You Already Won the Game

Let me open with a sentence you would not expect from a publication that carries a labor journalist's name: capital is the lifeblood of this economy, and the men and women who risk it, gather it, and build with it have done something most of us could not. Somebody had to believe in the thing before it existed. Somebody had to put money on the table and stand to lose it. Strip that away and the lights never come on, the wages are never paid, the great machine on the edge of town is never built at all.

I want that said plainly and up front, because everything that follows is meant in good faith — one American to another — and the first act of good faith is to give credit where it is owed. My grandfather covered the 1934 waterfront as what he called himself, plainly: a rank-and-file journalist.¹³ He did not write to burn the owners down. He wrote so that everyone doing the lifting would be counted in the ledger. I'll keep his habit — real figures, honestly come by — but I will count everyone in this ledger, the investor included.

This is the first of a companion series, addressed not to the picket line but to the boardroom. Its whole argument is that there is an open road here that almost nobody is taking, and that the people with the means to take it would be honored for it across generations. We begin where every honest accounting begins: with the cost.

The industry prices these machines by the megawatt of electricity the building can pull. An ordinary data center runs about $11.3 million per megawatt to build in 2026, by JLL's reckoning, and that figure has climbed roughly seven percent a year for half a decade.¹ An AI-grade hall — dense racks, liquid cooling, the works — clears $20 million per megawatt before a single server or acre of land is bought.² That is not waste. That is ambition rendered in concrete and copper.

Then stop thinking about one megawatt and think in gigawatts — a thousand at a time, the scale now being announced. Researchers at Epoch AI totaled the all-in capital for a single one-gigawatt AI data center and arrived at roughly $38 billion; other builders quote $45 to $55 billion.³ That is more than the annual budget of a good-sized state, marshaled by private hands and pointed at one project. Hold the wonder of it a moment. No king in history could command that. Capital can.

About sixty cents of every dollar in the lifetime cost is the servers themselves — the chips and racks, bought from a few firms, shipped in on trucks.³ The concrete and the local trades are real, and they matter, but they are the smaller and more temporary partner in the deal. That is not a complaint. It is the map — and a good map is the beginning of a good plan.

No king in history could command thirty-eight billion dollars for a single work. Capital can — and that is a kind of power worth deploying on purpose.

Once built, the year-to-year cost is comparatively small. For that one-gigawatt machine, annual operating expense runs around $0.9 billion, of which electricity — the largest line — is roughly $0.6 billion a year.³ Per megawatt, a well-run facility spends one to one-and-a-half million dollars a year to keep humming. A 100-megawatt center running flat out burns about 876 million kilowatt-hours a year — the appetite of a small city that never sleeps; a $41-million electric bill in a cheap-power state, north of $131 million in an expensive one.⁴

Read that arithmetic with care, because it is the hinge of everything. The cost is colossal and front-loaded; the running of it is modest and mostly an electric bill; the human payroll is small by design. This is a machine for turning a mountain of capital and a river of cheap power into compute. There is real genius in it. But it also means the value pools at the top of the structure, with the people who financed it — which is exactly why the holder of capital, and almost no one else, has the standing to choose what kind of neighbor this thing becomes.

When a machine that drinks like a small city plugs into a grid built for the county that was there before it, a bill arrives. Somebody must pay to enlarge the grid. In the PJM region — sixty-five million people across thirteen states — the cost of securing power supply jumped from $2.2 billion to $14.7 billion in a single year, with data centers accounting for nearly two-thirds of the rise.⁵ Residential rates nationally climbed about 32 percent from 2020 to 2025.⁶ The industry's consultants rightly note that data centers are not the sole cause of every hike — inflation, aging wires, and gas prices all play their parts, and fairness demands we say so. Yet Virginia's own auditors reckon a typical household there could see power costs rise $14 to $37 a month by 2040 as this load piles on.⁷

That is where the trouble starts, and I want to name it precisely, because heading it off is the reason this series exists. On one side of the table sits a return as high as anything in modern finance. On the other sits a family doing the plain arithmetic of a life — the light bill, the rent, the food on the table — and watching one of those numbers tick upward to feed a machine they will never own and rarely understand. Set those two across from each other and you have manufactured a fight: profit pulling against the kitchen table. It is an ugly, avoidable fight, and right now we are sleepwalking into it.

