New Discovery — AI Agents Evaluate Thousands of Potential Materials for Chip Cooling Daily

Discovered Materials is an early stage startup out of YC that builds a fleet of AI agents modeled after Anthropics to discover new semiconductor materials. These agents can examine thousands of candidate materials per day, compared to ~20/day for an average PhD level scientist. They then rapidly simulate these materials using proprietary physics models.

1The Target: Cooling AI Processors

They’re currently focused on materials for thermal management in AI chips. Too much heat will slow down the chip and require more energy intensive cooling in data centers. They’ve already profiled hundreds of novel materials, and recently announced a public Material Discovery Bench that ranks various top large language model’s performance on materials problems.

2Funding and Team

Discovered Materials recently closed a $9M round led by Lightspeed India Partners, with participation from Peak XV Partners and angels Paul Graham and Gokul Rajaram. The team is led by Akash Ramdas, who has a PhD in materials science from Stanford, and Advait Sridhar, who previously worked on building AI agents at Persona AI and Luma Labs. Building agentic systems to discover physical materials allows us to see the first signs of AI being used to speed up both the software and hardware layers of infrastructure.