AI-powered chips generate significant amounts of heat, adding to the energy and cooling demands of data centres. Startup Discovered Materials is using AI to search for new materials that could help make semiconductor components more efficient.

The company recently raised $9 million in a seed round led by Lightspeed India Partners after graduating from Y Combinator. Peak XV Partners and angel investors including Paul Graham, Gokul Rajaram and Thariq Shihipar also participated.

Follow THE FUTURE on LinkedIn, Facebook, Instagram, X and Telegram

Using AI To Search For New Materials

Founders Advaith Sridhar and Akash Ramdas have built a software pipeline that combines Anthropic models with physics models developed by the company. AI agents generate potential materials, while simulations assess whether those candidates could have useful properties.

Ramdas, who holds a doctorate in materials science from Stanford, previously spent his PhD research making around 20 material predictions a day. Discovered Materials says its AI agents can now generate thousands of candidates daily by running continuously in the cloud.

The company has also released examples of hundreds of new materials and introduced its Material Discovery Bench, designed to evaluate how advanced AI models approach materials research.

Finding A Material Is Only The First Step

Discovered Materials is focusing specifically on the thermal challenges facing semiconductor materials. The startup says it has already identified several materials with properties similar to those used by major chipmakers, although it has not disclosed further details.

Finding a promising candidate, however, does not guarantee that it can be used in a chip. A material might reduce heat generation but prove difficult to manufacture, or have electrical properties that make it unsuitable.

“A material is only useful in the real world if all of them converge at once,” said Hemant Mohapatra, the Lightspeed partner who led the funding round.

Mohapatra expects AI-based material prediction to become increasingly commoditised as models improve. He sees Discovered Materials’ advantage in Ramdas’ expertise and the company’s ability to test candidates experimentally.

Commercialisation Remains A Challenge

When the company identifies valuable materials, it plans to seek patents covering their use in GPUs or the processes needed to manufacture chips with them, then license the technology to chipmakers. Sridhar hopes the startup will have materials worth patenting within the next year.

AI-driven materials discovery, however, has yet to produce major commercial breakthroughs at scale. While companies have identified promising candidates, including rare-earth-free magnets and new semiconductor materials, widespread commercial deployment remains limited.

For Discovered Materials, that highlights a central challenge: generating candidates may be getting easier, but determining which ones work and producing them remains difficult.

As Sridhar acknowledged, some parts of the process still require physical laboratory work. “This is the process that cannot be sped up,” he said.