CuspAI: $450M Series B at $2.6B Valuation for AI Materials Discovery

CuspAI raised $450 million at a $2.6 billion valuation, co-led by Kleiner Perkins and NEA with Jeff Bezos's family office writing a check alongside John Doerr. That's the short answer. The longer answer is more interesting.

A Cambridge, UK company that helps discover new materials — the kind that go into chips, batteries, and industrial coatings — just went from a $520 million valuation to $2.6 billion in ten months, without a product category most people have heard of. This isn't a chatbot, an agent framework, or a coding tool. It's AI applied to one of the slowest, most trial-and-error-heavy fields in industrial science, and the capital markets just priced it like the next foundation model company.

CuspAI $450M Series B: Round Terms and Lead Investors

CuspAI closed a $450 million Series B on July 21, 2026, co-led by Kleiner Perkins and NEA at a $2.6 billion post-money valuation. Bezos Expeditions, Lux Capital, AMD Ventures, Glade Brook Capital Partners, Tru Arrow Partners, StepStone, and Britain's Sovereign AI Venture Fund all participated, alongside angel investor John Doerr.

What CuspAI actually builds

CuspAI's pitch is that materials science still runs on a discovery process that hasn't fundamentally changed in decades: a researcher hypothesizes a compound, synthesizes it, tests it, and iterates — often across years and thousands of failed candidates — before landing on something that works. CuspAI's models simulate the mechanical, thermal, and electronic properties of candidate materials computationally first, narrowing an enormous search space down to the handful of candidates actually worth synthesizing in a lab.

The commercial product is called the AI Materials Foundry, and CuspAI says it now has more than 45 partners feeding real-world problems and validation data into the platform, including NVIDIA, Meta, Samsung, Hyundai Motor Group, Applied Materials, Tokyo Electron, and Lam Research. That partner list is the tell: this is squarely aimed at the semiconductor and advanced-manufacturing supply chain, where a single better dielectric, coating, or thermal interface material can be worth billions in yield and performance gains across an entire fab.

From $520M to $2.6B in ten months

CuspAI's funding history is a clean case study in how fast AI-for-science valuations are compounding right now. The company raised a $30 million seed in June 2024, then a Series A of just over $100 million in September 2025 — co-led by NEA and Temasek, with NVIDIA's NVentures and Samsung Ventures both already in the cap table — that valued the company at $520 million. Ten months later, the Series B put a $2.6 billion price tag on the same business.

Total capital raised across all three rounds is now roughly $580 million. That's a company that's raised more money in the last ten months than most Series C SaaS companies raise across an entire lifecycle, backing a product category — AI-designed materials — that barely existed as a venture thesis three years ago.

Why Bezos and Kleiner Perkins are both in this deal

Jeff Bezos's family office, Bezos Expeditions, has been on an aggressive run backing physical-world AI companies through 2026 — it co-led Prometheus's $12 billion round at a $41 billion valuation and put capital into Flourish's $500 million round earlier this year. CuspAI fits the same thesis: AI applied to atoms, not just tokens, in categories with genuine physical-world moats rather than a thin wrapper around a foundation model API.

For Kleiner Perkins and John Doerr specifically, materials discovery is also a chip-supply-chain bet. As the AI buildout runs into physical bottlenecks — advanced packaging, thermal management, next-generation dielectrics — the firms funding the compute layer have obvious reasons to also fund the materials layer that determines how fast that compute layer can actually scale.

CuspAI vs. the AI-for-science funding wave, side by side

CuspAI isn't raising in a vacuum. It's part of a broader wave of large, fast rounds for AI applied to physical science and hard infrastructure that's defined mid-2026 dealmaking.

Figures from TechFundingNews, Forbes, and company announcements as of July 22, 2026.

Is a $2.6B valuation justified, or is this multiple expansion again?

CuspAI doesn't disclose revenue, and a company this early rarely has ARR that comes close to justifying a $2.6 billion price tag on fundamentals alone. What it does have is a partner list that reads like a who's-who of chipmaking and advanced manufacturing, and a lead investor bench — Kleiner Perkins, NEA, Bezos Expeditions, AMD Ventures — that's betting the partnerships convert into paid enterprise contracts faster than the typical deep-tech company.

That's the same dynamic we've tracked across the broader AI market all year: valuations are increasingly priced on distribution and strategic partnerships rather than trailing revenue, which is exactly the multiple expansion dynamic now showing up in categories well outside large language models. Materials science AI is a genuinely hard, capital-intensive problem with real physical-world validation cycles — but a 5x markup in ten months means the market is pricing in a lot of future contract conversion that hasn't happened yet.