AI
Discovered Materials offers $1000000 in prizes for AI to solve five major materials science challenges
San Francisco startup Discovered Materials announced it is putting up $1000000 in prize money for AI models to solve five open problems in materials science, including the theory of cuprate superconductivity and prediction of the silicon–silicon dioxide interface structure. The company hopes to use clearly defined problems to measure actual AI progress in materials.
San Francisco startup Discovered Materials, founded months ago and part of Y Combinator’s Spring 2026 batch, announced a $1000000 challenge to reward AI models capable of solving five key materials science problems. The company uses AI to find new substances that can be used in advanced chips.
Co-founder Advaith Sridhar said materials science problems are not as easy as mathematics problems to define precisely or answer with right-or-wrong outcomes, so the team worked with an advisory group of scholars from top universities worldwide to identify five open problems that can answer whether AI is actually making progress in the field.
The five problems are: a theory explaining the zero-resistance phenomenon of cuprates at extremely low temperatures; numerical methods for predicting the structure of the interface between silicon and silicon dioxide; numerical methods for mapping grain structure to thermal properties in order to design better heat-dissipation materials; a theory explaining why light changes material behavior, especially behavior related to superconductivity; and numerical methods for improving density functional theory, a key technique for modeling atomic and molecular structures.
Stanford professor Felipe Jornada, who helped formulate the challenge, said there are many observed but unexplained phenomena in materials science, and the answers would affect electronic devices, energy storage, and the future manufacturing of materials. Discovered Materials previously let frontier laboratory models work autonomously for several days with internet access and other resources, but progress was limited. Sridhar said whether the answer is found by the company, another team, a professor, or a frontier lab, the purpose of the challenge is to set goals for measuring progress.