Perspectives

CuspAI: A search engine for materials that don’t exist yet

July 20, 2026

Most big leaps in technology come down to a material. Polymers, steel, silicon. The next set – cheaper carbon capture, better batteries, faster chips, cleaner water – is stuck waiting on materials nobody has discovered yet. And the old way of finding them is painfully slow. Materials discovery has traditionally taken 10+ years of trial and error across a chemical space that is bigger than the number of grains of sand on Earth. It’s the ultimate needle in a haystack. 

That’s why we’re excited to co-lead CuspAI‘s Series B.

CuspAI has built an AI system for materials discovery that operates much like a search engine. You describe the properties you need (strength, conductivity, etc) and its agents work backwards from there: searching a proprietary database, generating entirely new candidates that have never existed, then simulating both how they perform and whether they can actually be made. That last part is the hard part. Earlier AI approaches were good at proposing materials no lab could ever synthesize. CuspAI designs for what can be built, which is what turns a clever model into a real product.

And it is already working in the field. For example, with the Finnish chemicals company Kemira, CuspAI went after materials that can remove PFAS from water, the “forever chemicals” now found in the blood of almost everyone alive. Out of a design space of roughly 300 trillion possible structures, the platform narrowed the field to 20 candidate designs now advancing into development, compressing a search that would have taken years into a matter of months. 

The team is why it works. CTO Max Welling did his PhD in physics under Nobel laureate  Gerard ‘t Hooft, a postdoc under Geoffrey Hinton, and co-invented the variational autoencoder, one of the ideas modern generative AI rests on. Chad grew up in a Welsh farming village, was the first in his family to attend university, and went on to help build the quantum computing company Quantinuum. Their advisory board includes Hinton and Yann LeCun, which says plenty about how the field views what they’re building.

Language models reshaped how we work with information. CuspAI is going after atoms, the physical stuff everything is made of. We could not be more excited to partner with Chad, Max, and the team as they pull it off. They’re hiring across the US, Europe, and Asia.