Discovered Materials, a San Francisco startup from Y Combinator's spring 2026 batch that uses AI models to find new substances for advanced chips, announced a $1 million challenge for AI systems to solve five critical open problems in materials science.

Startup Fortune reports the prize is $200,000 per problem, and that the company was founded by IIT Madras alumni Advaith Sridhar and Akash Ramdas, raising a $9 million seed round in August led by Lightspeed India Partners with Y Combinator, Peak XV Partners and Paul Graham participating.

Co-founder Advaith Sridhar told TechCrunch the difficulty was specifying problems precisely: "It's not like math, where you have a very specific definition and can get to a yes or no."

The company recruited an advisory group of academics, including Stanford professor Felipe Jornada, to spell out the challenges. TechCrunch; Startup Fortune.

Materials science has no equivalent of the Millennium Prize Problems, so the advisory panel had to define the goalposts themselves.

The five challenges: a theory explaining cuprate superconductivity, the ability of certain copper-oxide materials to conduct with zero resistance at ultra-low temperatures; a numerical method to predict the structures of silicon-silicon dioxide interfaces, a combination found in almost all semiconductors; a method mapping grain structure to thermal properties to design better heat-dissipation materials; a theory predicting why shining light on materials can change their behavior, including superconductivity; and a numerical improvement to Density Functional Theory, the key technique for modeling atoms and molecules.

Stanford's Jornada said the answers would "directly shape how we build electronics, store energy, and manufacture the materials of the future." TechCrunch; Startup Fortune.

What gives the challenge teeth is the test that preceded it.

Before announcing the prizes, Discovered Materials turned frontier lab models loose with internet access and other resources, letting AI agents work autonomously for days on each problem — and the models spent several days without making much headway, according to TechCrunch.

Startup Fortune frames this as a bet most AI labs won't make out loud: that today's frontier models still can't crack open physics problems without human help, and notes the contrast with Google DeepMind's GNoME, which generated 2.2 million candidate materials but was a prediction engine rather than a system asked to explain a specific unresolved question.

Sridhar told TechCrunch it is not a concern whether the answer comes from his company, a professor, or a frontier lab — the challenges exist to serve as goalposts showing "whether the models are climbing in this domain."

Startup Fortune notes the challenge previews a funding model that skips government grant cycles entirely: a seed-backed startup can simply post the prize and open a leaderboard, and the $1 million pool is a public, falsifiable marker next to what big labs spend on training runs.

The company's core business is finding new materials for semiconductors, partnering with chip giants while funding its own research.

When Discovered Materials announced its seed round in August, Sridhar told TechCrunch he hoped to generate a useful material within a year — and he now says those timelines have accelerated, expressing more confidence that industrially useful materials will be found in less than a year.

Whether any lab claims one of the five prizes will say more about AI for science than another benchmark score, and if none do, Startup Fortune argues that is a real data point the startup will have paid $1 million to generate.