CuspAI raises $450M at $2.6B valuation for AI materials platform

CuspAI Ltd. has raised $450 million in Series B funding at a $2.6 billion valuation to expand its AI-driven materials discovery work. The U.K.-based startup also announced a research consortium called the AI Materials Foundry, giving the round a broader strategic angle than another AI software financing.

Kleiner Perkins and NEA led the round, with what CuspAI described as a significant contribution from Bezos Expeditions. The company said more than half a dozen other backers joined the financing, including investor John Doerr. CuspAI did not disclose revenue, headcount, customer contract value or the split between primary and secondary capital.

The company is trying to shorten a slow and expensive part of chemistry research: screening large numbers of possible material and molecular structures before lab validation. CuspAI says its platform, called MIRA, can reduce work that often takes years to about six months. That is a company claim, and CuspAI did not provide independent benchmark results or commercial deployment metrics in the announcement.

What CuspAI is building

MIRA combines two open-source systems: UMA and kUPS. UMA is a set of AI models released by Meta Platforms Inc. last year that can simulate materials at the atomic level and help map electronic properties. Those properties can indicate how a compound may behave when used with other materials.

kUPS was developed by CuspAI and operates at the molecular level, which the company says is useful for work such as analyzing crystal structures. CuspAI also says kUPS can help automate some coding work tied to simulations, improve the handling of datasets that include both atomic and molecular information, and assist with troubleshooting simulation errors on graphics processing hardware.

CuspAI said MIRA was trained using several academic datasets for which it acquired exclusive AI training rights, along with information from scientific journals. The company did not name the datasets, disclose licensing terms or say how much of MIRA’s performance comes from proprietary data versus open-source components.

The consortium angle

The AI Materials Foundry has more than 45 members, according to CuspAI. Named participants include Nvidia Corp., Samsung Electronics Co. and Meta, along with other technology companies. The group is built around MIRA and is intended to speed up the search for new materials.

Under the model CuspAI described, members can submit requests for new compounds. A chipmaker, for example, could ask for a material suitable for interconnects. MIRA would generate candidate materials and route them to a lab capable of making them, with the lab selected based on factors such as production capacity.

That workflow is commercially interesting if it works at scale, particularly for semiconductor, energy and industrial customers that need materials with narrow performance requirements. The announcement, however, does not say how many candidate materials have moved from software prediction to lab production, how many have passed validation, or whether any are in customer products.

CuspAI Chief Executive Officer Chad Edwards said the company is focused on combining AI, domain expertise, data access and customer partnerships to address demand for new industrial materials. The new funding will be used to expand CuspAI’s international presence, the company said. It did not provide a hiring target, geographic breakdown or timeline for that expansion.

This story draws on original reporting from SiliconANGLE.