CuspAI Raises $450M Series B to Launch AI Materials Coalition

CuspAI, a two-year-old British startup that uses generative AI to design new materials, has raised roughly $450 million and launched an industry coalition built on the same technology, closing a large round and standing up a consortium on the same day.

The Series B was led by Kleiner Perkins and NEA, with backing from Jeff Bezos’s investment firm, Bezos Expeditions, according to Bloomberg. The raise more than triples the roughly $130 million CuspAI had previously disclosed and lands the Cambridge company among the best-funded startups in a market where AI valuations keep climbing on the bet that the technology’s next frontier is the physical world.

CuspAI was founded in 2024 by chemist Chad Edwards, its chief executive, and Max Welling, a machine-learning professor who previously held research leadership roles at Microsoft (MSFT ) and Qualcomm. Bezos’s involvement fits a broader push into what investors call “physical AI,” systems that act on the material world rather than generate text; materials discovery, which sits between generative models and manufacturing, is close to that thesis’s center.

From customer roster to consortium

On July 20, 2026, CuspAI also launched the AI Materials Foundry, which Bloomberg reported is a coalition of more than 45 technology companies, industrial firms and research facilities. Members include chipmaker Nvidia (NVDA ), Meta and Hyundai Motor Group, the latter two already CuspAI customers. The group’s stated aim is to pool computing power and scientific resources to build software that helps researchers develop new materials faster and more cheaply than lab-based trial and error allows.

The consortium reframes what CuspAI has been selling. Its commercial story so far has rested on bilateral deals: carbon-capture work with Meta, PFAS-filtering materials with Kemira and sustainable-energy work with Hyundai. A coalition converts that customer list into shared infrastructure — pooled compute, useful when your marquee member is the company that sells the GPUs, plus pooled datasets and a common software layer members can build on. For a startup whose edge rests on proprietary data and synthesis-aware models, turning customers into co-investors in that platform is a competitive move as much as a technical one.

Why chipmaking is the pitch

Bloomberg built its coverage around semiconductors, which need enormous energy and access to rare minerals to manufacture. That framing is deliberate. In its own year-end review, CuspAI said it had expanded work across semiconductors and advanced compute — logic, memory, interconnects, packaging, power delivery and thermal management — arguing that as conventional chip scaling slows, progress increasingly hinges on materials choices. The company counts chip-equipment maker ASML among its customers, and Nvidia chief executive Jensen Huang has publicly named CuspAI as a company he is watching, speaking alongside UK Prime Minister Keir Starmer.

CuspAI’s product is an inverse-design “search engine for materials”: a customer specifies the properties it needs, and the models propose candidate structures meant to be manufacturable, not merely plausible in simulation, then test them in physics-based simulations. The company says the approach yields synthesizable candidates up to ten times faster than traditional methods. Those are CuspAI’s own figures, not independently audited results, the standard caveat for a private firm selling a capability outsiders cannot easily benchmark.

A crowded, well-funded race

CuspAI is far from alone in betting that AI can compress materials discovery from years to months. Periodic Labs, founded by former OpenAI and Google DeepMind staff, raised a $300 million seed led by Andreessen Horowitz at a reported $1 billion valuation; Flagship Pioneering spun out Lila Sciences with a $200 million seed; and rivals from Orbital Materials and XtalPi to established players Schrödinger (SDGR ) and Dassault Systèmes are chasing the same market, alongside molecular-simulation platforms such as Matlantis. Microsoft and Meta have released their own materials models, which could eventually compete with the startups they now supply.

CuspAI’s climb has been fast: a $30 million seed in 2024, a $100 million-plus Series A in September 2025 reported at a $520 million valuation, and now a round the company says values it at $2.6 billion, a roughly fivefold markup in ten months. The talks had surfaced in June 2026 at about that price; the closed round confirmed it. What the number rests on is thinner than its size suggests: commercial contracts a private company does not have to detail, plus a marquee roster of backers and advisers including Nobel laureate Geoffrey Hinton and Turing Award winner Yann LeCun. It is the sort of private valuation now routinely attached to pedigreed AI startups, priced by investor conviction rather than audited revenue, and tested only when customers renew.

The harder question is whether the coalition holds. Members will either treat the AI Materials Foundry as shared infrastructure worth funding, or as a marketing layer over deals they would have signed anyway. The $450 million buys CuspAI time to answer a narrower one: whether its six-month discovery cycles survive contact with manufacturing and convert into repeat, expanding contracts.