Chips running AI workloads are too hot: This is one reason data centers consume so much power and require cooling systems. And inevitably, entrepreneurs will turn to AI to solve the problems it poses.
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. The company recently announced that it has exited Y Combinator and closed a $9 million seed round from Lightspeed India Partners with investment from Peak XV Partners and angel investors Paul Graham, Gokul Rajaram, and Thariq Shihipar.
Founders Advaith Sridhar and Akash Ramdas teamed up to launch the company, drawing on Ramdas’s PhD in materials science from Stanford University and Sridhar’s experience working on Persona AI and agents at Luma Labs.
The two created a software pipeline that uses a human body model in a custom harness to generate material leads. The trained underlying physical model is then used to run simulations to verify whether the candidate material is actually interesting.
“[Ramdas]was probably making 20 guesses a day during his PhD,” Sridhar told TechCrunch. “By running these agents 24/7 on the cloud and examining the research instructions he gives us, we can now make thousands of inferences a day.”
Discovered Materials today released the Materials Discovery Bench, designed to track hundreds of new material examples and how state-of-the-art models address this challenge.
Companies like MatNex, SandboxAQ, and CuspAI have all launched similar efforts, but Discovered Materials is betting that focusing on thermal issues in semiconductor materials is the path to success. The company says it has already discovered some materials that match the properties of existing materials used by major chipmakers, but it cannot share those details.
One of the challenges is the engineering trade space. If a material is found that has the potential to reduce heat generation or improve its dissipation, it may be too difficult to actually manufacture chips from that material, or its electrical properties may be compromised.
“This is like playing whack-a-mole with atomic structures,” Hemant Mohapatra, a Lightspeed partner who led the round, told TechCrunch. “Materials are only useful in the real world if they all converge at once. That’s what makes this exploration problem so interesting.”
Mohapatra predicts that as models continue to improve, the business of predicting new substances will become commoditized. What sets Discovered Materials apart is Ramdas’ deep experience in the field and ability to run a lab that can rapidly experiment and validate candidates. This is something the two founders are already doing with some new materials, he says.
Once it finds a worthy candidate, Sridhar said, the company intends to patent the use of the material in GPUs, or the process of making chips from it, and license it to chip makers. He hopes to see patent-worthy new materials in the next year.
But despite all this excitement, we have yet to see drugs and materials discovered through AI have real commercial impact. The closest one is probably Insilico Medicine’s Renterosib. This is the first drug discovered with a generated AI to move into Phase III clinical trials. On the materials front, promising candidates include rare earth-free permanent magnets from MatNex and new semiconductor materials developed by Panasonic and Citrin Informatics. However, these have not yet been introduced commercially on a large scale.
As AI continues to advance, these technologies may be coming into their own, but this is one reason why Mohapatra says he doesn’t think finding more candidates will hinder AI materials science. Rather, “filtering and synthesizing them correctly is the bottleneck.”
Sridhar believes that Discovered Materials’ unique data and expertise will help the startup compete with well-funded frontier labs, but acknowledged that the reality is that “a lot of it will also involve actually going into the wet lab and making things, and that’s a process that can’t be sped up.”
Correction: This article originally reported that Renterosib entered Phase II clinical trials. In fact, it started Phase III clinical trials in July.
If you buy through links in our articles, we may earn a small commission. This does not affect editorial independence.
