Field of Dreams

THE NUMBER: $1.1 trillion. That’s the US data-center build already in the pipeline — roughly 604 projects, about half the planet’s facilities sitting on American soil. Hold it next to the number we opened on Monday: $40 trillion of federal debt. Everyone is arguing about the debt. Almost nobody has noticed that the thing that pays it down and the thing everybody’s protesting are the same thing. The $1.1 trillion isn’t a tech budget. It’s what “grow our way out” actually costs to pour into the ground.

The movie you think you remember

Field of Dreams is not a baseball movie. People remember Kevin Costner and the cornfield and the catch with his dad, and they miss what the whole thing is actually about, which is a country that lost something and wants it back badly enough to pay for the ticket.

The dream in the field is a specific year. It’s the game before 1919 — before the Black Sox took the money and threw the World Series and taught every kid in America that the thing they loved could be bought. Shoeless Joe walks out of the corn to play again for nothing, for the love of it, in the world as it was before the fix. That’s the twenty dollars. People drive to Iowa and hand it over to stand for one afternoon in the country before it got cynical.

We are a long way into the after now. The politicians divide us on purpose because division is the business model. Social media divides us because the algorithm eats outrage and starves everything else. The networks pick a side before the anchor sits down. And a whole generation has come up inside that and never seen anything else — never watched Walter Cronkite read the news straight and then, separately, marked when he was giving you his opinion, because the line between the two was a thing that existed. They’ve never felt the country pull in one direction the way it did in a war, or the way it did for about three weeks after 9/11 before we let that go too.

I bring up a baseball movie because I think we are about to make a mistake with the biggest infrastructure project of our lifetimes, and the mistake is a failure of imagination about what it could be. Right now the data center is a symbol of everything people hate: tech run amok, a few men getting obscenely rich, a boom that shows up as a higher electric bill and a fenced-off building where the field used to be. That’s one reading. There’s another one sitting right there, and no politician has had the nerve to pick it up.

Door C

Monday we laid out the strait. Forty trillion in debt, funded by the tax withheld off a human paycheck; an AI industry whose entire product is doing the work without the paycheck. Scylla if the machines win and the wage base cracks, Charybdis if they stall and the whole levered market unwinds with the debt still sitting there. The only channel we could find between the rocks was the quiet one: inflate the debt away and send the bill to savers. We told you to pick a side. I said I’d rather be long the machines.

Scott Bessent has a louder answer. On CNBC on August 20 the Treasury Secretary waved the whole thing off: “There’s nothing magic about the $40 trillion number. And we can grow our way out of that.” No tax hikes anyone will feel. No spending cuts anyone will miss. Just growth — “pull back the slingshot,” he said, “as the factories come online.” Door C. The one door that doesn’t hurt on the way through.

Everybody dunked on it. The budget shops ran the tape and it’s ugly: to retire that pile on growth alone you’d need real GDP running north of four percent a year for a decade, and the CBO has us at 1.8. One economist put it flat — the country has no chance of growing out through growth only. And they’re right about the arithmetic. What they missed is that Bessent wasn’t giving a forecast. He was giving a build order. Because there is exactly one way for this economy to grow fast enough to matter and keep the market aloft while it does, and it’s pouring data centers and running the whole AI stack hot. The factories he’s loading into the slingshot aren’t a metaphor. They’re the $1.1 trillion already in the pipeline.

We have seen this movie

Here’s where my finance brain kicks in, because I’ve watched this exact film once before and so have you, even if you don’t think of it that way.

In the late 1990s the country decided the internet needed rails, and it built them the way America builds things: all at once, on borrowed money, way past what anyone could use. Telecom firms issued more than $500 billion in new bonds between 1996 and 2001 and buried the ground in glass. Then the music stopped. Less than five percent of the fiber they laid was ever lit. Global telecom stocks lost more than two trillion dollars in market value. WorldCom went down in what was then the largest bankruptcy in American history; Global Crossing and 360networks right behind it. Every cocktail-party genius called it the dumbest capital bonfire in a generation.

And then. That dark fiber — the ninety-five percent nobody could use, the stuff that bankrupted the men who laid it — got lit. It became the cheap capacity that made broadband, cloud computing, and video streaming possible in the decade that followed. You are reading this on that overbuild. Almost every dollar of growth the American internet has thrown off for twenty-five years runs on rails that were poured by companies that went to zero. The builders went broke. The country got rich. That is not a tragedy. That is how the rails get laid.

