The record moved this year. Nobody can wind a magnet from what moved it.
Superconductors are the future.
Everything that moves electricity loses some of it as heat. Below one temperature, that loss does not shrink. It vanishes.
The transition, drawn
One wire, up close
Resistance
0
Move across the field to cool it
One wire, seen close, at the temperature on the bar. Warm, its electrons scatter off the shivering lattice and every collision sheds heat. Cool it and they still scatter, until one temperature: then they pair, run straight through, and the resistance is not smaller, it is zero. The curve underneath is that drop. The mark carries no number: the search is for a material whose mark sits above the record.
Everything that moves electricity
Loses nothing between the plant and the plug, at any length, at any current.
Holds a field no permanent magnet can reach, without a refrigerator built around it.
A fraction of the weight for the same power, in a ship, a turbine, an aircraft.
Switches at a fraction of a transistor's energy, without the refrigerator that keeps superconducting logic in a lab today.
Carries several times the power in the same trench, and loses almost none of it on the way.
Floats on magnets that carry no cold, so the line is lighter, cheaper, and not the only one.
Fusion is a superconducting magnet before it is anything else. Warmer is smaller, and smaller gets built.
Energy held as a current in a loop, indefinitely, with no chemistry to wear out.
Where it stands today, in four numbers
30,000+Medicine
MRI scanners installed, nearly all of them a magnet in a bath of liquid helium at 4 K. Take the helium out and the scanner fits a clinic, an ambulance, a field hospital.
1,232Science
superconducting dipoles hold the beam at 1.9 K inside the largest cryogenic system on Earth. Every fusion reactor on a drawing board is built around the same kind of magnet, and warmer means smaller and cheaper at every scale.
603Transport
km/h, the fastest a train has ever run, on magnets riding at 4 K in liquid helium. The cold is the weight, the cost and the reason there is one line.
5%The grid
of the electricity the United States moves on its grid is lost in the wires. A superconducting line loses almost none of it and carries several times the power in the same trench.
How we beat the giants
Bypassing the compute bottleneck.
Trillions of possible compositions. One sweep.
Legacy brute-force compute
Full physics, one candidate at a time.
A hyperscaler's compute, spent
AI-native MEL pipeline
The critical variables, screened in parallel.
Ours
A schematic comparison. On the left, a grid of cells is evaluated one cell at a time, and after six seconds the scan has covered about a seventh of the first two rows. On the right, a stream of candidates converges on a narrow gate over the same six seconds; most deflect away and fade, and the few that pass fill an output column completely. A progress rail under each panel shows how far each got: the right one full, the left one a sliver. The play holds for four seconds and runs again.
- 01
The problem
The search below is a compute problem before it is a physics problem, and the labs with the most compute are expected to win it. We are not going to out-spend them, so we changed what has to be computed.
- 02
What we changed
The Modulated Electron Lattice (MEL) framework was written for machines from its first line rather than ported to a GPU after the fact. It isolates the few variables that decide whether a material superconducts, local electron crowding first among them, and hands the model exactly those, at a fraction of the compute a brute-force sweep burns.
- 03
Why the gap widens
Which is how a company without a hyperscaler's budget sweeps the whole space, and why the gap keeps widening: every bench result that comes home trains the model, so each sweep starts sharper than the last.
Independently validated
Phys. Rev. Lett.2026
The effect the model ranks on was measured independently at Stanford and SLAC.
Superconductivity reinforces charge-density-wave phase coherence across cuprates
The search we run for ourselves.
Trillions of possible materials. One worth the century.
Discovery runs here from day one. The same model that answers a device team's brief is pointed at the band above the record, and every program and every bench result sharpens it.
The record
Thirty-three years. What finally beat it lasts three days.
Houston got 151 K this year by squeezing Hg-1223 and letting go. It survives about three days at 77 K, in a flake eighty microns wide.
The hydrides go higher, at around 150 gigapascals. That pressure exists inside a diamond anvil cell and nowhere anyone can wind a magnet.
