AMPLIFICATION
AI has the potential to usher in the most prosperous period in human history.
Too few believe it.
Fifty-two percent of Americans say the growing use of AI makes them more concerned than excited. Nine percent say the reverse.1 In China, eighty-three percent say these products do more good than harm. Here it is thirty-nine.2 We are building the technology and someone else is going to deploy it.
We used to hold World’s Fairs. The last one in this country was in 1984.3 The future used to arrive as something to be shown. Now it mostly arrives as a warning, and nearly every story told about this technology is a story about the harm it does.
“No society can surely be flourishing and happy, of which the far greater part of the members are poor and miserable.”
Smith was not moralizing. He was describing how systems fail. An economy that creates enormous wealth and shares it narrowly loses its legitimacy, and then it loses its future.4
The doubt is earned, and the newest reason is the technology itself.
The models were trained on what people wrote, recorded, photographed and built. The companies that trained them are worth trillions. The people whose work made that possible were not asked and were not paid.
That is the template the country just watched. It is a reasonable basis for expecting the next round to work the same way.
The longer record says the same thing. Goods kept getting cheaper while the things people actually need did not.5 Ninety-two percent of Americans born in 1940 out-earned their parents. Of those born in 1984, half did.6
Owners have their own version, which is watching a competitor get bought, loaded with debt and sold for parts.
So the owner waits. Waiting is the rational choice. From where he sits there is no way to tell a firm that will build something from a firm that will take the company apart.
Meanwhile, the work goes untouched.
The cost of running a model falls about tenfold a year.7 Almost none of that has reached the companies that dig, build, move and make. Thirty percent of American workers show no observed AI use at all. Not light use. None.8 Ninety-five percent of organizations report no measurable return on generative AI.9
The reason is not the software. The engineers who could build it do not run anything. The people who run things are busy, and the last three vendors oversold. An advisor recommends and leaves. A vendor sells a product built for someone else. Nobody in that picture holds the data, the workflow and the authority at the same time.
Capital has two answers for a company in that position. Neither one fixes it.
Private equity buys it.
That was an innovation once. It put owners back in the room and made management answer for results in a way the public markets had stopped requiring.
But the model begins with a judgment: a company growing slower than the economy around it is in decline. Once you have decided a business is in decline, there is nothing to do but take value out of it. Leverage, consolidation, cost, exit. The industry is now holding roughly 32,000 companies it cannot sell, worth $3.8 trillion, with distributions under fifteen percent of net asset value four years running.10 Those businesses are not being improved. They are being held.
Venture funds the replacement.
It produced the personal computer, the smartphone, and several of the most valuable companies in history. Software had no weight: no permits, no trucks, no workforce to win over, no customer on a twelve-year contract.
Atoms are slower, and the incumbent already owns everything the challenger has to go and build. Even PitchBook expects execution rather than invention to decide returns in these markets.11 And in the real economy, disruption means telling a family that the thing they spent their life building is the problem.
Buy an established company and build the technology inside it.
Slow growth is not decline. It is capacity nobody has used. A fifty-year-old contractor has the customers, the licenses, the field data and the people who know exactly where the work breaks down. What it does not have is the software that the largest companies in its industry already run.
Ownership puts the data, the workflow and the authority in one place.
Not because control makes a rollout easier. Because the insight that matters only exists inside the operation, and cannot be bought from outside it. Because the regulated customer will not take a call from a startup. And because people stop resisting a new system when they can see what is in it for them, and only the owner can put it there.
So we buy the company and we staff it. Our engineers and operators work beside the people already running it and build from that company’s own data and its own workflows.
The structure is conservative on purpose. Low leverage, little cash at close, the seller keeping a stake. That is the floor, and it is set before we deliver any technology at all. The transformation is the upside. Occasionally what gets built has a market outside the company that built it, and becomes a business of its own.
Efficiency is how this works, not what it is for.
The question is what people do with the hours and the money it hands back.
Machines will take the work people want least, and the same tools are pointed at harder things: materials, medicine, longer lives. None of that reaches anyone automatically. It has not for a generation, which is why so few believe it will now.
So the structure has to carry it. The owner keeps a stake. The crew that has run the work for twenty years runs it after the sale, with better tools and a share of what those tools produce.
That is what we exist to build: the capitalist solution that makes frontier AI technology the reason ordinary companies become the most valuable businesses of the next century.
Human flourishing is an engineering and empathy challenge. Both are required for success.
- Pew Research Center, survey of U.S. adults, June 2026 ↑
- Stanford HAI, AI Index Report 2025, Ipsos data ↑
- Bureau International des Expositions, United States ↑
- Adam Smith, The Wealth of Nations, Book I, Chapter 8, 1776 ↑
- U.S. Bureau of Labor Statistics CPI, January 2000 to June 2021, compiled by Mark J. Perry, American Enterprise Institute ↑
- Chetty et al., “The Fading American Dream,” Science, 2017 ↑
- Andreessen Horowitz, “Welcome to LLMflation,” November 2024 ↑
- Massenkoff and McCrory, “Labor market impacts of AI: A new measure and early evidence,” Anthropic, March 2026 ↑
- MIT NANDA, The GenAI Divide: State of AI in Business 2025, July 2025, as reported by Fortune ↑
- Bain & Company, Global Private Equity Report 2026, February 2026 ↑
- PitchBook, as reported by Defense News, January 2026 ↑