Discovered Materials puts $1M in prizes on five materials problems AI has not cracked

What happened

Discovered Materials, a San Francisco startup founded only months ago that applies AI models to the search for new chip-building substances, has launched a prize pool totaling $1 million for solutions to five open problems in materials science, TechCrunch reported. Each accepted solution earns $200,000. Y Combinator took the company into its spring 2026 cohort.

An advisory group of academics from leading universities helped define the problems. They include a theory of cuprate superconductivity and a theory of why light exposure alters material behavior. The others call for numerical methods to predict silicon and silicon dioxide interface structures, to link grain structure to thermal properties, and to improve Density Functional Theory.

Before going public, the company let frontier lab models attack the problems as autonomous agents with internet access, spending several days on each with little progress. Co-founder Advaith Sridhar said it does not matter who reaches a solution. He said the aim is to set goalposts that show whether models are improving in the field.

Context

Frontier AI models have solved notable math problems and advanced in the life sciences but have made less headway on the physics of advanced materials. Discovered Materials announced a seed round in August and works with large chip companies on research while funding its own projects. Sridhar said he now expects to find a material useful to industry in under a year, faster than he projected then.