Artificial intelligence first transformed what is coded, then what is written; it now extends to what is manufactured. Starting with materials, where innovation is measured in years, capital and scientific uncertainty. This is the ground on which positions itself CuspAIa British startup founded in 2024 by Chad Edwards, former director of Quantinuum, and Max Welling, AI pioneer and ex-Microsoft Research, whose ambition is to transform the very conditions for the discovery of new materials.
Reversing laboratory logic
Thus the design of materials historically remains a largely empirical process, researchers formulate hypotheses, test, adjust, start again, sometimes for years, without any guarantee of results. CuspAI proposes to reverse this logic.
On its platform, the input is no longer the molecule, but the property sought: conductivity, thermal resistance, energy capacity. From these parameters, artificial intelligence generates compatible molecular configurations. The approach is part of the “AI for science” trend, which aims to integrate learning models at the very heart of the scientific process.
But this promise is based on a delicate articulation, which is to bring together simulation and experimental validation, two distinct temporalities, subject to different constraints, and whose alignment remains uncertain.
Industrial partners as a signal of credibility
To attack the market, CuspAI targets sectors where the material directly determines performance: semiconductors, batteries, energy, carbon capture. In these industries, a marginal gain in physical properties can produce an immediate competitive advantage.
The startup has already signed commercial contracts worth tens of millions of dollars with groups such as NVIDIA, ASML or Hyundai Motor Company. More than references, these partnerships constitute a full-scale test of the capacity of AI to be inserted into demanding industrial chains
A funding dynamic aligned with the AI for science trend
After seeding in 2024, CuspAI raised a Series A in 2025 from New Enterprise Associates and Temasek, bringing its valuation to nearly $800 million. A new fundraising, of at least 200 million dollars, would be underway, with a prospect of entry into the unicorn club.
The dynamics of CuspAI are part of an underlying trend. AI for science is attracting a growing volume of capital, particularly in the United States, where former OpenAI and Google DeepMind researchers have launched startups already valued at over a billion. Investors like Jeff Bezos also support these types of initiatives. However, European players remain faced with a financing gap which reflects as much a question of available capital as an ecosystem differential.
Support from deep learning figures
The startup benefits from a network of leading scientific and industrial support. Geoffrey Hinton and Yann LeCun, major figures in artificial intelligence, are among his advisors, alongside Martin van den Brink, former president and CTO of ASML, and Lord John Browne, ex-CEO of BP.
The critical passage: from calculation to reality
This is where the main sticking point lies. The rapid generation of promising molecular structures does not eliminate laboratory test cycles, certification processes, or industrial integration constraints.
While calculation speeds up a step, it does not necessarily reduce the overall duration of the innovation cycle. However, the targeted customers, starting with large industrial groups, operate over long horizons, with high reliability requirements. The adoption of such a tool is based less on the promise than on the accumulation of validated results.
Furthermore, the economic model itself remains under construction: software licenses, SaaS platform, industrial co-development. No standard has yet been established in this still emerging sector.
An indicator of a broader transformation
CuspAI embodies these startups where the role of artificial intelligence moves from the optimization of existing processes towards upstream intervention, in the very definition of industrial objects.
This trend should not only attract the attention of manufacturers and investors. It also calls for a structured response from public decision-makers, commensurate with the issues it reveals.
Because beyond the case of a startup, a strategic capacity is emerging: that of designing the materials that will condition the next generations of energy, digital and industrial infrastructures. In other words, a direct lever of sovereignty.
