AI Just Discovered Two New Superconductors - and a Faster Way to Find Thousands More

For a century, discovering a new superconductor was mostly a matter of luck. An international team has just shown it does not have to be. Researchers led by Finland's Aalto University used machine learning to predict which materials might be superconductors - and then proved it, as collaborators at Rice University synthesized the top candidates and confirmed two brand-new superconductors in the lab. The two materials themselves are modest. The method behind them may be the real breakthrough: a systematic, AI-guided way to hunt through the near-infinite space of possible materials for the field's ultimate prize - a superconductor that works at room temperature.

At a glance

What: AI-guided discovery of two new superconductors, YRu₃B₂ and LuRu₃B₂
Who: Aalto University (Finland - theory + machine learning) and Rice University (synthesis + experiment), within the international SuperC consortium
Method: machine-learning pre-screening → targeted quantum-physics calculations → lab synthesis and confirmation
Critical temperatures: 0.81 K and 0.95 K - very low, so the value is the method, not these particular materials
The prize: a scalable search that could one day screen billions of candidates toward a room-temperature superconductor
Published: Physical Review Research, June 2026

1. What the AI Found

The two new superconductors are YRu₃B₂ and LuRu₃B₂ - compounds built around a kagome lattice, a pattern of interlaced triangles named after a traditional Japanese basket weave. That geometry is not decorative. In a kagome lattice, quantum interference flattens the electrons' energy bands, so many electrons crowd into nearly the same energy and interact strongly - exactly the kind of condition that can tip a material into superconductivity. Both compounds showed nearly 100% superconducting volume fractions in the lab, confirmed by magnetization and specific-heat measurements - meaning the whole sample, not just a sliver, went superconducting.

2. How Machine Learning Changes the Hunt

Superconductivity - the ability to carry electricity with exactly zero resistance - was discovered in 1911, and physicists have catalogued roughly 7,000 superconductors since. Almost all of them were stumbled upon. “Over the decades researchers have recognized over 7,000 superconductors, but mostly serendipitously,” says Aalto physics professor Paivi Torma, who leads the effort.

The new workflow replaces luck with a funnel. First, a machine-learning model rapidly pre-screens an enormous list of candidate compounds, scoring which are most likely to superconduct. Only the most promising survivors go through slow, expensive, first-principles quantum-physics calculations. The handful that pass are handed to experimentalists to actually make and measure. “Our method uses machine-learning-based pre-screening followed by targeted calculations on the promising candidates,” Torma explains - and because the cheap first step does the heavy filtering, the search can scale enormously. “With machine learning, we may be able to push the number of materials we can process into the billions.”

The honest caveat

These two materials are not the answer - and the team is the first to say so. YRu₃B₂ and LuRu₃B₂ only superconduct below about 1 kelvin (0.81 K and 0.95 K) - colder than deep space, and nowhere near useful for everyday technology. They are a proof of concept. The achievement is that an AI predicted them and the lab confirmed them, validating a pipeline that can now be pointed at millions more candidates.

3. Why Superconductors Are Worth the Hunt

A superconductor carries current with no energy lost to resistance at all. The catch has always been temperature: known superconductors only work when chilled, often to hundreds of degrees below zero, which makes them expensive and hard to deploy. A material that superconducted at room temperature would, in Torma's words, “forever change the way we consume energy.” The payoff list is long:

Power grids that move electricity across continents with no losses
MRI and other medical scanners made smaller and cheaper (they already rely on superconducting magnets)
Faster, more efficient maglev transport
Stronger magnets for fusion reactors, a key ingredient in the push for clean power
More stable and scalable quantum computers

4. A Global Race With a Deadline

This is the first discovery from SuperC, described as the first coordinated global collaboration dedicated to finding new superconductors. Launched in 2023, the consortium has set itself a deliberately bold target: identify a room-temperature superconductor by 2033. The project spans Aalto University, Rice University, Princeton University, Ruhr University Bochum, and the Donostia International Physics Center in Spain, pairing machine-learning and quantum theorists with the experimental groups that turn predictions into real crystals. Two materials in, the AI-guided approach has its first proof that it works.

What We Still Don't Know

Whether the same pipeline can find superconductors at much higher, more useful temperatures - or mostly turns up more low-temperature ones.
How far the “billions of candidates” vision scales in practice, and how quickly experiment can keep up with the AI's predictions.
Whether room-temperature superconductivity at ordinary pressure is achievable at all - the deepest open question in the field.

Sources

Aalto University: Researchers identify new superconductors, unlocking process that could yield thousands more
ScienceDaily: AI just supercharged the race to find room-temperature superconductors · Phys.org coverage
Paper: Machine Learning-Guided Discovery of Kagome Superconductors YRu₃B₂ and LuRu₃B₂, Physical Review Research (2026), DOI 10.1103/lpqj-7hyg (preprint arXiv:2512.16945)

Curated by Jerry Cards - jerrycards.com. We research the week's most consequential tech, science, and business news so you don't have to. More at jerrycards.com/news.