Discovered Materials raises $9M to find cooler chip materials

Discovered Materials has closed a $9 million seed round led by Lightspeed India Partners, with participation from Peak XV Partners and angel investors Paul Graham, Gokul Rajaram, and Thariq Shihipar, as the startup emerges from Y Combinator. The company uses swarms of AI agents to search for new materials that can make integrated circuits run more efficiently — chips running AI workloads generate so much heat that cooling them is a major driver of data centers’ electricity use.

Co-founders Advaith Sridhar and Akash Ramdas built a software pipeline that uses Anthropic models in a custom harness to generate candidate material leads, then runs them through physics simulation models the company trained itself to verify whether the candidates are actually promising. Ramdas holds a materials-science doctorate from Stanford; Sridhar previously worked on AI agents at Persona AI and Luma Labs. Sridhar said Ramdas made roughly 20 guesses a day during his PhD, while the company’s agents can now generate thousands of guesses daily, running around the clock in the cloud on research directions Ramdas sets for them.

Discovered Materials released examples of hundreds of new materials this week, along with a “Material Discovery Bench” meant to track how frontier AI models perform on the problem, and says it has already found several materials that match the properties of substances currently used by major chipmakers, though it isn’t sharing details yet. Competitors including MatNex, SandboxAQ, and CuspAI have launched similar AI-driven materials-discovery efforts; Discovered Materials is betting that focusing narrowly on the thermal problems of semiconductor materials will set it apart. Lightspeed partner Hemant Mohapatra, who led the round, said the hardest part is less about generating candidates and more about finding materials whose properties — reducing heat, staying manufacturable, keeping their electrical performance intact — all hold up at once.

No AI-discovered drug or material has yet had a commercial impact at scale: the closest example is Insilico Medicine’s Renterosib, the first drug discovered with generative AI to reach a Phase III clinical trial, while promising materials candidates from other labs haven’t been deployed commercially yet either. Sridhar said the company plans to patent the specific use of any promising materials in GPUs, or the manufacturing process itself, and license them to chipmakers, and hopes to have material worth patenting within the next year — though he said turning candidates into working chips will still require physical lab work that “cannot be sped up.”

Why it matters

The approach reflects a broader bet that AI-driven materials discovery — still unproven commercially outside a handful of early drug and magnet candidates — could ease one of the physical constraints on scaling AI compute, even though converting a promising material into a manufacturable chip still depends on slow, hands-on lab validation.