CuspAI named more than 45 founding partners for an AI Materials Foundry on Monday 20 July, describing the coalition as "a global network of data, labs, compute and scientific expertise for the design of new materials"1. The Cambridge company uses machine learning to predict compounds with specified properties, and announced the same day a $450m Series B at a $2.6bn valuation led by Kleiner Perkins and NEA, with Bezos Expeditions participating.
The names on the partner list own machines. Applied Materials, Tokyo Electron and Lam Research build the process equipment that fabricates semiconductors; Samsung and Hyundai Motor Group run production lines; Nvidia, Meta and Henkel are on it as well. A compound predicted on a Monday morning still needs a furnace to make it, a diffractometer to read its crystal structure and a technician to run both. That gap between prediction and synthesis is what the coalition addresses, and CuspAI has bought its way across with partnership terms rather than capital expenditure.
We reported the $400m term sheets in June at the same $2.6bn mark, so the valuation carries little a reader did not have a month ago. Fortune put the September 2025 Series A at $520m, citing a person familiar with the round2, which makes the rise roughly fivefold in ten months while the physical work stayed in other people's buildings.
That list is CuspAI's own account of its own initiative, and we have not verified the membership independently. No partner has published a commitment of hours, materials or money. Coalitions of this shape have historically failed on scheduling rather than science: the United States ran into the same wall with the Materials Genome Initiative from 2011, where computational screening outran wet-lab throughput by orders of magnitude and queueing for instruments became the constraint. Until a partner names what it has contributed, the Foundry cannot be audited from outside.
