When Jeff Bezos and Kleiner Perkins write checks into the same startup, the message to the market is rarely subtle. This week that startup is CuspAI, a two-year-old company out of Cambridge, England, that is using artificial intelligence not to generate text or images but to invent the physical materials the AI era is starving for.
CuspAI is closing a funding round of roughly $400 million that would value the company at about $2.6 billion, according to reports first surfaced by the Financial Times and confirmed across multiple outlets. Term sheets have been signed. The lead backers include Bezos Expeditions, the personal investment vehicle of the Amazon founder, alongside storied Silicon Valley firm Kleiner Perkins.
The number that should stop readers is not the round size but the velocity behind it. CuspAI was valued at roughly $520 million just nine months ago. A $2.6 billion mark represents a near-fivefold jump in under a year, for a company founded in 2024 that has yet to bring a flagship material to mass market.
From seed to unicorn in twenty-four months
CuspAI's fundraising history reads like a compressed history of the current AI boom. The company raised a $30 million seed round in June 2024, followed by a $100 million Series A in September 2025. The reported $400 million round, if it closes as described, would push total funding well past half a billion dollars and vault CuspAI into the upper tier of European AI startups by valuation.
The company is led by Chief Executive Chad Edwards, a chemist and former vice president at quantum-computing firm Quantinuum, and Chief Technology Officer Max Welling, a machine-learning researcher who co-invented the variational autoencoder and previously held senior research roles at Microsoft and Qualcomm. Its advisory board is the kind that opens checkbooks on its own: Nobel laureate Geoffrey Hinton, often called a godfather of modern AI, and Meta chief AI scientist Yann LeCun.
Hinton's involvement was itself news when CuspAI emerged from stealth in 2024, marking one of the first startups the famously cautious researcher chose to back after leaving Google. That two of the most cited names in deep learning are attached to a materials-science company, rather than a chatbot, is a signal in its own right about where the frontier is drifting.
A search engine for matter
CuspAI describes its core technology as a set of "synthesis-aware generative AI models." In plain terms, the platform is built to propose new molecules and materials that can actually be manufactured in a lab, not merely simulated on a screen. Edwards has framed the approach as an inversion of how chemistry has worked for centuries.
"We're flipping the old process on its head and saying, 'Well, if you can put materials or molecules in and get properties out, then why can't you put properties in and get materials and molecules?'" Edwards has said of the company's method.
The applications track directly onto the physical bottlenecks of the AI build-out and the climate transition: materials for carbon capture, advanced cooling and semiconductors for AI chips, batteries, and the removal of PFAS "forever chemicals" from water. CuspAI has already signed partnerships with Meta on carbon capture, Hyundai Motor Group on sustainable energy, and chemicals maker Kemira on PFAS removal.
Welling, the company's technical leader, has repeatedly tied the mission to climate urgency. "Climate change really is something that deserves our full attention at this point," he said in an earlier interview. "It is real, it's important, and we must be putting a much greater effort into this problem today."
Capital pools where AI meets the physical world
The CuspAI round is best read not as a one-off but as a data point in a larger migration of capital. For three years, the money chased models that manipulate language and pixels. Increasingly it is chasing models that manipulate atoms.
The global market for AI in materials discovery was valued at roughly $2 billion in 2025 and is projected to reach nearly $18 billion by 2034, a compound annual growth rate of about 28 percent, according to industry estimates cited in coverage of the deal. The thesis binding investors like Bezos and Kleiner Perkins is straightforward: the constraints on AI's next decade are physical, not just computational. Chips need better cooling. Data centers need cleaner power and carbon removal. All of it depends on materials that do not yet exist, or that would take decades to discover through traditional trial-and-error.
That is the wager CuspAI is selling, that generative AI can compress materials discovery from decades into months. It is an unproven promise. No CuspAI-designed material has yet reshaped a supply chain, and a $2.6 billion valuation prices in a great deal of faith. But the composition of the cap table, a retail-empire founder, a top-tier venture firm, and two of AI's most decorated scientists, suggests the smart money now believes the most valuable thing artificial intelligence can generate may not be words at all. It may be the stuff the world is built from.
"We're flipping the old process on its head and saying, well, if you can put materials or molecules in and get properties out, then why can't you put properties in and get materials and molecules?"-- Chad Edwards, CEO, CuspAI