Discovered Materials is a startup formed to identify new semiconductor materials that could reduce heat generation or improve heat dissipation in chips, using swarms of AI agents and physics-based simulations. The company recently closed a $9 million seed round led by Lightspeed India Partners after graduating from Y Combinator; additional investors include Peak XV Partners and angel investors Paul Graham, Gokul Rajaram, and Thariq Shihipar.
Founders and technical approach
The company was founded by Advaith Sridhar and Akash Ramdas. Akash Ramdas holds a PhD in materials science from Stanford, while Advaith Sridhar previously worked on agents at Persona AI and Luma Labs. They built a software pipeline that uses Anthropic models in a custom harness to generate material leads, and then applies foundational physics models they trained to run simulations that validate whether candidate materials are promising.
According to Sridhar, Ramdas was making roughly 20 candidate guesses per day during his PhD. By running agents 24/7 in the cloud, the new system can test thousands of candidates per day while exploring research directions provided by the founders.
Outputs and public tooling
Discovered Materials published examples of hundreds of new materials and launched a “Material Discovery Bench” intended to track how frontier models perform on this discovery task. The startup is focusing specifically on thermal issues in semiconductor materials, believing that this narrow, application-driven focus will be its competitive path.
The company says it has already identified several materials that match the properties of materials used by major chipmakers, but it has not disclosed further details.
Engineering and industrial challenges
A key challenge is the engineering trade-space: a material that reduces heat or improves dissipation might nevertheless be impractical to manufacture into chips, or its electrical properties could be compromised. Hemant Mohapatra, the Lightspeed partner who led the round, described the work as “a bit of playing whack-a-mole with atomic structures,” noting that a material only becomes useful if multiple criteria converge simultaneously.
Mohapatra expects prediction of novel substances to become commoditized as models improve; he sees Discovered Materials’ advantage in Ramdas’ domain expertise and the team’s ability to run rapid lab experiments to validate candidates — capabilities the founders say they have already applied to several new materials.
Commercial outlook and bottlenecks
When they find valuable candidates, Sridhar says the startup will seek patents on the use of those materials in GPUs or on processes to manufacture chips from them, then license the IP to chipmakers. He hopes to have patentable materials within the next year.
Despite the progress, real-world commercial impact from AI-discovered drugs and materials remains limited so far. The closest example on the drug side is Insilico Medicine’s Renterosib entering Phase II trials; on the materials side, promising candidates such as MatNex’s rare-earth-free permanent magnets or semiconductor materials explored by Panasonic and Citrine Informatics have not yet seen large-scale industrial deployment.
Mohapatra argues that the bottleneck is less discovery than correctly filtering candidates and synthesizing them. Sridhar likewise acknowledges that much of the work will require wet-lab efforts and actual material fabrication — steps that cannot be significantly accelerated by software alone.
