Analysis
CuspAI, a two-year-old Cambridge, UK-based startup that uses AI agents to discover new materials, raised a $450 million Series B led by Kleiner Perkins and NEA, with Jeff Bezos's Bezos Expeditions participating, valuing the company at $2.6 billion -- up sharply from a $520 million valuation last September, according to Electronics Weekly and MLQ News. The company says semiconductor-focused projects will represent about 80% of its research effort this year, and alongside the raise it launched the AI Materials Foundry, a coalition of more than 48 technology, industrial and research organizations.
CuspAI's pitch overlaps directly with Discovered Materials, the Y Combinator-backed startup that raised a $9 million seed round this week to use swarms of AI agents hunting for materials that reduce chip heat generation -- both companies are chasing the same underlying problem (AI chips running hot, driving up data center cooling costs and power draw) at very different scale and stage. CuspAI's roughly 5x valuation step-up in eleven months reflects investor appetite for the more capital-intensive, later-stage version of that same materials-discovery thesis.
The semiconductor materials-discovery category sits one layer below chip design itself: rather than designing a faster chip, these companies are trying to find the physical substances -- alloys, compounds, dielectrics -- that make a chip design manufacturable and efficient in the first place. It's a slower-moving, more capital-intensive category than software-layer AI infrastructure, closer in capital intensity to Sila's battery-materials work than to a typical enterprise AI funding round.
A 5x valuation increase in under a year for a pre-revenue-at-scale materials-discovery company is a steep multiple to underwrite without disclosed commercial contracts -- CuspAI has not published revenue figures or named specific chipmaker customers beyond the general "coalition" framing of its new AI Materials Foundry initiative, leaving outside investors to price the round largely on trajectory and the credibility of its backers rather than verified commercial traction.