And here is the part that should trouble a thoughtful investor most, because it is being done in your name and it is not even buying you anything. The honest jobs evidence, freshly gathered by Brookings economists across some 770 facilities, is that these centers do create local jobs — total private employment up four to five percent over five or six years — but that booster estimates overstate the effect by a factor of three, and the permanent staff of a single facility is often only dozens to a few hundred.⁸ Worse: in the counties that generate the fewest jobs, public incentives run to 62 percent of total investment, while in the counties that generate the most, incentives are barely 2 percent — because those projects would have come anyway.⁸ We are handing the largest subsidies to the smallest givers. No wonder seven in ten Americans now say they would rather not have one built nearby; more than a hundred communities have passed moratoriums; over three hundred bills hit statehouses in the first six weeks of 2026.⁹ The towns have read their ledger and begun, quietly, to say no.

Now let me lay the investor's hand face-up, with respect, because the case I am making depends on its being true. A lender here looks for six to eight percent. An equity holder in a stabilized facility targets fifteen to twenty. The developers who get in early — land, power, ground-up, in a tight market — are underwriting internal rates of return of 25 to 40 percent, on margins of 50 to 65 percent, over holds of three or four years.¹⁰ Ten- and fifteen-year leases to the most creditworthy tenants on earth make the whole thing behave more like a toll bridge than a building. The hyperscalers alone will commit something like $674 billion this year, on the way to a forecast $1.7 trillion a year by decade's end.¹¹ By any measure ever kept, this is winning.

So here is the question this series is built around, and I ask it without a trace of resentment: how much more do you need? You have won the game so thoroughly that the binding constraint is no longer money, and no longer even chips. It is permission — the grid hookup the regulator will sign, the zoning the county will grant, the water the town will share, the welcome that is curdling into refusal across seventy percent of the country. And none of the winnings follow you out the door. No one in the history of the world has carried a balance sheet past the grave. What you carry — what your name carries — is what you built, and how you built it.

There is a road here that turns the avoidable fight into an alliance, and it costs less than the fight will. Reframe the deal from extraction to partnership and the family at the kitchen table stops being your adversary and becomes your advocate — because now the machine on the edge of town is partly theirs, too. Five terms open that road. None of them sinks the deal; together they trim a thirty-five-percent return into the twenties and buy you the one thing money cannot otherwise purchase: a community that holds the door open. Each is an offer, not a demand — but offered in earnest.

01Pay your own freight on power.

02Give the town a piece of the building.

03Steward the water, in writing.

04Build the payroll on purpose.

05Pay a fair tax; take a fair break.

Run those five through your model. The grid term costs capital you were going to spend anyway; the equity slice and the dividend trim the return from, say, thirty-five percent to the high twenties. You are still earning a multiple of what your bonds pay — and you have bought the welcome your competitors cannot underwrite. Here is the test to carry into every deal.

Henry Ford was no one's socialist, and he grasped that a worker who could afford the car was worth more than a worker squeezed to the bone. The Carnegie libraries still stand in towns that have long forgotten the steel. The men we remember well are not the ones who held the largest balance at the end — that number died with them — but the ones who turned their winnings into something a community would thank them for across generations. That is a return no market quotes, and it is the only one you get to keep.

You cannot take the profits with you. You can leave a legacy that makes America better — and be revered for it.

So this is the invitation, offered in earnest to the people who hold the means: you have already won. Build the machine — we need it, and you are extraordinary at building it. But build it the way you would on a street where your own mother pays the light bill. Do that, and the fight never comes; the moratorium never comes for the partner, only for the extractor. Do that, and a hundred years from now your name will be on the good thing in the town square.

The machine age will be financed one deal at a time, whether the community is welcomed into the deal or not. Let this one be remembered for how it treated the people who kept its lights on. That is the whole of the appeal. The rest of this series will lay the road out, brick by brick.