We are all learning how to stop paying each other.
I build with AI all day long, and even, no kidding, in my sleep. I have an agency, so I build for my own business but mostly for others.
Often, a founder tells me some version of the same sentence: I used to have to pay someone for this, and now I don’t. Sometimes AI is doing work they could never have afforded, unlocking a previously out-of-reach capability. Sometimes AI is filling a role a human used to do; now they can keep more of the margin. Either way, the logic is the same. Automate more, keep more. All’s fair. This is capitalism, and that’s the game.
WHO WINS?
But there’s another side of that ledger, and I’m not sure this is all working out as planned.
Many of those same people are watching their jobs wobble, disappear, or stop feeling safe in the way they did a year ago. Service providers I’ve worked with for years are getting less work. Small businesses that were fine two years ago are not fine now. When enough of us get very good at not paying each other, the money that used to move between studios and freelancers and shops thins.
I have a career too, and before my own capital dwindles, I decided to ask myself a cold, hard question: In this new world, who wins?
I’m not an analyst, and I’m not in the business of telling anyone what stocks to buy. But I do place bets—every project I take on is a bet. Every year I spend building something is a bet. So the question I ask myself is: Where should I be focusing my own work?
Here’s where I’ve landed.
THE MIDDLE GETS EATEN
Let’s start with who I don’t think is winning: the middle of the stack.
Consider the frontier labs. They are extraordinary, and they are also, by their own numbers, torching cash. They are unprofitable, heavily subsidized by investors, and have no obvious road to profit that doesn’t run straight into ever-scarcer high-quality data and ever-scarcer electrical power. And they can’t eat the world, because they’re too busy eating each other. Every product release and update resets the board. That’s a knife fight, not a moat.
So I go down. All the way down.
THE GROUND
Underneath all of it—the models, the chips, the hype—are the critical materials they need to survive. That market is not commoditized. It is scarce, and getting scarcer.
Copper is the tell. The U.S. formally designated it a critical mineral in late 2025. It is the metal of electrification, threaded through every wire, every motor, every server. A single large AI data center can require up to 50,000 tonnes of it. Demand is projected to climb 50% by 2040 while supply falls behind, opening a gap that S&P Global estimates could reach 10 million metric tons. A new domestic mine takes 29 years to enter production, partly due to the permitting process. Copper prices hit record highs this year. You cannot vibe-code a copper deposit into existence.
Recently, Emma Gallo and I started GridMine to map the copper and rare earth magnets already pulled out of the earth—the above-ground reserve sitting in old wire and dead EV batteries and retiring turbines. These are critical materials available in months, not decades, and at higher grade than anything that gets pulled from a mine.
Critical materials are a phenomenal investment because they’re finite, and every dazzling thing we’re building runs on them.
Count me in.
THE ROOM FULL OF PEOPLE
The middle gets eaten, and the ground below it is one end of a barbell I’ll invest in. The other end is up and out—out of the house, off the screen, into a room with other people in it.
In a world where anyone can generate anything—when a song, an image, a video, an essay, a whole synthetic friend costs virtually nothing to make—the price of all of it goes to zero. And when the supply becomes infinite, the value moves to what the machine cannot make more of.
The machine cannot make more Saturday nights. It cannot put you in section 114 when the game goes to overtime. It cannot make the music that carries you away, or the stranger next to you grab your arm in excitement, or the whole crowd sing the same words at the same time. It cannot manufacture the fact that you were there. This is a scarce, finite product.
Ticket prices for live music and live sports have been climbing far faster than inflation for years, and the more of our lives we spend inside social media feeds, the steeper they get. Media rights for live games keep setting records while every other kind of programming gets cheaper to produce. A live game has to be watched right now, at the same time as everyone else, which is why 96 of the 100 most-watched broadcasts last year were sports. Restaurants are hard to get into. Runs, races, festivals, comedy, church, the gym, the trivia night. Any unifying event with other humans who are physically present will continue to increase in value.
Count me in there, too.
THE BARBELL
As an investment thesis, it’s simple. Go maximally physical, at both ends. The critical materials beneath us, and the rooms full of people around us. The machine eats the middle—the models, the compute, the endless generated content. In the process, this makes the two ends more valuable than they’ve ever been: what only the Earth can supply, and what only being there can supply.
That’s where I’m pointing the next few years, and all of my time these days. Let’s see what happens.
Lindsey Witmer Collins is CEO of WLCM App Studio and GridMine.us.
