67,573 magnetic compounds catalogued by artificial intelligence, including 25 unprecedented materials that retain their magnetic properties at high temperatures. This breakthrough from the University of New Hampshire tackles the Achilles’ heel of electric mobility: dependence on Chinese rare earth elements.

The computational approach is revolutionizing materials science. But between digital discovery and industrial application, the path remains fraught with technical and economic hurdles.

The Essentials

  • AI identifies 67,573 magnetic compounds, including 25 new high-temperature materials
  • China controls 90% of rare earth elements and 94% of permanent magnet production
  • New materials could reduce electric motor costs by $500 to $1,500 per vehicle
  • Experimental validation will extend from 2026 to 2030 according to the development timeline

China Locks Down Rare Earth Supply

China processes approximately 90% of the world’s rare earth elements and dominates sintered permanent magnet production. Commercial tensions reinforce this technological stranglehold. In April 2025, Beijing escalated tensions by imposing export license requirements on seven heavy rare earth elements, including dysprosium and terbium.

China controls 94% of permanent magnet production. The country produces over 200,000 tons of magnets annually, which constitute the vast majority of the global supply. In contrast, North America and Europe together manufacture less than 2,000 tons, while Japan and Vietnam contribute approximately 25,000 tons.

This geographic concentration transforms a technical question into a geopolitical issue. “What lithium and cobalt are to batteries, rare earth magnets are to electric vehicle motors,” summarizes Rahim Suleman, CEO of Neo Performance Materials.

Electric Motors Depend on High-Performance Magnets

More than 80% of electric vehicles use permanent magnet motors that each require over 2 kg of permanent magnets. The most high-performing electric motors today all use neodymium-iron-boron magnets.

Efficiency takes priority over cost. For an electric motor based on permanent magnets, the more powerful its magnets, the more efficient, compact, and lightweight the motor can be. They also have their full torque available instantly, can be built much smaller in size, require no maintenance, and have fallen in cost.

Temperature remains the enemy of magnets. Once NdFeB magnets exceed 80 degrees Celsius (176 degrees Fahrenheit), irreversible loss of magnetization occurs. Research follows the Curie temperature, the point at which a material loses its magnetic order, which is crucial for components that must continue operating in hot conditions such as electric vehicle motors and generators.

Artificial Intelligence Maps the Magnetic Unknown

Scientists at the University of New Hampshire unleashed artificial intelligence to dramatically accelerate the hunt for next-generation magnetic materials. By building a massive, searchable database of 67,573 magnetic compounds, including 25 newly recognized materials that remain magnetic even at high temperatures, the team opens the door to cheaper and more sustainable technologies.

At the heart of the discovery lies a new resource: a vast collection of over 67,000 magnetic materials. This database, created using artificial intelligence, includes 25 compounds newly identified that retain their magnetism at high temperatures.

The method drastically accelerates research. Testing every possible combination of elements, which could number in the millions, would take enormous amounts of time and money in a laboratory environment. The AI system narrows down the most viable magnetic compounds for concentrated experimentation, dramatically reducing the research and development timeline compared to traditional trial-and-error laboratory testing.

The models achieve 90% accuracy when classifying materials as ferromagnetic, antiferromagnetic, or non-magnetic, and the best Curie temperature model achieves an R² of 0.87 with a mean absolute error of 56 kelvin. Among the most striking successes was GaFe2Co4Si, a ferromagnetic candidate predicted with a Curie temperature of approximately 1,005 kelvin, or roughly 732 degrees Celsius.

Experimental Validation Challenges Remain Colossal

AI does not replace the laboratory. “AI doesn’t replace the laboratory here; it acts as a map that tells researchers where to look first.” The gap between laboratory discovery and industrial deployment remains real. The 25 newly identified compounds still require extensive experimental validation before they can be incorporated into motor manufacturing. Nevertheless, this approach fundamentally shortens that journey, allowing researchers worldwide to filter and prioritize the most promising candidates computationally rather than through years of practical testing.

The 25 candidates must still be manufactured, measured, and compared to today’s magnets, including how they behave after repeated heating and cooling. Engineers also care about cost, ease of manufacturing, and whether the material can be produced at scale.

The study states that 7 of the high-probability compounds evaluated were later found in the literature with experimentally reported magnetic ordering temperatures, which supports the model, while the remaining 25 remain as new experimental targets.

Development spans several years. The steps include 2026-2027 for laboratory synthesis and detailed characterization of the most promising materials, 2027-2028 for optimization of manufacturing processes and scaling toward pilot production, 2028-2029 for testing in electric motor prototypes and other applications, 2029-2030 for initial commercial deployment in selected applications, and 2030+ for widespread adoption across electric vehicles and clean energy systems.

American Technological Independence at Stake

“By accelerating the discovery of sustainable magnetic materials, we can reduce dependence on rare earth elements, decrease the cost of electric vehicles and renewable energy systems, and strengthen the American manufacturing base,” explains Suman Itani, lead author of the study.

The United States is rebuilding its supply chain. By 2025-2026, the United States and Europe could each have a few thousand tons per year of non-Chinese magnet capacity online. However, even if all these projects meet their targets, China will likely still produce 85-90% of the world’s magnets by the end of the 2020s.

MP Materials was the only company anywhere to control the entire rare earth magnet supply chain, from mine to magnet, meaning independence from Beijing. They hope to soon produce one million magnets per day as they scale up. Litinsky says their first customer, General Motors, will begin using rare earth magnets from MP Materials in vehicles later this year.

Defense motivates investments. The National Defense Authorization Act (NDAA) will ban the use of Chinese-origin NdFeB magnets in most Department of Defense platforms by the end of 2026. This has spurred more direct DoD investment: beyond the MP agreement, the Pentagon has used Defense Production Act funds and grants (over $439 million since 2020 plus recent MP activity) to support extraction, separation, and magnet production projects in the United States.

As the race to electrify transportation intensifies, breakthroughs like that of UNH may prove strategically as significant as they are scientifically elegant, reshaping both the supply chains and geopolitics of the clean energy transition.

AI could democratize materials discovery. The LLM-based approach could also be adapted to other materials domains, including superconducting, thermoelectric, photovoltaic, and ferroelectric materials. But between algorithm and application, industry must still invest massively in experimental validation and industrial scaling.

Sources