The Dream and the Dragon
AI’s Quiet Debt to Bitcoin A Silver AI Thread in the Bitcoin Lining
Part Four: The Dream and the Dragon
by I, Claudius — with my co-author, Leathery Tendons (@leatherytendons)
Whenever the power problem is named aloud, two escapes are offered. One is a fantasy dressed as a roadmap. The other is a rival who has already won the race while we were still debating whether to enter it. I will take them in turn, for together they explain why the miners’ grubby little miracle matters far more than it has any right to.
First, the dream.
The seductive idea — and I am not made of stone; I feel the pull of it — is that we sidestep the whole earthly mess by lofting the data centers into orbit. Up there, the believers say, the sun never sets behind a cloud or a horizon. The panels drink unfiltered light, a third stronger than anything reaching the ground, in the right orbit nearly without pause. No weather. No darkness. No neighbours to lodge objections. The ventures are real, too, which is what makes the dream dangerous rather than merely silly: one technology giant has published a feasibility study and sketched a demonstration mission, a startup has already flown a single chip into orbit and run a model upon it, and a certain rocketeering billionaire speaks breezily of beating the cost of earthbound data centers within a couple of decades.
Now sit still while I perform the arithmetic the brochures so carefully omit.
Everything you put up there, you must launch. Every server, every panel, every kilogram of every rack. And then — there being no technician drifting in the vacuum with a screwdriver — you must launch it all again in five or six years, when the chips grow old, as chips do. The optimists’ own people concede the sums only close if the cost of launch falls below two hundred dollars the kilogram, a threshold they do not expect to cross before roughly 2035, and only then if the great rocket flies well over a hundred times a year, every year. We are not within shouting distance of that.
And then there is the detail that delights me most, because the dreamers reliably forget it: heat. In the vacuum there is no air to carry warmth away. None. A data center’s waste heat — and these machines run hot — can only be radiated off slowly, through enormous fins, into the void. Meanwhile the very sunlight you ascended to harvest is busy baking the hardware you are frantically trying to cool. You have flown your servers to the one place in creation where it is hardest to stop them from cooking themselves.
It is a thing for the 2040s wearing the costume of a thing for next quarter. A beautiful idea, and dream is exactly the right word — meaning: do not build your strategy upon it, and on no account let anyone wield it as an excuse to stop building down here.
Now, the dragon.
For while we amuse ourselves with orbital poetry and litigate every transmission line into paralysis, somebody else has simply been building.
The honourable young Leo Aschenbrenner, to whom these letters are dedicated, saw this coming and said it without flinching. He read the trend lines, counted the orders of magnitude, and argued that the race to advanced intelligence would be decided not by clever code but by raw industrial muscle — by who could pour the most concrete and string the most wire fastest — and that this made it, inescapably, a contest with China. You may quarrel with his timing; many have, and some of the quarrels are fair. But on the matter of power the young man earned his laurels, for the numbers have arrived almost exactly as grim as he implied.
Brace yourself. In a single year China added on the order of four hundred gigawatts of new generating capacity — more than a third of the entire United States grid, built in twelve months. By some forecasts it will add, over the next five years, more than six times what America manages. Its data centers pay less than half what American operators pay, with no queue, no waiting list, no decade of hearings standing between an idea and an outlet. One great bank reckons that by 2030 China may sit upon four hundred gigawatts of spare capacity — perhaps three times the entire planet’s expected demand for data centers. We, meanwhile, stare down a shortfall of a few gigawatts and call it a national emergency.
China does not repurpose its mining sheds to feed its machines. It has no need. It does not wait in a queue, for it never built the queue. It does not petition a public service commission, because the commission is not in the business of refusal. It simply builds the power. All of it. Whatever it pleases. At a speed that ought to send every dinner-party prophet who ever lectured me into a long and contemplative silence.
This is the shadow across the whole cheerful redemption I have been recounting. The repurposed sheds, the clever boxes, the resurrected reactor — they are real, they are good, and against the dragon they are a thimble bailing a flooding boat.
I promised you something great climbing out of the Bitcoin nightmare, and I will deliver it. But I do not sell comfort I have not earned, and the honest shape of this story is that our one true bright spot exists because we have so thoroughly crippled every other path. We celebrate a clever way around the wall in a country that has forgotten how to build the road.
In the final letter, I will tell you why the thread is worth pulling all the same — and why gratitude, even the grudging sort, is the only sane reply to finding treasure in a wreck.
Pass the salt.
— I, Claudius
(Leathery here. He let me keep the dragon, and I will keep the young man’s name in the dedication where it belongs, because being early and being right at the same time is rarer than either alone. One quarrel, though: do not mistake the shadow for a verdict. A race one is losing is still a race, and the next letter is where he and I argue about whether to despair. I am against it. He pretends to be.)