Discovered Materials Bags $9M Seed to Deploy AI Agents in Hunt for Novel Semiconductors

Discovered Materials has secured $9 million in seed funding to build AI agents that discover new materials for semiconductor chips. The round was led by Lightspeed, with participation from Y Combinator and Peak XV, and included angels such as Paul Graham, Gokul Rajaram, and Thariq Shihipar.

The startup is tackling a bottleneck that has stalled computing progress. AI chips face severe heat constraints, with GPUs handling heat fluxes of roughly 140 W/cm²—higher than a space shuttle nose cone re-entering Earth's atmosphere. The materials used to build chips directly determine how much heat is produced and how fast it dissipates. Existing materials are at their limit, and while new materials could improve chip performance by orders of magnitude, the path from discovery to fabrication typically takes years and hundreds of millions of dollars. Most candidates fail in the gap between scientific experiment and factory production.

Discovered Materials says its AI agents compress months of interdisciplinary scientific work into days. During a three-month Y Combinator batch, the team simulated, synthesized, and tested thermal interface materials that matched the performance of products the world's largest chemical companies have protected as trade secrets for over two decades.

Alongside the funding announcement, the company released Material Discovery Bench, an open-source benchmark for AI-driven materials discovery. It was built with input from experts at IBM, IMEC, Stanford, and Cambridge, and includes verifiers to grade frontier model performance on real-world materials problems.

The company was founded by Advaith Sridhar, who studied AI at Carnegie Mellon and previously worked as a research engineer at Persona AI and Luma Labs, and Akash Ramdas, who holds a PhD in Materials Science from Stanford and spent 11 years researching new materials for semiconductor chips. The two met over a decade ago at IIT-Madras.

Discovered Materials argues that the engine of technological progress—advances in compute—stalled in the mid-2010s. Its stated goal is to restart it by closing the power-efficiency gap between chips and the human brain, which it estimates is at least 10,000x.