University of Cambridge researchers have joined a new international network designed to shorten the path between artificial intelligence predictions and materials that can be tested and manufactured in the real world.
The AI Materials Foundry was launched by Cambridge-based start-up CuspAI and brings together data, advanced computing, experimental laboratories and scientific expertise. Its stated aim is to accelerate materials discovery across fields including semiconductors, clean energy and advanced manufacturing.
Cambridge’s involvement comes through its Department of Materials Science and Metallurgy, where researchers are working on systems that combine AI-led design with laboratory testing. The university said the wider network includes more than 45 members from academia, industry and research, including NVIDIA, Meta Platforms and the Henry Royce Institute, the UK’s national institute for advanced materials research and innovation.
Closing the gap between prediction and experiment
Generative AI systems can propose large numbers of possible materials, but a computer-generated candidate is not automatically useful. It still has to be physically produced, tested and shown to perform under real conditions.
That gap is one of the central problems the Foundry is intended to address. The network will link computational design and simulation with synthesis planning and experimental validation, allowing laboratory results to feed back into future AI predictions.
At Cambridge, the work connects with research led by Associate Professor Shijing Sun, whose group develops AI- and robotics-guided methods for designing, making and understanding advanced functional materials. The department said its capabilities in high-throughput synthesis and advanced characterisation would help build closed-loop workflows in which predictions and experiments continuously inform one another.
Sun has also received support from the Alchemy Frontier Fund for a related project with Professor Aron Walsh of Imperial College London, who is CuspAI’s chief scientific officer. According to the University of Cambridge, the project will use generative AI to identify candidate materials and then connect those models directly with experimental work.
Why materials discovery matters
Many technologies depend on materials with highly specific properties. Improvements in batteries, semiconductor performance, carbon capture and clean-water systems may require compounds that are more efficient, less expensive or less reliant on scarce metals than those in current use.
CuspAI has described the shortage of suitable materials as a constraint on industrial progress. Independent reporting by The Guardian said the company wants to reduce research times and limit dependence on rare metals used in technology supply chains.
The company raised $450 million in its latest funding round, according to the same report, giving it a valuation of $2.6 billion. The investors included the UK government’s sovereign AI fund and Amazon founder Jeff Bezos. The Foundry therefore sits at the intersection of university research, private investment and a broader policy effort to expand the UK’s capacity in advanced AI and materials science.
For Cambridge, the immediate significance is less about the size of the coalition than the structure of the research process. AI can narrow the field of possible materials, while university laboratories can test whether the most promising candidates are stable, practical and capable of being made at useful scale.
If that feedback loop works as intended, it could reduce the number of promising digital predictions that never progress beyond simulation. The Foundry’s success will ultimately depend on whether its shared data, computing resources and laboratory network produce materials that can move from research programmes into industrial use.