CuspAI, a two-year-old British startup using artificial intelligence to discover new materials, has raised $450 million as demand for more efficient semiconductor manufacturing continues to intensify. The Series B financing, led by venture firms Kleiner Perkins and NEA with significant participation from Jeff Bezos's fund Bezos Expeditions, values the Cambridge-based company at $2.6 billion, up from $520 million last September. CuspAI also launched the AI Materials Foundry, a coalition of more than 48 technology companies, industrial firms, and research facilities that plans to combine computing power and scientific resources to develop materials for chipmakers and other industries more quickly and at a lower cost. Members include Nvidia NASDAQ:NVDA, a chipmaker seeking new materials across areas such as energy storage and data-center cooling, Meta Platforms, a technology company whose researchers previously worked with CuspAI on carbon-capture materials, and Hyundai Motor Group, one of the coalition's participating companies.
CuspAI expects semiconductor projects to represent 80% of its research efforts this year as chip manufacturers and suppliers search for materials that may reduce production constraints and supply-chain exposure. One possible area of research involves limiting or eliminating the use of rare metals such as ruthenium and iridium, which the company identified as carrying supply-chain risks. Chief Executive Chad Edwards said much of the new funding will support laboratories with Foundry partners in Cambridge, Singapore, and the San Francisco Bay Area. Nvidia is likely to pursue three-way collaborations involving CuspAI and another partner, while Advanced Micro Devices, a semiconductor company whose venture unit joined the financing round, has gained another connection to the startup through the appointment of AMD NASDAQ:AMD director and semiconductor veteran Abhi Talwalkar to CuspAI's board. CuspAI also hired John Giannandrea, a former Google and Apple executive, to work part-time on establishing its California operation.
The financing and high-profile partnerships may attract further investor attention to AI-driven materials research, although CuspAI's results also highlight the technical and commercial uncertainty surrounding the field. In an earlier project with Meta NASDAQ:META researchers, CuspAI reduced 300 trillion possible carbon-capture structures to 10 candidates and successfully synthesized six, but none performed better than materials already commercially available. Edwards said the process was completed in six months, which he characterized as a substantial improvement over typical industry development timelines. CuspAI is now applying a similar approach to materials designed to remove toxic forever chemicals from water, with Kemira Oy, a Finnish chemicals company, expected to synthesize and test 20 structures this year. Despite the company's funding and industry support, none of its projects has yet produced a candidate molecule ready to leave the laboratory, suggesting investors may need to balance the platform's potential speed advantages against the long and expensive process of validating new materials.