San Francisco startup Discovered Materials has launched a $1 million prize pool to incentivize AI models to solve five critical unsolved problems in materials science. The company, which is part of the Y Combinator spring 2026 batch, aims to use the challenges to establish benchmarks for AI progress in the field of complex physics.
The five challenges were developed in collaboration with an advisory group of academics from leading global universities. The problems include developing a theory for cuprate superconductivity, creating a numerical method to predict silicon and silicon dioxide interface structures, mapping grain structure to thermal properties, predicting light-induced changes in material behavior, and improving Density Functional Theory for modeling atomic structures.
“Across materials science, there are phenomena we observe but cannot explain,” Stanford professor Felipe Jornada said in a statement. He noted that solving these phenomena would improve the understanding of physics and directly influence the manufacturing of electronics and energy storage.
Discovered Materials reported that preliminary tests involving frontier lab models working autonomously for several days failed to make significant headway on the problems. The company intends for the challenges to serve as "goalposts" to track whether AI models are gaining capability in the domain of materials science.
Co-founder Advaith Sridhar said the company's approach involves defining the verification and the challenges before allowing researchers or models to propose explanatory ideas. Sridhar noted that identifying the correct questions is particularly difficult in materials science because the field lacks the specific, binary definitions found in mathematics.
The startup focuses on developing new materials for semiconductors and currently partners with chip manufacturers on research projects. Sridhar previously indicated the company aims to generate a useful material within a year.
