AI startup pledges $1 million to solve five intractable materials science puzzles

Discovered Materials, a San Francisco startup utilizing AI to discover novel substances for advanced chips, has launched a $1 million challenge. The initiative calls upon AI models to tackle five significant, unresolved problems within materials science, an area where frontier AI models have yet to achieve deep comprehension of complex physics. Co-founder Advaith Sridhar noted that unlike mathematics, materials science lacks the precise definitions needed for simple yes/no answers. To define these challenges, the company consulted an advisory group of academics from top global universities. Professor Felipe Jornada of Stanford stated that solving these phenomena would significantly advance physics understanding and directly influence future electronics and energy storage. After initial autonomous testing of lab models yielded minimal progress, Sridhar emphasized that the goal is not necessarily finding the solution, but establishing benchmarks. He stated the challenges serve as goalposts to track whether AI is advancing in the domain. Discovered Materials, which was part of Y Combinator’s spring 2026 batch, is focused on semiconductors and has accelerated its internal timeline, expressing greater confidence in finding industrially useful materials in less than a year.