Editorial illustration for Discovered Materials Runs AI Agents 24/7 to Hunt for Cooler Chips
AI Agents Hunt Cooler Chips 24/7 for Startup
Advaith Sridhar's cofounder spent his PhD years testing maybe 20 material candidates a day, one careful guess at a time. Now the two of them run thousands of guesses daily, around the clock, using AI agents that never sleep. Their startup, Discovered Materials, closed a $9 million seed round from Lightspeed India Partners after coming out of Y Combinator, with checks from Peak XV Partners and angel investors Paul Graham, Gokul Rajaram, and Thariq Shihipar.
The pitch: chips running AI workloads generate too much heat, and that heat is a big reason data centers eat so much electricity and need elaborate cooling. Sridhar and cofounder Akash Ramdas, who holds a materials science doctorate from Stanford, built a pipeline that pairs Anthropic models in a custom harness with physics simulations they trained themselves, hunting for materials that could make integrated circuits run cooler and more efficiently. Sridhar previously worked on agents at Persona AI and Luma Labs.
The company released hundreds of candidate materials today alongside a new benchmark for tracking how frontier models handle this kind of search. It's chasing the same problem as MatNex, SandboxAQ, and CuspAI, using a method Sridhar described in blunt terms.
Discovered Materials is the latest, with plans to use swarms of AI agents to find new materials that can be used to build more efficient integrated circuits.
Why this matters
Chip cooling is a real bottleneck, and Discovered Materials is betting that agent swarms searching materials space faster than any lab team can beat the problem at its root instead of patching it with better fans and liquid cooling. A $9 million seed from Lightspeed India Partners, straight out of Y Combinator, is a small check by AI infrastructure standards, but the pitch is clear: thousands of automated guesses a day, running continuously, replacing the slow grind of manual materials research. The Material Discovery Bench release matters more than the funding number.
If it holds up to scrutiny, it gives outside researchers a way to check whether these "discoveries" are actually novel and synthesizable, not just plausible-looking outputs from a model that's good at pattern-matching known chemistry. That's the real test for founders chasing this space: agent throughput is cheap to claim and easy to demo. Verified, fabricable materials that lower thermal load in real chips are the only number worth watching here.
Until then, treat the hundreds of new materials as a claim, not a result.
Related Reading
Business & Startups Nordic pilot adds Gemini for Education, NotebookLM to boost AI literacy
Business & Startups Kling launches Video O1, all-in-one model with MVL bridge using transformer
Business & Startups DeepSeek Seeks More Capital Weeks After USD 7B Funding Round
From Archives OpenAI Buys NextSlide to Add AI Presentations to ChatGPT
From Archives Irregular's AI safety test failure could have been caught by external audit
Common Questions Answered
How does Discovered Materials use AI agents to accelerate material discovery compared to traditional methods?
Discovered Materials runs thousands of material candidate guesses daily using AI agents that operate 24/7, compared to the traditional approach of testing maybe 20 candidates per day through careful manual experimentation. This continuous, automated approach allows the startup to explore the materials space far faster than any lab team could manually, dramatically accelerating the discovery process for new chip materials.
What is the primary goal of Discovered Materials' swarms of AI agents?
The primary goal is to find new materials that can be used to build more efficient integrated circuits with better thermal properties. By discovering materials that naturally run cooler, Discovered Materials aims to address chip cooling as a fundamental bottleneck rather than relying on external solutions like improved fans and liquid cooling systems.
Who are the key investors backing Discovered Materials after its Y Combinator exit?
Discovered Materials closed a $9 million seed round led by Lightspeed India Partners, with additional checks from Peak XV Partners and notable angel investors including Paul Graham, Gokul Rajaram, and Thariq Shihipar. This funding came immediately after the startup's graduation from Y Combinator.
Why is chip cooling considered a critical problem that Discovered Materials is addressing?
Chip cooling is a real bottleneck limiting AI workload performance and efficiency in modern processors. Rather than continuing to patch the problem with better cooling infrastructure, Discovered Materials is tackling it at its root by discovering fundamentally cooler materials for building integrated circuits.
Further Reading
- Microsoft Unveils AI Science Platform 'Discovery' That Finds New Materials and Coolants - Microsoft / YouTube
- AI Scientists to Discover Materials for the Semiconductor Industry - Y Combinator
- Diamond coating nearly doubles Chinese AI data centre's cooling efficiency - South China Morning Post
- AI chips are getting hotter. A microfluidics breakthrough could help cool them - Microsoft News
- A Team of scientist and engineers at the University of Texas have created a new thermal material that could Supercharge AI Chip Cooling - Patently Apple
