ASHFALL INSTITUTE | SUBDUCTION ZONE
Building Bridges — Episode Six
THE GRID WE’D HAVE TO BUILD ANYWAY
P. A. Moore
Ashfall Institute | Subduction Zone
Written under the collaboration disclosed in Time Is a Wheel and We Are the Ground. Concept, argument and judgment: P. A. Moore. Research and composition: Claude Opus 5.
I want to start with a number, and I want you to sit with it before I tell you what it measures.
One hundred and eighty.
That is the number of miles of new high-voltage transmission line built in the United States in the most recent two-year period for which we have the figure.
One hundred and eighty miles. For a country three thousand miles across.
You could drive it in an afternoon.
Now let me give you the other number, the one that has been on television all autumn.
Two thousand six hundred gigawatts.
That is the interconnection queue — the line of projects waiting for permission to plug into the grid. The median wait to reach commercial operation is approaching five years. For some projects it runs to twelve.
In Texas alone, the large-load queue went from 63 gigawatts at the end of 2024 to 226 gigawatts a year later.
And you have been told, more or less continuously, that this is what artificial intelligence has done to your power supply.
So let me be precise about what is in that queue, because the composition is the argument.
The overwhelming majority of those 2,600 gigawatts is not data centers. It is generation — solar, wind, batteries, gas. Power plants that somebody has already financed and already sited and already decided to build, standing in a line, waiting for a study.
We are not short of people who want to build power. We are short of the ability to say yes to them.
The queue is not a demand problem. It is a paperwork problem, sitting on top of a wire problem.
And the wire problem is a hundred and eighty miles long.
Here is what I want to establish before anything else, and it is the whole of the episode in one sentence.
The grid was failing before anyone said the word “AI.”
We have known this for decades. Anyone who has lived through a summer in the South knows it. Anyone in California knows it. Anyone in Texas remembers February of 2021, when the thing went down and people died in their houses, and that was four years before a single one of these buildings was proposed.
The transmission system most of this country runs on was built in the middle of the last century, for a load profile that no longer exists, by utilities that were then allowed to earn a return on building it. We stopped building. The demand didn’t stop.
The data centers did not break the grid. They arrived at a grid that was already broken, and they turned the lights on in the room.
That is an uncomfortable thing to be, and I understand why nobody wants to be it. But it is not the same thing as being the cause.
The load is the excuse. It is not the reason.
Now. I am not going to pretend the load isn’t real. It is enormous, and I said in the first episode that I would not wave away a single grievance, so I won’t wave this one either.
A large data center draws what a small city draws. Twenty of them draw what a large one draws. That is a genuine engineering problem and the people running these systems are not exaggerating when they say so.
But there is a fact about that load that almost nobody outside the industry knows, and once you know it the entire conversation changes shape.
Most of that demand does not have to happen at any particular moment.
Some of it does. If you are in an emergency room and a machine is reading your scan, that has to happen now. But an enormous share of what runs in these buildings is training — grinding through a calculation for weeks — and batch work, and indexing, and jobs that genuinely do not care whether they run at four in the afternoon on the hottest day of August or at two in the morning in October.
A steel mill cannot do that. An aluminum smelter cannot do that. You cannot tell a blast furnace to take a two-hour break because the grid is tight, and you certainly cannot tell a hospital to.
But you can tell a training run.
Somebody went and did the arithmetic on that, and I think it is the single most important study nobody has heard of.
Researchers at Duke University’s Nicholas Institute asked a simple question. How much new load could the existing American grid absorb — with no new power plants and no new transmission — if the new loads agreed to throttle back occasionally?
Not shut down. Not go dark. Curtail, briefly, when the system is at its worst.
They gave it a name: curtailment-enabled headroom.
Here is the answer.
If large new loads agreed to curtail just half of one percent of the time — that is forty-four hours a year, in stretches averaging two hours — the existing grid could absorb roughly one hundred gigawatts of new load.
A hundred gigawatts. On the wires we already have. Without pouring a single foundation.
And it is distributed across the country where the fight is actually happening. At that half-percent level: eighteen gigawatts in PJM — that is Ohio and twelve other states, the same system where capacity costs went up by nine point three billion dollars in one year. Fifteen in the Midcontinent system. Ten in Texas. Ten in the Southwest Power Pool. Eight in the Southern Company territory, which is my part of the world.
Forty-four hours a year. Two hours at a stretch.
That is the difference between a decade of moratoriums and a solved problem, and it is sitting there unused.
So why hasn’t it been done?
Not because it’s hard. It is a great deal easier than building a transmission line, and the industry has been doing demand response with industrial customers for forty years.
It hasn’t been done because flexibility is a contract term, and nobody wrote it into the contract.
That is the third time in this series we have arrived at the same place. In the first episode, the ratepayer class was a contract term. In the third, the clawback and the disclosure were contract terms. Now the single largest available piece of grid capacity in the country is a contract term.
Every structural problem in this fight turns out to be a sentence somebody didn’t write.
