CuspAI was founded in Cambridge in 2024 by two people: Dr. Chad Edwards, who previously helped scale a quantum computing company, and Professor Max Welling, a machine learning researcher who was VP of technology at Qualcomm and a distinguished scientist at Microsoft Research. Their shared observation was straightforward: the physical bottleneck to the next wave of industrial progress is not software or processing power, it is materials.
Chips require rare metals like iridium, ruthenium, germanium, and gallium. China controls the refining of many of them, and it has already tightened export restrictions multiple times since 2023. New mines typically take 15 to 20 years from discovery to production. That timeline does not match the speed at which demand for chips, batteries, and clean energy infrastructure is growing.
What CuspAI does is flip the search. Instead of starting with what materials exist and testing their properties, the platform starts with the properties you need and works backwards to suggest chemical compositions that could deliver them. It then runs simulations to test performance and manufacturability before anything goes near a lab. The company describes this as "synthesis-aware" design: materials the AI suggests can actually be built, not just modeled.
In a six-month project, CuspAI's system narrowed 300 trillion possible material structures down to 20 candidate designs for water treatment. That kind of compression, from astronomical possibility space to a short list of buildable options, is what makes this commercially interesting beyond the semiconductor sector.
The $450 million Series B round values CuspAI at $2.6 billion. That is five times its valuation from September 2025. Backers now include Kleiner Perkins, NEA, Jeff Bezos's personal family office, AMD Ventures, Lux Capital, Singapore's sovereign wealth fund Temasek, and the UK government's sovereign AI fund. Alongside the raise, CuspAI launched the AI Materials Foundry: a coalition of more than 48 companies and research facilities including Nvidia, Meta, Samsung, Hyundai, and Lam Research, pooling data and computing resources.
Bezos's involvement is worth noting separately. He is also the co-CEO of Prometheus, his own physical AI venture that raised $12 billion at a $41 billion valuation and focuses on AI for design and manufacturing. His investment in CuspAI is personal, not through Prometheus, but the pattern is clear: he is placing capital across the bet that AI's most significant near-term impact will be in the physical world, not in language.
The competition in this space is real. Orbital Materials, co-founded by people from DeepMind, and XtalPi, valued at $2.5 billion, are also working on AI-driven materials discovery. Established simulation software companies like Schrödinger and Dassault Systèmes are active in adjacent areas. Periodic Labs, founded by former OpenAI and Google DeepMind staff, raised $200 million in seed funding at a $1 billion valuation.
For business operators outside the tech sector, the direct implication is this: the supply chain constraints that are already affecting semiconductor availability, battery production, and water treatment costs are unlikely to ease through traditional mining alone. Companies that depend on these inputs should watch whether AI-designed material substitutes start reaching commercial production, because that is the path that could actually move the needle on costs and availability within this decade.
CuspAI's own timeline from seed round to $2.6 billion valuation in under two years reflects how seriously large industrial players, not just venture investors, are treating this problem. Nvidia, Hyundai, and Samsung are not joining a coalition for press coverage; they are customers and partners with materials problems they need solved.