8.8T
4Today's price
A slice of payroll — the whole world's, every year, forever.
you are hererev ~$4.2T · 6.7% of world wages
$25T
5Jensen's number
Huang's own arithmetic: $10T of AI spend against $50T of thinking.
never happenedrev ~$10T · 16% of world wages
$60T
6Every American job
AI does all of it and bills the entire US wage bill. Employers keep nothing. The US is a quarter of world payroll, so this is a ceiling on the American half of the bet, not the global one.
the US ceilingrev ~$16.2T · 26% of world wages
$97T
today · ~$25T
Company value →
$0.5T $1T $5T $10T $50T $100T
Read it as a ladder, not a forecast. The three rungs below today's price are the ones with something behind them — money collected, the historical record, the whole software market — and not one of them clears $9 trillion. Rung four is where the price already is, and it is a different kind of claim: not on a market, but on payroll. Above it sit the two rungs the industry talks about and nobody has ever reached. The whole argument between bulls and bears is a disagreement about which rung is the honest one — and everyone is quoting a different denominator.
Method: each rung's annual revenue × a 30% net margin × a 20× earnings multiple = the company value it supports. Horizontal axis is that value on a log scale, $0.5T→$100T. "World wages" = the ILO's 52.7% global labour income share applied to World Bank world GDP of ~$118T ≈ $62T; this is a derived figure, not an ILO publication. Sources: Nordhaus, Schumpeterian Profits in the American Economy (NBER 10433) for the 2.2%; Gartner (Jul 2026) for the $1.47T world software market; Jensen Huang on the BG2 podcast, Sep 2025, for the $50T/$10T; BEA for the $16.2T US compensation of employees; S&P DJI (30 Nov 2022) and Slickcharts (6 Aug 2026) for index capitalisation, with Morgan Stanley Wealth Management's Oct-2025 decomposition attributing ~75% of the gain to AI names — the ~$25T is an estimate with a wide error bar. Margin and multiple are illustrative and applied uniformly so the rungs are comparable — change either and every rung moves together. Not investment advice.