CuspAI raises $450 million and launches AI Materials Foundry

CuspAI, the materials discovery company co-founded by Amsterdam AI professor Max Welling and CEO Chad Edwards, completed a $450 million Series B on 20 July 2026. The post-money valuation stands at $2.6 billion, compared with $520 million at the Series A in September 2025, a nearly fivefold increase in under a year. Total funding now exceeds $650 million.

Simultaneously, CuspAI launched the AI Materials Foundry, a global network of more than 45 founding partners bringing together data, laboratory capacity, computing power and scientific expertise. The company has offices in Cambridge (headquarters), Amsterdam, Berlin, Tokyo and Singapore, and will also open a London office in 2026.

Dutch Startup published a profile of CuspAI on 10 July 2026, at a time when the company had raised approximately $130 million in funding. The jump to $450 million in a single round at a $2.6 billion valuation illustrates how quickly capital flows into AI-driven materials science have accelerated.

Who is investing and how much

The Series B is led by Kleiner Perkins and NEA. New investors include Bezos Expeditions (Jeff Bezos's investment vehicle), Lux Capital, AMD Ventures, Glade Brook Capital Partners, StepStone, the UK's Sovereign AI Venture Fund, the Dutch Invest-NL, John Doerr and Tru Arrow Partners. Returning investors are Temasek, Prosus, Northzone, Basis Set Ventures, Giant Ventures, Touring Capital and Phoenix Court.

The participation of AMD Ventures and chip equipment manufacturers Applied Materials, Lam Research and Tokyo Electron as foundry partners signals that the semiconductor industry sees value in AI-designed materials as a complement to its own R&D. NVIDIA is providing compute infrastructure to the foundry network; Meta's Fundamental AI Research team is contributing to the Universal Model for Atoms, or UMA, a model being deployed within the network.

Abhi Talwalkar, board member at AMD and chairman of Lam Research, has joined the advisory board. Nobel Prize winner Geoffrey Hinton and Turing Award winner Yann LeCun had previously joined as advisors.

How the AI Materials Foundry works

The core of the network is MIRA, CuspAI's proprietary AI platform for materials discovery. The platform links computational design to physical production: partners contribute specialised labs, measurement equipment and manufacturing expertise, enabling designed materials to actually be produced and tested. This closes the gap between a computationally generated molecule and a substance that works in an industrial process.

In addition to NVIDIA, Meta, Samsung and Hyundai Motor Group, ASML, Hitachi High-Tech, Merck and Henkel also participate in the network. The application areas in which partners are active range from CO2 capture and battery materials to chip materials, water purification and advanced manufacturing processes.

A concrete example of the approach: in a project with Kemira, CuspAI screened 300 trillion potential molecular structures and delivered twenty validated candidates within six months. According to the company, the same process via conventional methods would have taken years. John Giannandrea, former head of AI at Apple and Google, is coordinating the development of the US foundry operations.

Manufacturability as a guiding principle

Many materials discovery platforms focus primarily on generating candidate molecules. CuspAI places the emphasis on materials that can also be produced in practice, and has structured the foundry so that manufacturing knowledge is factored in early in the design process. Industrial partners contribute process expertise, which is combined with MIRA's computational models.

According to CuspAI, this approach is building a flywheel of proprietary training data, customer partnerships and computing power. The more partners participate and the more experiments are conducted, the richer the dataset used to train the models. Foundry partners thus simultaneously serve as customers, data suppliers and validators of the results.

Dutch involvement

Max Welling is a professor at the University of Amsterdam and is known as a co-inventor of variational autoencoders and equivariant neural network architectures. He serves as CTO at CuspAI. The Amsterdam team is based at Lab42 on Amsterdam Science Park.

Invest-NL, the Dutch development finance institution, participated as a new investor in the Series B. This gives the round both a scientific and a financial Dutch connection, in a raise otherwise dominated by large American and international funds.