Anthropic is about to go public, and according to Wall Street Journal reporting, the pitch to investors includes a number. The total addressable market for what they sell is more than $30 trillion a year.
Total addressable market is a simple idea. It’s a company telling investors how big the pie is in their sector. They aren’t necessarily projecting their share of it. They’re just saying what they think it is.
The entire global economy in 2025 was $118 trillion. The IMF projects $126 trillion for 2026. Back China out, because nobody expects Anthropic to be selling into Shenzhen, and you’re at about $105 trillion. SpaceX’s own IPO filing doesn’t even pretend on this. It cuts China and Russia out of its market figures entirely.
A Fight Worth Having is a reader-supported newsletter. Want to help me do more of what I do? You can give a one-time gift or become a monthly supporter.
$30 trillion out of $105 trillion is 28% of everything produced on this planet outside of China.
Nobody’s published the math on how they got there, but there’s really only one way:
Take the world’s GDP. If you multiply by 53%, which is the amount that the ILO estimates goes to workers (as opposed to owners), you get $66 trillion. That’s every wage or salary for every working person on earth. Then multiply that by 46%, which is the share of American wages from jobs that can be done entirely at a screen.
$126 trillion times 52.3% times 46% is $30.4 trillion.
That’s the number. When Anthropic says this is their market, they’re assuming every wage for every desk job in the world will go not to workers but to AI. Fortune called the estimate “outlandish” and I couldn’t agree more.
Let’s have a closer look at what’s in that 46%.
That’s an American number. It comes from an American study of American jobs. Running it against the whole world’s wages assumes every country on earth works like us: Canada, the UK, the EU, Japan, and also Nigeria, Bangladesh, Bolivia, Vietnam. It assumes the planet is one big office park.
It ain’t. A quarter of the world’s workers, 25.8% as of 2025, still work in agriculture. Close to 58% work informally, off the books entirely, with no payroll to replace in the first place.
The ILO went and measured this directly. Their global index of AI exposure, published last May, found that only 3.3% of global employment falls in the category most at risk of being replaced. 3.3 is somewhat smaller than 46. If you combine all categories with jobs at risk, you’re still only looking at a range of 11% of jobs in low income countries to 34% of jobs in rich ones, and the ILO’s conclusion was that most of those jobs will get transformed rather than eliminated.
Anthropic’s $30 trillion describes a planet where everybody has a desk. That planet doesn’t exist.
Their actual revenue run rate hit $65 billion at the end of July. Another slice of the pie, software spending that could genuinely be replaced by AI agents in the next four years, is about $234 billion according to Gartner. That’s 0.008% of what they pitched.
The great wealth heist over the last forty years has left our economy built on paper. Administrative fees and administrative bloat, waste and fraud and abuse, and it runs through every sector we depend on. Health care. Housing. Education. Finance. Every year those cost more for the same exact results. Look at these numbers:
Management: this includes 11 million workers, 7% of the American work force. They take home 15% of all wages.
Office and administrative support: 17.8 million.
Professional and business services: 13% of our GDP.
If you throw in finance and insurance and information, and you’re at nearly 27% of the American economy before anybody’s built a thing. So when AI companies promise to cut down on paperwork, I’m with them on the diagnosis. Neoliberalism is a cancer and we’re in the later stages of it. Most of that busywork shouldn’t exist in the first place.
The question is what it should be replaced with.
What we need is manufacturing capacity. Industrial capacity. Infrastructure. More people who know how to deliver babies instead of more people who know how to bill for it. Processing facilities for our food. Distribution networks that work. Things that get made and things that get built.
Nobody’s planning for what we actually need. AI companies aren’t, they’re planning an IPO. The government isn’t. The Democratic Party isn’t for damn sure.
Handing a bloated economy to seven AI companies may cut down on bureaucracy but it leaves our economy and us at the mercy of a handful of mad scientist billionaires.
Meanwhile the economy propping them up is running on the buildout itself. AI data centers and hardware are 1.4% of US GDP, double the 2015 to 2022 average. The financing has gotten circular enough that Bloomberg built a graphic to track who’s paying whom. Chip makers are investing in the model companies that buy their chips. Everybody’s revenue is somebody else’s capital expenditure.
Sam Altman and Elon Musk have both been clear about where this leads. Altman writes that intelligence will be “too cheap to meter” and admits to “whole classes of jobs going away.” Musk gives roughly 80% odds that eventually none of us will have jobs. Goldman Sachs estimates that 300 million full time jobs are in danger worldwide.