Liquid nitrogen77 K
A warm room293 K
158 K
of nothing
Hg
1911 · 4 K
Hg-1223
1993 · 135 K
Hg-1223, pressure-quenched
2026 · 151 K
about three days at 77 K
1911
1950
2026
Highest measured transition temperature at ambient pressure, by year of first report. Linear scale, zero at the foot. The dashed point is a metastable phase, not a standing record.
The model
Our model of superconductivity, pointed at the band above the record.
The target
Above liquid nitrogen, at one atmosphere, makeable in length. Nobody sent it.
The search
Six constraints at once: phase, pressure, current, route, cost, length. Everything that clears them is new.
Into the lab
What survives is built exactly as the physics wrote it.
Made with
Measured
You cannot teach a model their intuition. So they work with it.
Called first
We put our number on the record before the instrument runs. Then we go the other way: in a closed test, we erase pieces of a century of settled research, and the platform gives them back, matching the published record.
Independently validated
The effect the platform relies on has been measured independently at Stanford and SLAC, by people who have never worked with us and owe us nothing.
Phys. Rev. Lett., 2026Measured, not by us
The return
Hit or miss, every result comes back and rewrites the model.
The journals only print what worked. Every miss stays with us, and the next search starts from what nobody else knows.
What the search finds, we file. Materials, routes and process know-how are licensed, and a licence is priced per unit built.
The scene opens on the highest transition temperature anyone has measured at ambient pressure, plotted against the year it was first reported, from mercury at 4.2 kelvin in 1911 to Hg-1223 at 135 kelvin in 1993. It does not move again for thirty-three years. In 2026 a pressure-quenched phase of that same compound reached 151 kelvin at ambient pressure, and it survives about three days at 77 kelvin. The empty band between the record and a warm room fills with points, and those points become the model: our model of superconductivity, built on the MEL framework, pointed at the band it has just drawn. The target is the material the field has waited a century for. The space gathers round the model out of the dark, trillions of possibilities, and the model sweeps it in rings from the model outward, each ring one constraint, bypassing the compute the search would otherwise cost; what a ring takes ignites and falls away. Six of the constraints are shown: whether the material forms a phase at all, whether it is stable at one atmosphere, whether it carries current at the field the brief asks for, whether a synthesis route exists, what it costs per kiloamp metre, and whether it holds up in length. Everything the search keeps is new. What survives is made with our partners: Brookhaven National Laboratory, the Walther-Meissner-Institut, UC Berkeley, CAN Superconductors and Eloi Materials. Nobody at SuperMatics then measures any of it. Most of the measurement runs at the bench that made the sample, at UC Berkeley and CAN Superconductors, and the hardest of it at the University of Illinois Urbana-Champaign and at Georgia Tech. Every result, on spec or not, returns to the model.
Built with the people who know
You cannot teach a model their intuition. So they work with it.
The physics was written here
The framework the model runs on has one author on the team and two decades behind it. The people who wrote it decide what it searches.
The measurers publish under their own names
Crystal growers and scanning tunneling groups whose careers are the intuition a model cannot be taught. They work with it, hand in hand.
Every result comes home
Hit or miss, with the conditions it was taken under. The journals only print what worked; the misses are ours, and the next search starts from them.
Whoever finds it holds the most valuable material of the century.
We are searching for it, and we are paid on the way.
$10T+ / yr
of systems built on a superconductor are waiting for it
If it is found elsewhere first, whoever found it still has to make it work in a magnet, a chip and a cable, and that is our platform. Either way, the upside runs through here.
Superconductors are the future of the magnets, the chips, the cables and the grid. That future is capped by one number, how cold, and the company that moves it gets paid by all of them.
Read the narration
For centuries, human progress has been measured by our struggle against resistance. Every machine we build pays a heavy tax to nature in the form of heat. But deep within the quantum world, there exists a hidden loophole, a state of absolute physical perfection.
When the temperature falls, the rules of reality change. Electrons move in perfect unison, without collision, without loss. This is a glimpse into a frictionless future.
It holds the power to reshape our cities, propel us into the air, and contain the fire of stars. The age of fighting nature is over. We are stepping into the era of perfect efficiency.
The only challenge left is discovering the exact material to unlock it.