CuspAI

CuspAI develops AI systems for discovering materials with specified properties. Its platform combines generative models, molecular and physics-based simulations, and experimental feedback. Co-founders Chad Edwards, CEO, and Max Welling, CTO, have built the company around industrial partnerships, including a PFAS-removal project with Kemira and a materials-innovation agreement with Hyundai Motor Group.[3][1][4][6]

CuspAI was founded in 2024 by Edwards and Welling. Looking back on its origins in an interview published in February 2026, Welling described concern about climate change, particularly the need to remove carbon dioxide from the atmosphere, as a motivation for starting the company. He also believed the technology was ready to support materials discovery. The company is headquartered in Cambridge, UK.[4][1][6]

Welling described a platform in which a generative component produces candidate materials and a multiscale, multifidelity digital twin screens them. Inexpensive calculations eliminate unsuitable candidates first; progressively more expensive calculations narrow the remaining set before experiments provide feedback. By the February 2026 interview, the platform covered several material classes, with further expansion and connections to high-throughput experimentation under development.[1]

The platform also incorporated agents that searched chemical literature to suggest experiments and agents that orchestrated calculations and experiments. Welling said these components were at varying stages of maturity. He credited Felix Hanke with building the scientific part of the platform and Alessandro De Maria with its platform and MLOps engineering. CuspAI lists them as VP Scientific Engineering and VP Platform Engineering, respectively, and Aron Walsh as Chief Scientific Officer.[1][3]

Automation developed from workflows that scientists initially assembled and ran manually. Welling illustrated the process with a porous material that “actually collapses if you shake it a bit”: that missing constraint prompted the addition of a stability-testing tool. Once the tools and their order were established, an optimizer or agent could begin using them. He described automating smaller tasks, such as selecting calculation settings, before attempting larger workflows.[1]

Welling's intended materials “search engine” would let chemists request candidates while retaining their judgment about which materials were useful. He did not expect chemists and domain experts to leave the loop soon. Automation was also specific to an application: moving to another material class could require different experimental equipment, retrained models and renewed expert input.[1]

Welling said CuspAI invested in an application direction only after finding an industrial partner. Its strategy combined an undisclosed long-term material-development goal with narrower work, such as training a force field or analyzing a particular problem. He preferred deep partnerships that paired CuspAI's tools with companies' domain expertise.[1]

Kemira announced its partnership with CuspAI on July 9, 2025. The initial focus was in-silico molecule and formulation development, with removal of PFAS from water as the first project. Kemira supplied water-treatment chemistry expertise and industrial requirements; CuspAI supplied AI-driven materials design.[4][5]

On November 6, 2025, Hyundai Motor Group announced a signed framework agreement with CuspAI for collaboration across multiple domains. The partnership targeted materials innovation for next-generation mobility, including improvements in efficiency, durability and stability.[6]

In its May 21, 2026 project update, Kemira reported that CuspAI had designed new metal-organic frameworks (MOFs)—porous crystalline materials with tunable structures and chemistry. The brief called for materials targeting GenX, PFBS and PFOS at sub-parts-per-billion concentrations, with requirements for water stability, environmental compatibility, synthesizability and cost effectiveness.[5]

Kemira reported a design space of approximately 300 trillion possible structures, more than 5,000 novel designs with property data, and approximately 20 selected priority candidates. The program reached that discovery stage in six months. Those candidates were advancing to further development and testing; the announcement described a path toward a remediation product rather than a finished filtration product.[5]

In the interview published February 25, 2026, Welling said CuspAI had grown to about 40 people. Kemira's May 2026 update reported more than $200 million raised in Seed and Series A funding. In August 2026, Sovereign AI announced its investment as part of CuspAI's $450 million Series B financing, intended to expand platform development, research capabilities and industrial applications.[1][5][7]

CuspAI's website, as accessed on September 18, 2026, described the AI Materials Foundry as a global network of more than 45 partners, including NVIDIA and Meta, connecting industrial teams and laboratories with data and technology providers. It reported cumulative financing of more than $650 million. The Foundry's partner count describes the ecosystem's membership, rather than a count of paying customers.[3]