Take them at their word.
A company replaces a person with a machine because it’s cheaper. That’s the single diving principle of capitalism. Cheaper. Faster. More profitable. So the entire $30 trillion is, by design, wages that will go to stockholders instead of workers.
Those wages were somebody’s income. That income paid someone else’s mortgage. If you pull a third of the world’s purchasing power, who’s going to buy what the robots make? If you break the demand, the supply doesn’t matter. Your AI isn’t worth $30 trillion if customers no longer have $30 trillion to spend. You can’t count it twice.
What they’re describing is the Star Trek replicator. Or the Jetsons. AI does the administrative work and the management work, then robots build the goods we need. People are free to do whatever they want. Read, play, swim. If I stood up and described that future, I’d be called a utopian at best. Everyone knows you eventually run out of other people’s money.
But when the same future gets described by men raising a hundred billion dollars at a two trillion dollar valuation, we call it a growth story. But if nobody’s working, nobody’s earning, and nobody’s buying. So who pays the $30 trillion, and with what?
And if human labor isn’t necessary anymore, how is a trillionaire different from anyone else? Think about what a trillionaire actually is in that world. He didn’t build anything, because the machines build. He didn’t figure anything out, because the machines think, and they were trained on the collected output of every person who ever wrote anything down or picked up a wrench. His claim comes down to this: he happened to own the assets on the day the music stopped.
That’s the whole thesis. Freeze everybody in place. Everybody gets a UBI check calibrated to roughly where they already were. Rich stay rich, poor stay poor, and the justification is timing.
We are funded by supporters that do it because they want to, not because they have to. Wanna help? Give s one-time gift or become a monthly supporter.
In that world labor is free and knowledge is free. What still has value is the physical stuff. Land. Water. Lumber. Nickel. Platinum. Lithium. The rights to dig and the rights to build. If robots do the work and AI does the thinking, owning raw material is the entire game. Look at who’s buying land right now. Look at who’s buying mineral rights and water rights and power generation. So far it ain’t many people.
That handful of people are going to own the replicator, and the rest of us are going to be shopping with vouchers at their company store. We won’t own the land. We won’t own the machine. We’ll get an allowance and spend it back where it came from. And the owners will call it freedom because we didn’t have to go to work
A lot of Christians I know talk about trying to be Christ-like. They don’t expect to get there. They know they’ll fall short, because they’re people. But they hold it up anyway, and it makes them better, because they’ve got something to strive for.
If one of them said we might as well go all in on being bad because eventually we’re gonna give in to temptation, we’d know that guy was looking for an excuse. Yet in politics we put up with that all the time. Somebody says everybody ought to have housing, or care, or enough to eat, and before we’ve even sat down to talk about how, the answer comes down from on high that it’s unrealistic. Human nature. Won’t work. We’ve decided we’re going to fail before we’ve started.
The AI people have put a real question on the table, and it’s a better one than they meant to ask. They’re describing genuine abundance, a world where it no longer works to use the old excuse that there isn’t enough to go around. That’s a world worth wanting. I want that. Most people would.
What they haven’t done is say who the abundance belongs to. They assumed it was themselves, and moved on to the fundraising.
That’s the real conversation, and we ought to have it now, before the ownership is decided. A world where machines meet human needs could be the best thing that ever happened to us. It could also be the most complete concentration of power in history. Technology doesn’t get to decide that for us. That’s a political question. It always was.
So dream big. Dream of a better country and a better world, and don’t let anybody talk you out of wanting it.
Then let’s build it, because it isn’t going to arrive on its own. Building will take power, and we need to take ours back from the people hoarding more and more of it. They aren’t going to hand it over because we asked nicely.
Corbin Trent
Lenoir City, Tennessee
Here’s where you can find us in the fray, covering elections, grabbing third rails, arguing for an affordable and democratic future.
Here’s where you can learn more about us and join the fight.
Boiled down, distilled and described pretty much *on point*. Thanks Corbin! See you in the streets friend.
AI points to things we could just get rid of. As someone who has yet to use AI in anyway, talking to people has led me to believe that most things that they’re using AI for could probably just go away. Overwrought emails, over complicated forms, the busywork you speak of here. It is at least in part an over complicated answer to an over complicated question. If we were to pull that out of the percent of GDP then what’s left?