Discovered Materials is hunting cooler-chip materials with AI

Startup Discovered Materials raised a $9 million seed round led by Lightspeed India Partners, with participation from Peak XV Partners and angels Paul Graham, Gokul Rajaram, and Thariq Shihipar, TechCrunch reports. The company is hunting for new materials for the semiconductor industry that could help solve one of AI infrastructure's biggest problems — chips overheating under sustained load, and the massive data-center cooling costs that come with it.

Rather than relying on the slow trial-and-error of a physical lab, Discovered Materials has deployed swarms of AI agents that run in the cloud around the clock, generating candidate materials and validating them against physics-based simulation models the team trained. Co-founder Advaith Sridhar described the difference in scale this way:

“[Ramdas] was doing maybe 20 guesses a day during his PhD. We're able to do thousands of guesses a day now by having these agents run 24/7 on the cloud, exploring research directions that he gives them.”

— Advaith Sridhar, Co-founder, Discovered Materials

Lightspeed partner Hemant Mohapatra, who led the round, explained why finding the right material isn't just a matter of brute-force search but a genuine game of whack-a-mole: a material is only useful in the real world once several of its properties — thermal, electrical, manufacturability — converge at once, and improving one often breaks another. Co-founder Akash Ramdas notes that AI agents speed up the hypothesis-generation stage specifically, but the final step — synthesizing and testing materials in an actual wet lab — can't be sped up the same way; that part of the process stays physical and takes time.

• $9 million seed round, led by Lightspeed India Partners
• Swarms of AI agents generate thousands of material hypotheses a day, versus ~20 for a human researcher
• The company went through Y Combinator in 2026 and is based in the Bay Area

Discovered Materials is part of a broader wave of startups applying AI agents not to generate text or code, but to speed up fundamental science and materials development — a field where the traditional experimentation cycle has historically been measured in years, not days.

None of this should be read as personalized investment advice.