AI’s data centers are that cycle again, and the parallel is close enough to be spooky — the same borrowed money, the same too-much-too-fast, the same certainty from smart people that it’s a bubble. Some of it is a bubble. That’s the part I want to be very clear about, because it’s the through-line from Monday: be as bearish as you want on the paper. The equities priced for perfection, the take-or-pay contracts, the data-center bonds the smart money is busy packaging up for everyone else — a lot of that gets torched, exactly like the telecom paper did. Monday’s warning stands in full. Own that view.

But the paper is not the point. The point is the rail. When the financiers are carried out, the buildings are still standing, the power is still flowing, the fiber is still lit, and the country still owns the physical layer of the intelligence economy. Bearish on the paper, bullish on the build. Both. At once.

Who gets rich is not the issue

The loudest objection to all of this isn’t really about kilowatts. It’s about the men. A handful of people are going to get obscenely, generationally rich off this buildout, and a lot of Americans find that intolerable, and I understand why. But it’s the wrong thing to fixate on, and the fixation is going to cost us.

The problem-solvers are almost all easy to hate. Elon Musk rubs half the country the wrong way and seems to enjoy it. Jeff Bezos is the face of the guy who gets automated off the line and the boss who flew off to a wedding while it happened. Dario Amodei is a genuinely strange bird. Sam Altman spends half his time explaining the last thing he said. None of that is the measure that matters. The measure is: do they ship? Musk solves problems and moves technology forward at a pace no committee on earth can match. Altman marshaled the capital and the talent that dragged this whole field into the daylight. These are the people who actually build the thing.

Now look at who gets the applause. We pinned a halo on MacKenzie Scott for giving away twenty-six billion dollars, and it was a genuinely generous act, and I’m glad she did it. But be honest about the scoreboard. That money went, in large part, into the same nonprofit machine that’s been fed for decades — and homelessness is worse, poverty is right where it was, hunger persists, and cancer is still cancer. The problems did not move. I’m not knocking charity. I’m saying we have the incentives exactly backwards: we canonize the giving-away and revile the building, and the building is the only one of the two that has ever actually solved a problem at scale. Electricity solved a problem. The polio vaccine solved a problem. The transistor solved a problem. AI, and the data centers underneath it, is how this generation gets a real shot at the ones still standing.

And the meter is running on the alternative. Forty trillion dollars isn’t a number that sits still. It mortgages your kids, and their kids, and everyone who comes after. Doing nothing is not neutral. Doing nothing is the most expensive option on the board.

These are the manufacturing jobs

Here is the part no one running for office has been willing to say plainly, so I will: this is the reindustrialization of America. It is the thing every politician of my lifetime has promised and none has delivered, and it’s arriving in a shape nobody campaigned on.

Washington has spent a decade promising the manufacturing jobs would come home. They’re not coming home, not the way they left. The assembly line that employed a town isn’t being rebuilt. But the jobs are real, and they’re enormous, and they’re the ones that can’t be shipped overseas or handed to an agent: the construction crews, the electricians, the plumbers, the steel and the concrete, the people who spend years standing up a million square feet of building and then spend decades keeping it running. What’s inside the building is the automated factory that lets the country grow. What builds and tends the building is a generation of trades work with a paycheck attached.

I drove it this summer. Route 81 from Connecticut down to Georgia, and another stretch across Minnesota into North Dakota — hundreds of miles of trees and mountains and farms and road and almost nobody. This country is not full. It is mostly empty, and the empty part is exactly what this buildout needs. We have the land. We have the engineers. What the empty land needs is two things: power and fiber. Add those, and the field grows.

And underneath the jobs argument is the one that actually keeps me up: sovereignty. A country that lets the future get built somewhere else does not get to keep a standard of living higher than the rest of the world. Higher real incomes, first-world wages, the whole arrangement we take for granted — that is a function of building the next thing here, on our soil, under our law. Let it get built in another country and we become a country that rents the future instead of owning it, and the standard of living follows the ownership out the door. This is not a tech story. It’s a question of whether America leads the next fifty years or watches someone else do it.

The governor who hasn’t shown up yet

So here’s what boggles my mind. We have empty land by the millions of acres. The federal government alone owns something like 640 million acres — roughly a quarter of the country — most of it out West where the land is emptiest and the sun and gas and uranium are. The demand is screaming. The capital is lined up. And not one governor has walked to a podium and done the obvious thing.

The obvious thing is this. Some governor looks at a few square miles of empty state land, or the federal land right next to it, and says: I will build the data center. I will build the nuclear plant to power it. I will build the waste handling and the substation and run the fiber to the door. I’ll bring the trades and the tax base and the cheaper power for the town next door. All I need from Washington is the red tape cut so it can happen this decade instead of never.

