CuspAI has unveiled the AI Materials Foundry, a global ecosystem designed to accelerate the discovery of advanced materials by combining artificial intelligence, high-performance computing, scientific data, laboratory infrastructure and domain expertise on a single agentic platform. More than 45 organisations have joined as founding members, including NVIDIA, which will provide the computing infrastructure, and Meta’s Fundamental AI Research (FAIR) team, whose Universal Model for Atoms (UMA) will support atomistic simulations for materials science.
The launch reflects growing industry efforts to address what many researchers see as one of the biggest barriers to technological progress. While advances in semiconductor design, clean energy systems and advanced manufacturing continue to gather pace, the development of new materials has remained a lengthy and resource-intensive process, often taking years of laboratory research before promising candidates emerge. CuspAI aims to shorten that timeline by integrating every stage of materials discovery—from molecular design and simulation to synthesis planning and experimental validation—within its proprietary AI platform, MIRA.
The company said MIRA is built on what it describes as the world’s largest curated experimental materials dataset, enabling researchers to conduct complete discovery cycles through a single platform. Rather than operating as isolated research projects, organisations within the Foundry can leverage shared computing resources, specialised laboratories and scientific expertise while maintaining the confidentiality of their proprietary research. The platform continuously learns from validated experimental results, allowing each discovery cycle to improve future predictions and accelerate subsequent programmes.
“The next era of industrial innovation depends on materials that do not yet exist,” said Dr Chad Edwards, CEO and Co-Founder of CuspAI. “If we don’t make progress fast, the next 50 years of industrial progress will be constrained by a single challenge: the world needs materials that don’t yet exist. That’s what we’re on a mission to solve—combining frontier agentic AI with deep domain expertise, exclusive data access and close customer partnerships.”
The Foundry brings together four capabilities that have traditionally existed in separate environments: large-scale experimental datasets, GPU-accelerated computing, synthesis infrastructure and scientific expertise. According to CuspAI, this integrated approach enables AI models not only to generate new material candidates but also to determine viable synthesis routes, assign experiments to appropriate laboratory facilities based on expertise and capacity, and feed validated results back into the models to improve prediction accuracy over time.
NVIDIA will provide the accelerated computing infrastructure required to screen billions of molecular candidates and run simulations at molecular resolution. Ian Buck, Vice President of Hyperscale and HPC at NVIDIA, said advances in AI will increasingly depend on breakthroughs in physical materials. “As AI transforms the physical world, new materials will open up new frontiers across semiconductors, energy and advanced manufacturing. The AI Materials Foundry brings NVIDIA accelerated computing infrastructure together with world-class chemistry and materials expertise to help power the next generation of materials discovery.”
Meta’s FAIR team joins the initiative through its Universal Model for Atoms (UMA), an open-source foundation model for materials science capable of simulating atomic interactions across the periodic table. Rob Fergus, Vice President of AI Research and Head of FAIR at Meta, said the collaboration would allow researchers to tackle materials challenges that have historically proved difficult to solve. “We’re proud to be longstanding partners with CuspAI and now as founding members of the AI Materials Foundry ecosystem. Our open-source frontier models for materials science research will enable more teams to tackle previously intractable challenges, more precisely and more quickly than before.”
At the core of the Foundry’s workflow is MIRA, an autonomous scientific agent that enables researchers to specify target material characteristics—such as thermal stability, semiconductor bandgap, catalytic activity or cost thresholds—and automatically generates candidate molecular structures using generative AI models trained on experimental data. The platform evaluates millions of potential compounds, identifies the most promising candidates, designs synthesis pathways compatible with available laboratory capabilities and coordinates experimental validation across partner facilities. Each completed experiment is incorporated into the platform’s knowledge base, creating what CuspAI describes as a continuously improving intelligence loop.
Supporting this workflow is kUPS, an open-source molecular simulation toolkit developed jointly by CuspAI and NVIDIA’s ALCHEMI (AI Lab for Chemistry and Materials Innovation). Integrated with Meta’s UMA model, the toolkit enables GPU-scale simulation and property prediction, creating an end-to-end computational pipeline from molecular generation to physical validation.
The company points to earlier industrial deployments as evidence of the platform’s capabilities. In collaboration with Finnish chemicals company Kemira, CuspAI screened a search space of approximately 300 trillion molecular structures and identified 20 experimentally validated candidates for further testing. According to the company, the programme was completed in six months, compared with timelines that would traditionally extend over several years using conventional discovery methods.
One of the first major initiatives under the AI Materials Foundry is a multi-year partnership with Singapore’s Agency for Science, Technology and Research (A*STAR). The collaboration will combine AI-driven materials discovery with autonomous synthesis capabilities to develop new materials for semiconductors, advanced electronics and carbon capture technologies, highlighting the Foundry’s focus on translating computational discoveries into industrial applications.
CuspAI’s competitive advantage, the company argues, lies not only in AI models but also in the quality of the data used to train them. It has secured exclusive AI training rights to foundational experimental materials databases, including the Cambridge Structural Database through the Cambridge Crystallographic Data Centre (CCDC) and the Inorganic Crystal Structure Database through FIZ Karlsruhe. These are complemented by licensed scientific content from Wiley and other research publishers, while every validated result generated through Foundry programmes is incorporated into the training pipeline to expand and refine the platform’s knowledge base.
The company has also assembled a leadership team with expertise spanning artificial intelligence and computational materials science. Co-founder and Chief Technology Officer Professor Max Welling is recognised for pioneering work on variational autoencoders and equivariant neural networks, technologies that underpin many modern generative molecular models. Chief Scientific Officer Professor Aron Walsh is a leading computational materials scientist, while former Apple Senior Vice President of Machine Learning and AI Strategy John Giannandrea will support the establishment of the company’s US Foundry operations.
