A startup focused on finding materials for advanced chips is putting $1 million in prizes behind five unsolved problems in materials science. Discovered Materials hopes the challenges will give researchers a way to measure whether AI systems are making meaningful progress in a field where the questions can be difficult to define and verify.
Five problems to measure progress
Discovered Materials, a San Francisco startup that was part of Y Combinator’s spring 2026 batch, recruited an advisory group of academics at leading global universities to identify five open problems. The company is offering a total prize pool of $1 million for solutions.
The effort addresses a measurement challenge. In mathematics, a problem may have a precise definition and a yes-or-no answer. Materials science deals with complex physical behavior, and the company’s co-founder Advaith Sridhar said that makes it harder to state clearly what a model must solve.
There is no direct equivalent in materials science to the Millennium Prize Problems. The company’s approach is to ask professors to help define important questions, then use attempts to solve them as indicators of progress in the domain.
AI models have not cracked the challenges
Before opening the challenges, Discovered Materials tested frontier AI models on them. The company gave AI agents internet access and other resources, allowing them to work autonomously for several days on each problem. The systems made little headway.
That result describes those attempts, rather than setting a limit on what AI might do next. More computing power or a newer model could change the outcome. Sridhar said the company is not concerned about whether a solution comes from Discovered Materials, another company, a professor, or a frontier AI lab.
Instead, the value of the prize challenges is in the questions themselves. Tracking attempts and solutions could show whether models are improving at materials science, while any successful explanation could deepen understanding of how materials behave.
Why the questions matter
Materials science connects fundamental physics to practical technologies. Felipe Jornada, a Stanford professor who helped develop the challenges, said in a statement that explaining observed but poorly understood phenomena could shape how people build electronics, store energy, and manufacture future materials.
The challenges also reflect a change in how scientific work might be organized. Rather than waiting for researchers or AI systems to produce ideas and then deciding how to assess them, the company is establishing questions and verification goals first. The professors helped identify which questions are worth asking in a difficult field.
That approach makes the challenge useful even before it produces a winning answer. A clearly framed problem gives researchers and model developers a shared target, and the results of attempts can offer a way to assess progress over time.
From research targets to semiconductor materials
Discovered Materials’ main focus is developing new materials for semiconductors. The startup works with large chip companies on research projects and also funds its own scientific efforts.
When the company announced its seed round in August, Sridhar told TechCrunch he hoped it could generate a useful material within a year. He now says those timelines have accelerated and that he is more confident the company can find industrially useful materials in less than a year.
The $1 million challenge and the company’s semiconductor work are related by a broader aim: using AI to improve understanding of materials and help identify substances with practical uses. The five open problems provide a public way to watch whether AI systems are advancing toward that aim.