A British Startup Raised $450 Million to Build a Search Engine for New Materials
CuspAI closed a $450 million round at a $2.6 billion valuation, backed by Jeff Bezos' fund, Nvidia and the UK government, to use AI for finding materials that do not exist yet.
Cambridge-based CuspAI has raised $450 million in a Series B round that values the company at $2.6 billion. Kleiner Perkins and New Enterprise Associates led it, with the UK government, Jeff Bezos’ investment fund, AMD Ventures and Lux Capital taking part. The round is roughly a month after reports that CuspAI was seeking $400 million at the same valuation, which is a small sign of how quickly money is moving toward AI applied to science rather than to chat.
What the company does is easier to picture than most AI startups. Instead of asking a model to write for you, CuspAI asks it to find a material with properties you specify: a compound that conducts at a certain temperature, or one that works without a scarce metal such as iridium or ruthenium. It combines generative AI, physical simulation, planning for how a compound could actually be synthesised, and experimental validation in the lab. The company has also set up the AI Materials Foundry, a coalition of more than 45 partners including Nvidia, Meta and Hyundai, and plans to expand across the US, Europe and Asia-Pacific with operations in Singapore and several research hubs.
The reason investors like this is a matter of arithmetic rather than enthusiasm. Discovering a new material has traditionally meant years of laboratory work, testing candidates mostly by informed trial and error, with each experiment costing time and money. If AI can narrow millions of theoretical candidates down to a few dozen worth actually making, the expensive physical part of the process gets much shorter. Note the word “narrow”: the model does not replace the lab, it decides what the lab should try first. That is a more modest claim than “AI discovers new materials”, and it is also the claim most likely to hold up.
What this means for you: Nothing in your daily tools, and nothing this year. This is the slow kind of AI story, and its payoff, if it comes, shows up as better batteries, cheaper semiconductors and less dependence on scarce metals somewhere in the 2030s. What is worth taking from it now is the shape of the field: alongside the chatbots, a serious amount of capital is going into AI that reads simulations and molecules rather than sentences. A fair caveat: funding rounds measure investor conviction, not results, and materials discovery has produced confident predictions before that the laboratory did not confirm.
Sources
Google Moves Recipe Links Higher in AI Mode, and the Fight Is Bigger Than Dinner
Google is placing links to original recipe pages more prominently in AI Mode answers. It is a small change that goes to the heart of who pays for the web that AI summarises.
