Inside CuspAI: The Cambridge Startup Using AI to Redesign Molecules, Backed Now by Bezos, Nvidia, and Meta

2026-08-03

From PFAS Removal to Chip Materials, a Company Built Around One Core Bet

CuspAI, a Cambridge-based startup founded in 2024, has spent the past two years pursuing a fairly unconventional thesis in the AI world: rather than building better chatbots or image generators, it's using generative AI and molecular simulation to design entirely new physical materials from scratch. Co-founder and CTO Max Welling, a machine learning researcher at the University of Amsterdam and co-inventor of the variational autoencoder, described the company's mission bluntly in a recent interview: "We are building molecules to remove forever chemicals from water."

That's not just a talking point. Earlier this year, CuspAI partnered with Kemira, a Finnish chemicals company, to design new materials capable of removing PFAS, so-called "forever chemicals," from drinking and process water at trace concentrations. The project explored a design space of roughly 300 trillion possible material structures using CuspAI's MIRA platform, ultimately producing more than 5,000 novel material designs with full property data for three priority PFAS compounds, GenX, PFBS, and PFOS. That list was narrowed down to about 20 priority candidates now moving into further development, a process the companies say took just six months and represents the first commercial partnership to apply generative AI end-to-end to PFAS remediation specifically.

A Fivefold Valuation Jump in Under a Year

CuspAI's momentum has translated into a dramatic funding trajectory. The company started with a $30 million seed round in June 2024, grew to a $100 million Series A by September 2025 at a $520 million valuation, and has now closed a $450 million Series B round that pushes its valuation to $2.6 billion, roughly five times higher than where it stood less than a year earlier. Total funding raised now sits above $580 million.

The round was co-led by Kleiner Perkins and NEA, with what CuspAI describes as significant participation from Jeff Bezos' personal investment vehicle, Bezos Expeditions, which also owns Blue Origin and The Washington Post. Additional backers include AMD Ventures, Britain's Sovereign AI Venture Fund, Greenoaks Capital Partners, Lux Capital, and Glade Brook Capital Partners. Bezos' involvement follows his earlier backing of Prometheus, a separate AI-for-manufacturing venture that raised $12 billion at a $41 billion valuation earlier this year, underscoring his broader bet on AI moving from software into physical engineering and industrial applications.

Why Chip Materials Are Now the Company's Main Focus

Alongside the funding announcement, CuspAI unveiled what it's calling the "AI Materials Foundry," a coalition of more than 45 founding partners spanning the companies that physically build AI infrastructure. Nvidia is providing compute infrastructure for the initiative, while Meta's Fundamental AI Research team is also participating. Other founding members include Samsung, Hyundai Motor Group, AMD, Applied Materials, Tokyo Electron, Lam Research, and Henkel.

CEO and co-founder Chad Edwards said roughly 80 percent of the company's research will now focus specifically on chip materials, with a central goal of reducing dependence on supply-constrained rare metals like ruthenium and iridium, materials that have become bottlenecks in semiconductor manufacturing as demand for AI chips accelerates. "As AI transforms the physical world, new materials will open up new frontiers across semiconductors, energy and advanced manufacturing," Edwards said. Notably, both AMD and Nvidia, direct competitors in the AI chip market, are now aligned around CuspAI's platform, a signal that the industry views the materials constraint as significant enough to justify cross-competitor collaboration.

The Science Behind the Platform

CuspAI's core technology inverts how materials discovery traditionally works. Rather than testing candidate materials one at a time in a lab, a process that can take years, the company's AI platform simulates the performance of enormous numbers of potential material structures computationally first, then narrows that list down to a small set of the most promising candidates for real-world laboratory testing. The company describes its platform as functioning like "a search engine for the material world."

Welling brings a notable pedigree to the effort, having previously served as a Distinguished Scientist and VP at Microsoft Research AI4Science and as VP of Technology at Qualcomm before co-founding CuspAI with Dr. Chad Edwards, a chemist and former VP at Quantinuum. The company's advisory board includes Nobel laureate and Turing Award winner Geoffrey Hinton, Turing Award winner and Meta chief AI scientist Yann LeCun, former ASML President and CTO Martin van den Brink, and former BP CEO Lord John Browne. The company has also hired former Apple and Google executive John Giannandrea to lead its US operations.

What Comes Next

With the fresh capital, CuspAI is opening a new office in Singapore and expanding its teams across the UK, Netherlands, Germany, Japan, and the United States. The broader bet underlying the company's rapid valuation growth is that as AI infrastructure buildout accelerates, the physical materials needed to manufacture chips, batteries, and other hardware will become an increasingly critical bottleneck, one that traditional lab-based materials science moves too slowly to address. Whether CuspAI's AI-driven approach can consistently deliver commercially viable materials at scale remains to be proven, but the scale of industry buy-in, spanning direct competitors like Nvidia and AMD, suggests the underlying problem is one few in the chip industry are willing to ignore.

Sources