$450m · Series B · Cleantech · Cambridge, UK
CuspAI, a Cambridge company building an AI platform for industrial materials discovery, has raised $450m in a Series B co-led by Kleiner Perkins and NEA, with significant participation from Bezos Expeditions. The company said the round values it at $2.6bn and will support expansion across the US, Asia-Pacific and Europe as it scales its AI Materials Foundry.
The foundry is the commercial model
Materials discovery is not only a prediction problem. An industrial customer has to define a useful target, identify candidates, simulate their properties, work out how to synthesise them and validate the result in a laboratory. CuspAI says its MIRA platform coordinates that cycle: it generates candidate structures, predicts properties across millions of possibilities, plans synthesis routes around available facilities and feeds experimental outcomes back into the models.
The AI Materials Foundry turns that workflow into a network rather than a single laboratory. CuspAI has assembled more than 45 founding partners spanning industrial users, scientific datasets, compute infrastructure and experimental facilities, including NVIDIA, Meta, Samsung, Hyundai Motor Group, Applied Materials and A*STAR. For customers, the proposition is access to a joined-up discovery process without having to build every model, dataset and laboratory capability internally.
Commercial value will depend on whether the network repeatedly converts promising calculations into materials that meet a partner's performance, cost and production constraints. CuspAI's official launch release says a project with Finnish chemicals group Kemira screened 300tn molecular structures and produced 20 validated candidates for further testing in six months. That is company-reported evidence from one programme, but it shows the metric that matters: validated candidates and shorter experimental cycles, not simply the number of structures a model can generate.
Industrial partners supply both distribution and feedback
The partner network can also become a distribution advantage. Industrial groups enter with defined R&D problems and domain expertise, while lab and data partners provide the infrastructure needed to test proposed materials. Each completed programme can create experimental results that improve later searches, although private partner data and confidentiality requirements will determine how much learning can travel across projects.
CuspAI has announced a multi-year project with Singapore's A*STAR covering semiconductors, carbon capture and advanced electronics. The Series B therefore finances more than model development: it supports the global operating footprint, foundry integrations and laboratory coordination required to move from software predictions to partner-led validation. The execution test is whether CuspAI can preserve that closed loop as the number of customers, scientific domains and facilities grows.
Explore this theme
Why Europe's deep tech problem isn't funding — Martin Schilling & Deep Tech Momentum, hosted on EUVC · 4 March 2026. The piece argues that early corporate customers and workable integration paths are central to commercialising deep technology, the same buyer-led mechanism CuspAI is building into its foundry network.
Sources checked:
CuspAI funding announcement · CuspAI foundry announcement · CuspAI company site · Reuters · Tech.eu Funding Explorer
