Amazon founder Jeff Bezos and the UK government have invested in a £2bn British AI startup that’s building what it calls a “search engine for rare materials” — frontier AI trained to discover new compounds that could replace the scarce metals underpinning the semiconductor and clean energy industries.
The Cambridge-based CuspAI has raised $450m (£330m) in Series B funding at a $2.6bn valuation, with investors including Bezos and the UK government’s sovereign AI unit. The two-year-old company is the fourth to receive backing from the sovereign AI fund, which targets early-stage British AI firms with equity investments typically between £1m and £10m.
What is CuspAI? CuspAI is an AI company focused on materials discovery — using large language models and generative AI to predict which chemical combinations will produce useful new materials, the way AlphaFold predicted protein structures for biology. The goal: cut research times from years to weeks, and reduce or eliminate reliance on rare metals like iridium and ruthenium that are both expensive and geopolitically concentrated.
🔍 THE BOTTOM LINE
The AI chip boom is bottlenecked not by compute but by materials. Iridium, ruthenium, and other rare metals are scarce, expensive, and largely sourced from a handful of countries. CuspAI’s pitch is that frontier AI can discover replacements — and a 48-company coalition including Nvidia, Meta, and Hyundai is betting that’s not wishful thinking. The £2bn valuation says the market agrees.
The AI Materials Foundry
CuspAI isn’t going it alone. The company has launched the AI Materials Foundry, a coalition of more than 48 tech companies, industrial firms, and research facilities. The names carry weight: Nvidia (the chipmaker at the centre of the AI hardware boom), Meta Platforms (which needs advanced materials for its own data centre infrastructure), and Hyundai (which needs them for electric vehicle batteries and motors).
The foundry’s goal is to use frontier AI — the most advanced, general-purpose models — to discover new materials that could make energy networks and batteries more efficient, and advance the semiconductors that underpin the AI revolution itself. It’s a closed loop: AI needs better chips, better chips need better materials, and CuspAI wants AI to find those materials.
“If we don’t make progress fast, the next 50 years of industrial progress will be constrained by a single challenge: the world needs materials that don’t yet exist,” said CuspAI co-founders Dr Chad Edwards and Prof Max Welling. “The world simply can’t make the progress it needs to with known materials, that much is clear.”
Why Rare Metals Are the Problem
The AI industry’s hardware stack runs on materials that are neither abundant nor easily substitutable. Iridium — used in some semiconductor processes and electrolysis — costs roughly $5,000 per ounce and is primarily sourced from South Africa and Russia. Ruthenium, another platinum-group metal, is similarly concentrated. These supply chains are vulnerable to export controls, geopolitical disruption, and simple geological scarcity.
The problem isn’t hypothetical. When export controls on advanced chips tightened in 2024-2025, the ripple effects through the materials supply chain were immediate. A company that can identify viable substitutes — or find ways to use less of these metals per chip — has a product every semiconductor manufacturer wants.
Josh Coyne, a partner at Kleiner Perkins (which co-led the funding round), framed it bluntly: “Most big leaps in technology come down to a material, and the next set, from cheaper carbon capture to semiconductors and cleaner water, is stuck waiting on materials nobody has discovered yet. CuspAI built a search engine that changes that.”
The Sovereign AI Angle
The UK government’s involvement is the part that elevates this beyond a standard venture deal. CuspAI is the fourth company to receive funding from the sovereign AI fund, which targets early-stage British AI firms to help them scale and compete globally. The criteria: a main office in the UK and British founders.
This is a different model from the US approach, where AI investment is almost entirely private. The UK is explicitly using public capital to build national AI champions in strategically important domains — and materials science is about as strategic as it gets. Science and technology secretary Liz Kendall framed it as economic policy: “By backing British companies such as this we are paving the way for breakthroughs that are set to grow our economy, protect the natural world and improve everyone’s lives.”
The sovereign AI fund’s typical investment of £1m-£10m is small relative to the $450m total round, but the signal matters. A government stake means regulatory alignment, potential defence procurement pathways, and a seat at the table when industrial strategy is being written.
The Talent Pipeline
CuspAI has brought on two names that tell you where it’s positioning: John Giannandrea, a former Apple and Google executive who led machine learning at both companies, is helping set up US operations. Abhi Talwalkar, a board member at AMD and chair of Lam Research, brings deep semiconductor industry connections.
These aren’t advisory-board trophies — they’re operational hires at a company that plans to expand internationally (Singapore office incoming, staff growth in Cambridge, Amsterdam, Berlin, Tokyo, and the US). The Series B follows a $100m Series A last year, meaning CuspAI has raised over $550m in two years.
NZ Angle
New Zealand has its own materials science ecosystem — particularly around advanced manufacturing and rare earth processing — but nothing at the scale of a CuspAI. The relevance is indirect: if CuspAI succeeds in finding substitutes for iridium and ruthenium, the cost structure of the entire semiconductor industry shifts, which affects the price of every AI-capable chip that reaches NZ. For a country that imports all its advanced silicon, cheaper materials upstream means cheaper hardware downstream. The NZ AI Blueprint emphasises sovereign AI capability — but sovereign AI starts with access to the hardware, and the hardware starts with materials.
❓ FAQ
Is CuspAI actually producing new materials yet, or is this still research-stage?
The company is two years old and has raised $550m+. The coalition structure (48 companies including Nvidia and Hyundai) suggests it’s past pure research but the founders’ framing — “materials that don’t yet exist” — indicates the commercial outputs are still emerging. The foundry model is designed to accelerate this.
Why is the UK government investing in a company Bezos is also backing?
The sovereign AI fund makes small equity investments (£1m-£10m) in British-founded AI companies. The goal is strategic: build national AI champions in domains that matter for industrial policy. A government stake alongside private capital is standard for sovereign fund models.
What does a “search engine for rare materials” actually do?
It uses frontier AI models to predict which chemical combinations will produce useful materials — the way AlphaFold predicts protein structures. Instead of running physical experiments for years, the AI narrows the candidate space, and labs then validate the most promising candidates.
Does this matter if you’re not in the semiconductor industry?
Yes — rare metals aren’t just in chips. They’re in batteries, solar panels, hydrogen electrolysis, and medical devices. If CuspAI’s approach works, the cost curve for the entire clean energy transition shifts.
🔍 THE BOTTOM LINE
CuspAI is betting that the next bottleneck in AI isn’t compute — it’s the materials that compute runs on. With Bezos’s capital, the UK government’s strategic backing, and a 48-company coalition that includes the biggest names in chips and manufacturing, the £2bn startup is positioned to attack a supply chain problem that every industrialised economy is quietly worried about. Whether frontier AI can actually discover new materials at industrial scale is the open question — but if it can, the company that does it first won’t just be worth £2bn.
