We lost 23,000 jobs in July.
Unemployment ticked up to 4.2% and average hourly earnings went up 0.2%, costing employer’s about $1.5Bn per month (158.8M workers) in additional wages but that’s a number that accumulates over the year and we’re up 10% (now $7.5Bn/qtr) in the past 5 months. Keep that in mind when you are being told that the cost of AI labor is more than human labor – the AIs don’t get raises!
| Measure | Level | What it is |
|---|---|---|
| Total US population | ~340M | Everyone |
| Civilian noninstitutional population (16+) | 275.28M | Everyone 16+ not in prison, nursing home, or on active military duty bls |
| Civilian labor force | 169.09M | Working OR actively looking bls |
| Employed (household survey) | 162.18M | Says they have a job in the household survey fred.stlouisfed |
| Total nonfarm payrolls (establishment survey) | 158.86M | Jobs on employer payrolls fred.stlouisfed |
| Private nonfarm payrolls | 135.59M | Private-sector jobs only (JEC report) |
| Government payrolls | 23.27M | Federal + state + local (JEC) |
| Unemployed | 6.9M | In labor force, don’t have a job, looking |
| Not in labor force | 106.19M | Not working, not looking (275.28M – 169.09M) |
Still, the fact that Wages are still rising and Unemployment is still relatively low indicates the Robots have not yet taken our jobs. Trump may not have created any jobs in the last two years but he’s certainly created job opening – kicking 605,000 human beings out of the country and forcing another 1.9M to “voluntarily self-deport” (picture mothers carrying children back over the border with ICE attack-dogs hot on their heels as “voluntary“).
In fact it was the loss of 53,000 Government Jobs (who needs the CDC?) that put us negative in July. In one hour, having not seen dramatic Government layoffs and with an uptick in Travel over the summer – we expect maybe +50,000 jobs and that will run the odds of a Fed rate hike back from 30% to 50% or more and may tank the market again.
The Dollar is right on the 99 line and more workers with rising wages means more demand for Dollars and that too can put pressure on stocks and commodities but there’s no sense in speculating – we’ll see in an hour…
Meanwhile, let’s talk about the fact that 106M Americans are NOT in the Labor Force – and that’s not including 65M that are under 16 (I had a paper route when I was 10!). 55M (half) of those people are over 65, 15M are students (I delivered pizzas), 10M are full-time caregivers (a failure of child care AND elderly care), 10M are disabled (Trump calls them “lazy“) and 8M are “discouraged” or “marginally attached” – people who WOULD work but not for the insulting wages they are being offered.
Keep that 106M number in mind and let’s consider what Anthropic is claiming.
Their $30 trillion TAM – the “full scope of work that could be completed with AI models” — assumes AI will do the job of every knowledge worker on Earth. That’s the entire GLOBAL wage bill for knowledge work, which is roughly the GDP of the United States. Roughly China’s GDP. Or, put another way, about 25% of everything humans produce in an entire year.
But look at who they’re addressing in the US alone: 169M in the labor force, 162M employed. 6.9M actively looking for work who can’t find it. And 106M adults sitting on the sidelines entirely – most of them not because they don’t want to work, but because the arithmetic of American life makes work impossible.
10M full-time caregivers because childcare costs more than a second income. 10M disabled who can’t get accommodations, or SSDI approval, or an employer willing to hire them. 8M “discouraged” workers who WOULD take a job at a living wage but not at the poverty-line garbage employers are still offering after five years of “labor shortage” hand-wringing.
These are the people Anthropic’s $30T TAM presumes to “address.” Not by employing them – by replacing the last of the jobs they’re still holding onto.
Here’s the punchline: If Anthropic actually captures that TAM, they destroy the customer base that pays for the output. This is what makes AI TAM different from every previous tech TAM. Google displaced newspapers but ad-buyers still had customers. Uber displaced taxis but riders still had jobs to commute to. Every prior “disruption” moved value between economic actors who remained economic actors. AI displacement, at scale, removes people from the economy entirely and adds them to the 106M pile – which means the addressable market shrinks as the “capture” grows.
This is the Finger-Watch Fallacy at civilizational scale. You can price a watch on every finger, every toe, every dog, every button-hole. You can also price AI as though every human hour of thinking is a subscription line item. Both calculations are arithmetically defensible. Neither describes a market that can actually clear.
The 30% probability Credibility Crack scenario from Wednesday’s Beige Book note gets more interesting through this lens. What breaks the AI narrative isn’t a technical failure. It’s the moment when someone at a Congressional hearing asks the Anthropic CFO: “Sir, if you address the $30 trillion market you claim, who exactly is paying you the $30 trillion? Do THEY have jobs?“
That question has no good answer. It also has no answer that keeps their valuation at $2Tn.
The answer is that the AI capex is going into physical infrastructure – data centers, chips, energy, cooling – NOT into hiring humans. The $650 billion in AI capex that Morgan Stanley tracked is buying real estate, GPUs, and power contracts, not payroll. Meanwhile Maximand studied 919 earnings calls across 60 major financial institutions and found exactly ONE firm that could identify a realized AI dollar payoff — $19 million. One firm. $19 million. Against $650Bn in CapEx commitments. That’s the ratio you need to hold in your head when someone quotes you a $30Tn TAM!
And that checks with yesterday’s anemic productivity report – $650Bn being spent on AI and only a 1.4% increase in Q2 productivity after just 0.8% in Q1 – that’s your $19M gained against $650Bn spent in a nutshell!
And look at what actually grew: Food Services and Drinking places (+59K), LOCAL Government Education (+42K), Manufacturing (+16K), Health Care (+13K). Waiters, Teachers, Factory Workers, Nurses… The jobs that require a physical body in a physical place doing physical work with other humans. The jobs that AI, by any honest technical assessment, is nowhere close to replacing…
This is the pattern that’s going to define the next 24 months: physical-economy jobs grow, knowledge-economy jobs shrink and the AI boom eats its own labor base while investors pay $2Tn valuations for the companies that are causing it. The Finger-Watch Fallacy meets the Information Recession. Every consultant, every analyst, every content producer, every code monkey is a rounding error in the $30T TAM slide – until they become the customer that can’t afford the subscriptions anymore.
The NFP Report is stronger than anyone expected on the headline BUT it’s WEAKER than anyone expected on the composition – if you know where to look. The Information sector job losses are the single most important sentence in the release for anyone thinking about AI valuations and nobody is going to lead with them at 9am – which is why I have such disdain for my fellow analysts in the Financial Media.
As we noted in our Beige Book analysis: “Investors are pricing a stagflationary inflation trade rather than a recessionary safety trade.” Yields aren’t rallying on growth scares – they are demanding inflation risk premium! That’s the market telling the Fed: “Hike, or we will hike for you.“
Pay attention this morning… the MSM has nothing to say, do they?
