CuspAI Raises $450M Series B, Launches AI Materials Foundry With Nvidia and Meta

CuspAI closed a $450 million Series B at a $2.6 billion valuation on July 20, 2026, and launched the AI Materials Foundry, a network of 45-plus founding partners that draws Nvidia, Meta, Samsung and ASML into one program to design the physical inputs behind the chip supply chain.

Kleiner Perkins and NEA co-led the round, joined by Bezos Expeditions, Lux Capital, AMD Ventures, Glade Brook Capital Partners, StepStone, Britain's Sovereign AI Venture Fund, Invest-NL and John Doerr. The valuation climbed roughly five times in nine months, moving from $520 million to $2.6 billion, and total capital raised now passes $650 million. For enterprise buyers, this marks a shift: investor money flows toward AI that designs materials, the semiconductors, membranes and catalysts that hardware roadmaps depend on.

Government money joins the round through Britain's Sovereign AI Venture Fund and the Netherlands' Invest-NL, a signal that materials AI now reads as strategic infrastructure. Nations that host chip fabs and battery plants want a stake in the software that designs what those plants produce.

What changed

CuspAI, based in Cambridge with offices in Amsterdam, Berlin, Tokyo, Singapore and the United States, builds a platform called MIRA. Researchers define the properties they want, and MIRA uses generative and agentic AI to generate, evaluate and prioritize candidate materials. The company frames this as inverse design: start from the target performance, then work back to the molecule. The AI Materials Foundry turns that engine into a coalition. Founding partners span the full stack, Nvidia and AMD on compute, Meta on research, Samsung and Hyundai Motor Group on manufacturing, and Applied Materials, Tokyo Electron, Lam Research and ASML across semiconductor equipment. John Giannandrea leads US foundry operations, Abhi Talwalkar sits on the advisory board, and Prof Max Welling and Dr. Chad Edwards co-founded the company.

The use cases run wide. CuspAI aims MIRA at materials that gate three markets: semiconductors, where new dielectrics and photoresists set the ceiling on chip density; clean energy, where better catalysts and battery chemistries decide cost per kilowatt-hour; and advanced manufacturing, where novel compounds shape everything from adhesives to coatings. Henkel's presence in the coalition points to that industrial breadth, and the equipment makers anchor the semiconductor path.

What it means for the vendor map

Materials discovery has run on trial and error for decades, with lab cycles measured in years. CuspAI compresses that loop into a software problem, and the foundry model spreads the cost and the data across 45-plus companies at once. That reshapes several vendor categories. Semiconductor equipment makers, Applied Materials, Tokyo Electron, Lam Research, ASML, gain a shared pipeline for the next generation of etch chemistries, photoresists and dielectric films. Compute vendors Nvidia and AMD lock in demand: every simulation run and every model training cycle consumes their silicon, so materials AI becomes another engine driving GPU sales. Chemical and manufacturing incumbents such as Henkel, Samsung and Hyundai gain early access to designs that competitors outside the coalition will chase later.

The foundry structure also changes the build-versus-buy calculus. A single enterprise attempting in-house materials AI faces steep costs: compute, rare talent in generative chemistry, and years of proprietary data. The coalition amortizes all three across 45-plus members, so a mid-size manufacturer gains capability that once sat behind the walls of the largest chemical majors. That democratization pressures independent R&D labs and specialty-materials suppliers to show why their standalone offering beats shared access to CuspAI's engine.

The competitive signal is sharp. A five-times valuation jump in nine months tells rivals that capital views AI-designed materials as a category with room to compound. Microsoft, Google DeepMind and Meta each run their own materials-science efforts, and DeepMind's GNoME work already showed the scale AI reaches in crystal discovery. CuspAI answers with a commercial coalition rather than a research lab, binding customers and suppliers into one program from day one. Enterprises that source advanced materials, chipmakers, battery producers, industrial manufacturers, face a choice: which coalition to join and how fast.

The 90-day decision

CTOs and heads of R&D at hardware, energy and manufacturing firms should map their materials dependency now. Identify the three to five materials constraints that gate your product roadmap, thermal interface compounds, cathode chemistries, low-k dielectrics, membrane films, and score each on how much a two-to-three-times faster discovery cycle would move your margin or time-to-market. Then open a direct conversation with CuspAI or a rival platform about a scoped pilot. The foundry has 45-plus seats filled already; the firms that secure access this quarter shape the roadmap, while later entrants inherit designs set by others. Treat materials AI as a procurement decision for 2026, and put a named owner on it before the next planning cycle.