Tech giants bet on “AI alchemy”! British startup CuSpai raised nearly $500 million to target the chip manufacturing materials revolution

Making the world's most in-demand chip consumes a lot of energy and relies on access to scarce mineral resources. The British startup CuSpai was founded just two years ago and has now raised nearly $500 million, betting that artificial intelligence (AI) can improve this process.

The Zhitong Finance App learned that on Monday, CuSPai launched the “AI Materials Foundry” (AI Materials Foundry), an alliance of more than 48 tech giants, industrial companies, and research institutes. The organization's goal is to integrate computational resources and scientific research capabilities to develop software that helps researchers develop new materials for chip makers and other industries faster and at a lower cost than current methods.

To achieve this goal, the Cambridge, UK-based company has completed Series B financing of 450 million US dollars. Investors include funds owned by Amazon founder Jeff Bezos. CuSPai co-founder and CEO Chad Edwards said most of the funding will be used to support laboratories in Cambridge, Singapore, and the San Francisco Bay Area, built with foundry partners.

In terms of R&D, Edwards pointed out that 80% of the company's R&D investment this year will be used in the field of material discovery. He added that one of CuSPai's goals is to reduce or eliminate the use of rare metals such as ruthenium and iridium in chip manufacturing that present supply chain risks.

Cuspai said that the latest financing was led by venture capital agencies Kleiner Perkins and NEA, and that Bezos Expeditions (Bezos Expeditions) “participated heavily”. The financing brought the startup's valuation to $2.6 billion, a significant increase from $520 million in September last year.

CuSPai initially focused on finding new materials for carbon capture and water purification. However, Edwards said that in the past year, the company decided to adjust direction due to strong demand in the chip manufacturing supply chain, and chip products are essential to support the development of AI. In an interview, Edwards said that chip makers and semiconductor suppliers are “frantically searching for new materials,” and “we are actually being driven into this field by demand from all sides.”

Nvidia senior director Jitika Gupta said that the chip maker is seeing huge demand for new materials across the entire market, from energy storage systems to data center cooling technology. Gupta said that through material foundries, Nvidia is most likely to participate in “tripartite cooperation” projects carried out by CuSPai and third party partners. “All of these companies regard vast amounts of relevant information as their core secrets,” she said.

Cuspai is one of the AI developers that are increasingly using AI to explore scientific fields to open up new markets. Early AI systems have proven that they can learn from biological data such as protein structures, and may be used to discover new drugs. Newer-generation AI models can theoretically screen massive amounts of molecular data to design new materials suitable for manufacturing and renewable energy.

Investors are pouring into the field to support related projects promoted by former OpenAI and DeepMind researchers, as well as Prometheus, a new company launched by Bezos and valued at 41 billion US dollars. Google DeepMind, a subsidiary of Google, also stated that it will soon begin AI material discovery research.

CuSPai aims to build differentiated competitiveness by leveraging talent advantages. Its other co-founder, Max Welling, is a highly respected AI researcher. Modern AI pioneers Yang Likun and Jeffrey Hinton both serve on the company's advisory board.

On Monday, CuSPai also announced the appointment of chip industry veteran Abi Tarvokar, director of AMD (AMD.U), to the board of directors. AMD's venture capital division participated in Cuspai's latest round of financing. Additionally, CuSpai said the company has hired former Google and Apple executive John Gennandrea part-time to establish the company's business base in California.

David Phelan-Jimenez, professor of molecular engineering at Cambridge University, said machine learning systems have “indeed” accelerated the process of developing new materials. But he warned that designing new materials that can vastly surpass existing materials is a long and expensive process, and that artificial intelligence can only provide limited help. “It's very useful,” he said. But what I don't agree with is that machine learning alone would make some kind of magical breakthrough.” Ferran Jiménez founded two materials science startups, but is not currently working with or acting as an advisor to CuSPai.

Cuspai is already familiar with the extensive trial and error required in the AI material discovery process. One of the company's early projects was to collaborate with Meta researchers to find new materials that can capture carbon dioxide, which causes global warming, and mitigate climate change.

To achieve this goal, CuSPai has built a database of 300 trillion potential structures belonging to a class of chemicals called metal-organic framework materials (MOFs). MOF materials are composed of metal atoms surrounded by carbon-based molecular chains and have special advantages in the field of carbon capture. The system then screened this huge list of candidates to 10 types.

But translating theoretical research into practical results has proven to be more difficult. Edwards said that in the end, CuSPai can only synthesize less than 10 of these compounds, and there is no one that surpasses the performance of commercial products currently on the market. “They're not the most advanced materials,” he said.

However, CuSPai did not abandon this approach. Currently, the company is using the same database to find new materials that can remove toxic “permanent chemicals” from water. Finnish chemical company Kemira Oy will try to synthesize and test 20 of these new structures this year.

Despite having many well-known partners and investors, up to now, none of CuSPai's projects have produced enough candidates to leave the lab and be commercialized.

Co-founder Welling isn't concerned about this, however. He pointed out that insufficient reliable data, testing facilities, and computational capabilities are problems faced by companies in this field for a long time, and CuSPai hopes to solve these challenges by constructing material foundries. He said, “People have underestimated the resistance they need to overcome to actually conduct experiments. It may seem simple, but it really isn't.”