Chips powering AI workloads generate excessive heat, a key driver of the massive electricity consumption and cooling demands in data centers. In a fitting twist, entrepreneurs are now turning to AI to solve the problem it helped create.
Discovered Materials is the latest entrant, leveraging swarms of AI agents to identify new materials for building more efficient integrated circuits. The startup recently closed a $9 million seed round led by Lightspeed India Partners, following its emergence from Y Combinator. The round also included Peak XV Partners and angel investors Paul Graham, Gokul Rajaram, and Thariq Shihipar.
Founded by Advaith Sridhar and Akash Ramdas, the company draws on Ramdas's doctorate in materials science from Stanford and Sridhar's experience building agents at Persona AI and Luma Labs. Their software pipeline uses Anthropic models in a custom harness to generate material candidates, then employs trained foundational physics models to simulate and verify whether these candidates hold real promise.
“During his PhD, Ramdas was making maybe 20 guesses a day,” Sridhar told TechCrunch. “Now, we can make thousands of guesses daily by running these agents 24/7 in the cloud, exploring research directions he provides.”
The company has already released hundreds of new materials and launched its “Material Discovery Bench,” a benchmark designed to gauge how frontier models tackle this challenge.
While competitors like MatNex, SandboxAQ, and CuspAI pursue similar goals, Discovered Materials bets that a sharp focus on semiconductor thermal issues will be key. The startup claims it has already found materials matching the properties of those used by major chipmakers, though it remains tight-lipped on specifics.
One major hurdle is the engineering trade-space: a material that reduces heat generation or improves dissipation might be too difficult to manufacture into a chip, or its electrical properties could suffer. “It’s a bit of playing whack-a-mole with atomic structures,” said Hemant Mohapatra, the Lightspeed partner who led the round. “A material is only useful in the real world if all of them converge at once, which is what makes this a really interesting search problem.”
Mohapatra foresees that predicting novel substances will become commoditized as models improve. What sets Discovered Materials apart, he notes, is Ramdas's deep domain expertise and the ability to run a lab that rapidly experiments and validates candidates—something the founders have already done with several new materials.
When promising candidates emerge, Sridhar says the company will seek patents on their use in GPUs or the manufacturing processes, licensing them to chipmakers. He hopes to have patent-worthy materials within the next year.
Still, despite the excitement, no AI-discovered drug or material has yet achieved commercial impact. The closest case is perhaps Insilico Medicine's Renterosib, the first drug discovered with generative AI to enter clinical trials—a reminder that the path from discovery to market remains long and uncertain.
via TechCrunch AI