Questions
Straight answers about a hard problem.
Every answer here is checkable against something published, and where a claim carries a condition, the condition is in the answer rather than in a footnote.
What is SuperMatics?
- SuperMatics is the first SaaS platform built only for superconductors. It screens millions of candidate materials in software, ranks them against the constraint a device actually has to hit, and sends only the strongest to a bench. The physics is the Modulated Electron Lattice (MEL) framework, licensed exclusively from Hyunsung TNC, where it has been in development since 2006.
What does SuperMatics sell?
- We sell the platform as a subscription. We patent what it finds, and manufacturers pay to use it. Four of the five ways this earns sell into materials that are already in use, so most of the business does not wait on a new discovery to start working.
How long does a superconductor R&D cycle take with SuperMatics?
- An R&D cycle that took years now runs in weeks. A sweep runs in about a week today, iteration cycles included, and the leg that still runs on partner-lab calendars is building and measuring, which is why synthesis is moving in-house at Berkeley. Brief to an industrial-scale sample runs about two months, every step included. The claim is about our own loop, brief to a sample in hand, and never about a customer's shipped product.
How do you know the predictions are any good?
- We validate in both directions, against the field's own archive. Forward, our number goes on the record before the instrument runs, so a prediction is committed while it can still be wrong. Backward, in a closed testing environment: superconductors the field settled decades ago go in with part of what is known withheld from the platform's inputs, and the platform recovers in minutes what the literature took decades to establish, scored against the published record and nothing else. There is no shortlist: every superconductor the published record has settled is in scope, and the suite reruns in full as the models improve. What this establishes is agreement with a record more than a century deep. It is not a precision figure, and we deliberately do not publish one.
Has the physics been independently validated?
- Yes. 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. The measurement was published in Physical Review Letters in 2026, and independent validation means exactly that here: we did not commission it, fund it, or take part in it. One distinction stays sharp on purpose: an independent measurement of the physics underneath the platform is not an independent measurement of a candidate.
Who builds and measures what the software finds?
- Partner labs, in fixed roles, and that division is deliberate. Crystals come from Brookhaven National Laboratory and the Walther-Meissner-Institut, synthesis runs at UC Berkeley and CAN Superconductors, and UIUC and Georgia Tech measure what comes back. We own the search and the learning. The benches belong to people whose results carry their own name.
What is the MEL framework?
- The Modulated Electron Lattice framework is the physics the platform runs on, and it is the reason the search is tractable at all. It has been in development at Hyunsung TNC since 2006 and is licensed exclusively to SuperMatics. Two papers describe it, arXiv:2512.03368 and arXiv:2601.14500, both public, and the effect it relies on has been independently validated by a measurement we had no part in.
What is the highest-temperature superconductor?
- At ambient pressure the durable record is Hg-1223 at 135 K, set in 1993, and it stood for 33 years. In 2026 a group in Houston reported 151 K in pressure-quenched Hg-1223: it is metastable, surviving about three days at 77 K, in flakes tens of microns across. Room temperature at ambient pressure has not been achieved.
Why does a superconductor search need AI at all?
- Because the space is too large to walk. Billions of compositions could exist, and the traditional method is to pick one, spend months making it, and measure. We generate candidates against a brief and cut them on the axes that decide whether a conductor is real: whether the phase forms, whether it holds at one atmosphere, whether it carries current in a field, whether there is a route to make it, what it costs per kA·m, and whether it scales to length. What reaches a bench is what survived all of them.
Why can't a bigger lab just out-compute this?
- Because the bottleneck is the theory, not the compute. The large materials-AI efforts run on density functional theory: DeepMind's GNoME trained on it to predict 2.2 million new crystals, and Periodic Labs raised $300 million to pair models with autonomous synthesis labs, with a superconductor as its stated north star. We take both seriously. But DFT computes ground-state energies, and it is documented to fail for the strongly interacting electrons that make a high-temperature superconductor work. Scaling that pipeline scales its blind spot, and an autonomous lab makes the testing faster without changing which materials the theory can see. The MEL framework was built for exactly this class of problem, in development since 2006 and written from the start to be computed by machines rather than retrofitted onto them. That is why the search is tractable at all, and why our loop gets faster with more compute instead of being stuck behind it.
Do you have a product or revenue today?
- No. There is no shipping product, no working device, and no revenue, and nothing on this site should be read as saying otherwise. We are at seed stage. What exists is the physics, the platform, the papers, the partner network, an independent measurement of the effect underneath it all, and, new this quarter, a superconductor of our own that never existed before: synthesized, measured, and in characterization now. The details stay ours until the filings are done.
How big is the market?
- The beachhead we enter first is $25B+ by 2035, growing around 12% a year, and the systems those materials sit inside are $10T+ / yr. Internal sizing estimate · not a third-party forecast. The more useful way to read it: none of these are markets anybody has to create, because every one of them already runs on a superconductor and is capped by how cold it has to be kept.
What intellectual property does SuperMatics hold?
- 20+ patents issued and pending, with counsel by Wilson Sonsini, and an exclusive license to the MEL framework from Hyunsung TNC. That license is the part worth understanding: the papers are public, but the right to build on the framework commercially is ours alone. The count is filings rather than inventions, and most are pending rather than granted.
What are you not claiming?
- We are not claiming a room-temperature superconductor, a shipped product, or a measured accuracy figure for predicted transition temperatures. What the platform produces are candidates, ranked against a brief. The back-test is retrospective, and an outside measurement of the underlying effect is not an outside measurement of a candidate. We would rather publish the limits than have a physicist find them for us.
Who are the founders and where is the company?
- The people who wrote the physics run the company. James Kim invented the MEL framework and is CTO at Hyunsung TNC; Davis Rens built the platform with him and co-authored both papers; Charlie Moon ran an electrification hardware business for fifteen years and owns the commercial side. SuperMatics, Inc. is a Delaware C-Corporation with offices in Berkeley, California and Suwon, Korea.
We already know our material. Can you help?
- Yes. Most teams do not need a new superconductor; they need the one they already use to survive one more step: a film that has to meet a fabrication process, wiring into a cryostat, a stack that fails where two layers meet. The model runs a bounded question the same way it runs a full search, and the answer comes back in weeks. Anything that then needs a bench runs on partner-lab calendars, which is why synthesis is moving in-house at Berkeley.
Do you model how materials interact?
- Yes, and that is where many briefs start, because where two materials meet is where devices fail. The model treats the pair as one system: superconductor on substrate, superconductor against superconductor, and the junction between them. Send the pairing, not just the material, and the search runs against the constraint the device has to hit.
How do I work with SuperMatics?
- Send us the constraint your device has to hit and we will run the search against it. Every desk has its own address and each one reaches a founder: research@supermatics.io, investors@supermatics.io, partnerships@supermatics.io, careers@supermatics.io, and contact@supermatics.io for anything else.
If your question is not here, ask it directly and a founder will answer it. Get in touch