Discovered Materials
Discovered Materials builds AI agents that accelerate the discovery of new semiconductor chip materials, targeting thermal interface solutions that can dramatically reduce heat and power consumption. In its Y Combinator batch the team simulated, synthesized and experimentally validated thermal interface materials that match the performance of trade‑secret products from the world’s largest chemical companies, and it has released the open‑source Material Discovery Bench, a benchmark created with IBM, IMEC, Stanford and Cambridge to evaluate AI‑driven materials research.
- Artificial Intelligence
- AI Agents
- New Materials
- Semiconductor
Funding
Founders
Product
Problem
AI‑driven semiconductor chips generate high heat fluxes, and the efficiency of heat dissipation is limited by the thermal interface materials currently available. Discovering new materials that can improve thermal performance typically requires years of interdisciplinary research and large capital expenditures, causing many promising concepts to stall before reaching fabrication.
Solution
Discovered Materials offers AI agents that automate the end‑to‑end workflow of thermal interface material discovery, including computational simulation, synthesis planning, and experimental validation. By leveraging these agents, the company reduces the time required to identify viable candidates from months to days. During its Y Combinator program, the platform produced and tested materials that match the performance of long‑standing proprietary products from major chemical firms. In parallel, the company maintains the open‑source Material Discovery Bench, a benchmark suite co‑developed with IBM, IMEC, Stanford, and Cambridge that evaluates AI models on real‑world materials problems. The benchmark includes multiple verifiers for both simulated and experimental results, enabling continuous improvement of discovery algorithms. Together, the AI platform and benchmark accelerate the material innovation cycle for semiconductor thermal management.
Target Audience
Primary customers are semiconductor manufacturers, chip designers, and material R&D teams that require faster development of high‑performance thermal interface solutions, as well as AI researchers focused on materials science who need a realistic benchmark for model evaluation.
Features
- Autonomous AI pipeline that integrates physics‑based simulations, synthesis route planning, and experimental testing for thermal interface materials
- Rapid iteration loop that compresses traditional multi‑disciplinary research timelines from months to days
- Open‑source Material Discovery Bench with curated datasets, evaluation metrics, and verifiers for both computational and laboratory results
- Collaborative development with leading research institutions (IBM, IMEC, Stanford, Cambridge) to ensure benchmark relevance to industry challenges
- Capability to benchmark frontier AI models on real‑world material discovery tasks, supporting both simulation‑only and hybrid simulation‑experiment workflows