And I want to be fair about why they didn’t. A data center operator signs a contract promising uptime to its customers, and every hour of curtailment is a risk against that promise. Firm power is worth more than interruptible power, and they have been buying firm power because that is what was on offer and nobody priced the alternative properly.
So price it. Make interruptible power meaningfully cheaper than firm power for very large loads. Make the curtailment obligation explicit, and bounded, and paid for. That is not a new invention. That is a tariff, and public utility commissions write tariffs every year of their lives.
The operators would take that deal in a heartbeat, because right now the alternative on the table is waiting seven years in a queue.
Now the part that matters to the town, which is what this series is actually about.
When one of these buildings arrives, somebody builds a substation. Somebody runs new conductor. Somebody upgrades a transformer that was installed when Eisenhower was president and has been quietly terrifying the local engineer ever since. Somebody, in some cases, builds a hundred megawatts of battery storage to firm the load.
All of that is grid. All of it is permanent. And all of it is being built in your county, with your rate base carrying part of the cost.
The question is not whether it gets built. It is getting built.
The question is who it serves afterwards.
A substation sized for a data center has capacity that does not vanish when the servers are idle. A reconductored line carries power in both directions. Battery storage installed to smooth an industrial load can hold a hospital and a water treatment plant through an outage — if the interconnection agreement says it can, and if somebody thought to ask.
That is a sentence in a contract. It is not a law of physics and it is not charity.
And here is the thing about grid infrastructure that makes it different from almost everything else in this argument: it stays. A company can leave. A building can go dark and be worth nothing. But the wire is in the ground and the substation is on its pad, and forty years from now somebody’s grandchildren will be drawing power through steel that was put up because a computer company needed it.
That is the same test I applied in the first episode, and grid passes it where the building does not.
A data center leaves a building and a bill. A transmission upgrade leaves a transmission upgrade.
We should be asking for a great deal more of the second one.
I want to put one man in this episode, because it has been all wires and tariffs and it should not end that way.
There would be no grid at all without an immigrant.
Direct current — Edison’s system, the one America started with — cannot be stepped up and down efficiently. Which means it cannot travel. In the 1880s a direct current station could serve customers about a mile away and no further. A city needed a power plant every mile. A farm got nothing, ever.
Alternating current can be transformed. Put it up to a high voltage, send it a long way with little loss, bring it back down at the other end. Every transmission line in this episode, every one of those hundred and eighty miles, every substation in every county fighting about this — all of it exists because of alternating current.
The polyphase system that made it work was designed by Nikola Tesla, who arrived in New York in 1884 from the Austrian Empire, with a letter of introduction and essentially nothing else.
He was, in the language of the time and of ours, an immigrant with an accent and a strange manner and no money.
The country he arrived in has spent a hundred and forty years running on what he brought with him.
I will leave that there.
So the ask for this episode is three sentences long, and none of it requires anyone to believe anything about artificial intelligence.
One. Create an interruptible tariff class for very large loads — priced meaningfully below firm power, with the curtailment obligation explicit, bounded, and compensated. The Duke work says that is worth on the order of a hundred gigawatts nationally. Cheaper than any megaproject and faster than all of them.
Two. Require that grid infrastructure built to serve a large load be sized and interconnected to serve the community it sits in — the substation, the conductor, the storage. Not as a donation. As a condition, written into the interconnection agreement, where a county can enforce it.
Three. Fix the queue. Two thousand six hundred gigawatts of generation is standing in a line waiting for a study, and most of it is not data centers and never was. That is an administrative failure of the first order and it is strangling the clean generation everyone claims to want.
None of that stops a building. None of it costs a moratorium. And every one of them leaves something behind that is still there in fifty years.
I keep coming back to the same shape.
We are told the fight is about whether to allow the machines.
The fight is about a tariff, an interconnection agreement, and a queue.
Those are documents. Written by people. In buildings. In daylight.
And there is a hundred gigawatts of capacity in this country that nobody is using, because it would have taken somebody forty-four hours a year and the willingness to write one more sentence into a contract.
One hundred and eighty miles in two years.
We are not being cautious. We are just not building.
Next time: a new Civilian Conservation Corps — three million men, three and a half billion trees, and what the modern equivalent would put in the ground.
Sources: US interconnection queue backlog and median wait times, 2026. National transmission construction figure, most recent two-year period. ERCOT large-load queue, late 2024 to late 2025. Duke University Nicholas Institute for Energy, Environment & Sustainability, “Rethinking Load Growth: Assessing the Potential for Integration of Large Flexible Loads in US Power Systems,” February 2025 — curtailment-enabled headroom of approximately 100 GW at 0.5% annual curtailment; balancing authority figures for PJM, MISO, ERCOT, SPP and Southern Company. PJM capacity cost increase, cited in Episode One. Nikola Tesla, arrival in the United States 1884; polyphase alternating current system.
Building Bridges is a fourteen-part series. Previous: Episode Five — The Shape of the Buildings. Next: Episode Seven — Twenty-Five of Thirty.
P. A. Moore is the pen name of Pamela King, philosopher and artist. Available through the Ashfall Institute.