SuperMatics Is Building AI to Find the Next Superconductors for Commercial Wire
By Matt Swayne · · 2 min read
Insider Brief
- SuperMatics is developing AI models to identify superconducting materials that can be manufactured into affordable wire for data centers, quantum computing, fusion magnets and other applications.
- Its AI models and agents predict material properties, generate synthesis recipes and simulate finished wires or devices, using experimental results, including failures, to improve predictions.
- Researchers at UC Berkeley and Georgia Tech make and test candidate materials, connecting the company’s AI development process with laboratory experiments.
PRESS RELEASE — Since 1911, scientists have discovered thousands of superconductors, materials that carry electricity with zero resistance. Only a handful have ever been made into commercial wire. SuperMatics, a Berkeley company building AI exclusively for superconductors, is targeting that gap and will present its approach on November 5 at the Plug and Play Silicon Valley November Summit in Sunnyvale, California.
The gap is getting expensive. The International Energy Agency expects data center electricity use to more than double by 2030, and Microsoft is exploring superconducting power delivery for data centers. Fusion developers build magnets from high-temperature superconducting tape, which remains costly and scarce.
Records alone will not close it. In 2026, University of Houston researchers raised the ambient-pressure superconductivity record for the first time since 1993, but for power and magnets, a material matters only if it can become tough, affordable wire. Because standard simulations cannot reliably predict high-temperature superconductivity at everyday pressures, most teams still rely on trial and error.
SuperMatics designs for manufacturing from the start. Its multimodal foundation models and AI agents, built on physics including the Modulated Electron Lattice (MEL) framework, predict material properties, write the synthesis recipe, and simulate the finished wire or device. Candidates are made and tested with researchers at UC Berkeley and Georgia Tech, and every result, failures included, retrains the models.
“Records make headlines, but wires change industries,” said Davis Rens, Co-founder of SuperMatics and a former quantum engineer at Rigetti Computing. “The next breakthrough is not the highest temperature on paper. It is the superconductor someone can actually make, and that is what our AI is built to find.”
SuperMatics is in talks with global materials and technology companies on joint programs for existing high-temperature superconductors while building its own new materials pipeline. CEO Charlie Moon leads the company with co-founder Dr. Jaehwahn Kim, Hyunsung T&C’s CTO and a 25-year superconductivity veteran.