That’s the move. And it should be a competition. We used to fight over this stuff. Towns lobbied and bribed and begged to get the transcontinental railroad to run through them, because the ones the rail skipped died and everyone knew it. Cities competed for the new airport, for the interstate interchange, because being on the grid was the difference between a future and a slow fade. This is that, exactly that, and instead of competing to host it we’re running attack ads against it. We wrote a memo to the governors three weeks ago making this case. Bessent and the President just spent the last two weeks confirming it. The map is going to get redrawn around this buildout whether a given state participates or not. The only question each governor actually gets to answer is whether their state is on the map or off it.

The objections, corrected

The case against gets made in three parts, and all three are weaker than they sound.

The water panic is aimed at the last war. The data centers people picture — the ones that drink millions of gallons through evaporative cooling — are the old design. The new AI builds run closed-loop; Microsoft says one of its new centers uses about as much water in a year as a restaurant. Meanwhile a single California almond drinks a gallon and nobody’s marching on the orchards. The water complaint was never really about hydrology. It’s a proxy for a fair question — who gets the benefit — and that question has a real answer.

The tax fights came from bad early deals, and the towns are learning. The first wave of communities got rolled: they handed out abatements and ate the interconnection cost on their own power bills. The ones paying attention now are writing the opposite deal — make the hyperscaler overbuild the grid so residents get cheaper, more reliable power, tax the racks to rebuild roads and schools, and land the trades jobs on the way through. Done right, the data center lowers your bill. Done wrong, it raises it. The difference is the deal, not the building.

And the real constraint isn’t water or votes at all. It’s power. The country is something like 90 gigawatts short of the firm power this needs by 2030 — call it a hundred power plants we haven’t built. That’s the actual gate, and it’s not a physics problem. We know how to make firm power; the Navy has run reactors for six thousand-plus reactor-years without a reactor accident. It’s a permitting problem, which means it’s a nerve problem, which means it’s solvable by exactly the governor described above and the red-tape cut described above. Wind and solar can’t scale fast enough for a load that runs flat around the clock. This is going to be nuclear and gas, and the sooner we say so, the sooner it gets built.

The fair hearing

Let me give the bear his due, because I’ve been early on enough calls to respect the other side, and because Monday’s piece was the bear.

The honest risk is timing, and it’s real. The debt is long, the take-or-pay contracts are long, the data-center bonds are long, and the productivity payoff that justifies all of it is late and uneven. Something can give in that gap. The financiers securitizing this exposure and selling it down to pensions and retail are doing precisely what the mortgage machine did in 2006, and the late money is going to get hurt precisely the way it did then and the way the telecom bondholders did in 2001. If you’re reaching for a compute-backed yield product right now because it looks safe, you are the bagholder this cycle is manufacturing. That view — Monday’s view — is intact.

But read the fair hearing all the way through, because the telecom bust is the answer to it, not the argument against the build. The 2000-2002 wipeout was catastrophic for the people who owned the paper and irrelevant to the fact that the rails got laid and the country spent the next twenty-five years getting rich off them. The bust and the build are two different events that happen to the same asset. You can lose your shirt on the equity and still want the fiber in the ground. That’s the whole trick of holding both ideas at once, and it’s the thing the doom writers and the boosters both refuse to do.

Where I stand

So where does that leave a person trying to figure out what to actually do?

Bearish on the paper. I wouldn’t be long the frontier at these prices, I wouldn’t touch the securitized data-center debt the smart money is handing to the public, and I’d assume a market that overcorrects hard on every leg of this — up on the hype, down on the reset, and probably both inside the same year. Tax policy will lag the economics by years and throw off distortions and headlines the whole way. All of that is coming. Plan for it.

And bullish on the build, without reservation. Own the fortresses and the physical layer, not the story stocks — the balance sheets that survive the reset, the power upstream of the rack, the land, the trades. Because when the paper is done cratering and the financiers are carried off, the buildings are still there, the power is still flowing, and the country still owns the machine that runs the next economy. We will have the resources right here to lead the world for fifty years — until the next thing comes along, whether that’s space, where our lead is already unassailable, or fusion, or some miracle we can’t name yet from a lab that only exists because we built the compute underneath it.

I keep coming back to a town in West Virginia with the mine closed and the young people gone. Everyone in Washington has promised that town its old life back for forty years, and everyone has lied, because coal mining is not coming back and everybody there knows it. But that town has land, and it has people who know how to build hard things and keep them running, and it sits on a grid that needs everything this buildout produces. Coal mining isn’t coming back to West Virginia. Data mining might. The only question left is whether we have the nerve to build it here.

---

The email edition — THE NUMBER, the three moves, and the day’s five stories — is in your inbox. If figuring out which side of this you’re on — bearish on the paper, bullish on the build, and where that leaves your own capital or your own town — is the question keeping you up, that’s the conversation we run at Outsider Labs.