CuspAI Raises $450M to Turn AI Loose on Materials Discovery — and Signs Up Nvidia, Meta and Samsung

The AI-for-science boom has a new flagship, and this time the target isn't drugs — it's materials. CuspAI, a Cambridge, UK startup founded in 2024, has raised a $450 million Series B at a $2.6 billion valuation, up from a $520 million valuation just nine months earlier. The round was co-led by Kleiner Perkins and NEA, with a backer list that reads like a who's-who of the AI era: Jeff Bezos' Bezos Expeditions, AMD Ventures, Lux Capital, Glade Brook, StepStone, legendary VC John Doerr, and Britain's new Sovereign AI Venture Fund — a rare direct bet by the UK government on a homegrown AI champion.

CuspAI's pitch is to do for materials what AI drug-design labs are doing for molecules. Its platform combines AI models, scientific data, computational simulation, and lab validation to design materials with specific target properties — then filters an otherwise astronomically large search space down to the most promising candidates before anything is ever synthesized. The applications it's chasing are squarely in the industrial economy: semiconductors, energy storage, climate technologies, and advanced manufacturing — domains where a single better material can reset an entire supply chain.

The founders explain the ambition. CuspAI was started by Max Welling — a machine-learning pioneer who co-invented the variational autoencoder, a foundational generative architecture, and served as a distinguished scientist at Microsoft Research — and Chad Edwards, a chemist who helped scale quantum-computing firm Quantinuum. It's an unusually credible pairing of deep-learning and hard-science pedigree for a field where most attempts founder on the gap between a model's prediction and what actually comes out of a furnace.

The more strategic move came alongside the money: CuspAI launched an AI Materials Foundry, a coalition of more than 45 technology and industrial organizations — including Nvidia, Meta, Samsung, Hyundai Motor Group, and Lam Research — that pool computing resources, laboratory access, and scientific expertise on CuspAI's platform. That's a telling roster: chipmakers, a social-media giant with its own hardware ambitions, an automaker, and a semiconductor-equipment maker all want a seat at a shared materials-discovery table, and CuspAI is positioning itself as the neutral layer they build on rather than a single company's captive lab.

The raise is a marker for where AI's frontier is heading. The first wave of "AI for science" money chased biology and drug discovery; CuspAI's $450 million — and the caliber of names writing the checks — signals that materials is now seen as the next trillion-dollar problem an AI foundation model can crack. If it works, the payoff is enormous: better batteries, cheaper solar, faster chips, and lower-carbon manufacturing all bottleneck on the same slow, expensive trial-and-error search that CuspAI is betting AI can compress from decades into months. The hard part, as always in the physical world, is proving the model's brilliant candidates survive contact with a lab bench.