Jyunmi

They are hidden inside our phones, laptops, speakers, cars, and wind turbines.

They just quietly do their job, but those strong magnets often depend on rare earth elements, names like neodymium and dysprosium.

They're not magic, but they are hard to mine, messy for the environment, and heavily concentrated in just a few countries.

When that supply shakes, prices and projects can shake too.

A team at the University of New Hampshire decided to attack this problem in a different way.

Instead of running only new experiments, they asked, what happens if we let AI read everything we have already done?

They focused on magnetic materials, decades of research, and thousands of papers and data tables, old results buried in PDFs, appendices, and scan charts.

They built an AI system that acts like a tireless reader.

For each paper, it tries to pull out the important details.

What elements are in the sample, how the material was made, how strong the magnet is, and at what temperature it loses its magnetism.

All of those details get turned into structured data, not just text, but a big table that the computer can search and analyze.

From this, the team built the Northeast Materials Database, roughly 67,000 magnetic materials in one place.

From that huge list, the AI helped highlight 25 new compounds that look like strong candidates for high Curie temperature magnets.

Curie temperature is just the point in which a magnet stops being magnetic as it gets hot.

For motors and generators, higher is better.

Now, this work was published in the Journal of Nature Communications, so it's published and gone through the peer review system.

The database is public, so other labs and companies can explore it instead of repeating the same trials by hand.

Even if a few of those 25 candidates turn into real products, the effects are concrete.

Motor makers might be able to use less neodymium.

EV makers could see more stable pricings for key parts.