A startup called CuspAI just assembled 45 of the world's largest companies and research institutions into a single network designed to solve one of the biggest problems holding back industrial progress: the world doesn't have the materials it needs yet. The AI Materials Foundry, announced today, brings together data, computing power, laboratory infrastructure, and scientific expertise under one orchestrated platform to design new materials faster than ever before.

The founding members include semiconductor giants like NVIDIA, Samsung, and Intel-adjacent companies; energy firms like Oxford PV; chemical manufacturers like Kemira and Henkel; and research powerhouses like Cambridge University and Singapore's A*STAR agency. What makes this different from previous AI initiatives is the scope: rather than a single lab or company trying to solve materials discovery alone, this is a compounding intelligence network where breakthroughs achieved by one partner can accelerate discovery across the entire ecosystem.

What's the Real Problem These Companies Are Trying to Solve?

For decades, the bottleneck in industries like semiconductors, clean energy, and advanced manufacturing hasn't been engineering know-how. Engineers understand how to build better chips, batteries, and manufacturing systems. The constraint is materials. We need semiconductors with specific electrical properties that don't exist yet. We need catalysts for carbon capture that work at scale. We need polymers that meet cost and performance targets simultaneously. Creating these materials traditionally requires years of trial-and-error experimentation.

CuspAI's track record hints at what's possible. With a single customer, Finnish chemicals company Kemira, the startup screened a search space of 300 trillion potential molecular structures and delivered 20 validated novel candidates for further testing. This process previously took Kemira years to accomplish; CuspAI completed it in six months.

How Does the AI Materials Foundry Actually Work?

At the heart of the network sits MIRA, CuspAI's proprietary AI platform. A partner company defines what it needs, such as a compound with specific thermal stability, a semiconductor with a target bandgap, a catalyst with a defined reaction profile, or a polymer meeting a cost threshold. MIRA then generates candidate structures using generative models trained on the world's most comprehensive experimental materials dataset.

The platform doesn't stop at generating ideas. It screens millions of candidates using property prediction at scale, selects the most promising ones, designs synthesis routes matched to the available lab infrastructure in the network, and routes the work to the right facility based on capability, geography, and throughput. As results come back from physical experiments, MIRA feeds them back into the model and sharpens its predictions with every cycle.

What Makes This Network Technically Possible?

Four critical components had to come together for this to work, and the Foundry assembles all of them for the first time at industrial scale:

• High-Quality Training Data: CuspAI has secured exclusive AI training rights to foundational materials science datasets, including the Cambridge Structural Database and the Inorganic Crystal Structure Database, plus licensed access to materials science content from publishers like Wiley. This data advantage compounds with every validated experimental result returned through Foundry programs.
• Compute Power: NVIDIA is providing accelerated computing infrastructure capable of screening at molecular resolution across billions of candidates. The simulation layer runs on kUPS, an open-source molecular simulation toolkit built by CuspAI in collaboration with NVIDIA's ALCHEMI team.
• Synthesis Infrastructure: The network includes autonomous synthesis capabilities and physical laboratories across multiple geographies, allowing designs to move from digital to physical reality without bottlenecks.
• Domain Expertise: CuspAI's leadership includes Professor Max Welling, who co-invented the variational autoencoder and equivariant neural network architectures underlying generative molecular design; Professor Aron Walsh, one of the world's foremost computational materials scientists; and John Giannandrea, who built AI research at Google and served as Apple's SVP of Machine Learning and AI Strategy.

Meta's Fundamental AI Research Team contributes the Universal Model for Atoms (UMA), a frontier atomistic chemistry model that enables fast, accurate simulation of atomic interactions across the periodic table. This model integrates with ALCHEMI to create a continuous pipeline from candidate generation to physical property prediction at GPU scale.

How to Deploy AI Materials Discovery in Your Organization?

For companies and research institutions joining the Foundry, the path forward involves several concrete steps:

• Define Your Materials Challenge: Partners specify exactly what they need, whether a compound with specific thermal stability, a semiconductor with a target bandgap, a catalyst with a defined reaction profile, or a polymer meeting cost thresholds.
• Access the MIRA Platform: Members learn state-of-the-art methods in AI for science and agentic materials discovery, including how to deploy CuspAI's discovery platform and autonomous scientific agent within their existing R&D infrastructure.
• Leverage Network Resources: Rather than relying on isolated laboratory constraints, partners tap into the global network's compute, synthesis capabilities, and experimental validation infrastructure, with work routed to the most suitable facility based on capability and geography.
• Protect Proprietary Data: Partner data is protected in private Foundry instances designed for industrial confidentiality at the scale of multinational and government operations.

"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," said Dr. Chad Edwards, CEO and Co-Founder of CuspAI.

Dr. Chad Edwards, CEO and Co-Founder, CuspAI

What's Already in Motion?

One Foundry project already underway is a multi-year partnership between CuspAI and Singapore's A*STAR agency, which will combine AI-driven discovery with autonomous synthesis capability across semiconductors, carbon capture, and advanced electronics. This signals that the network isn't theoretical; it's operational and addressing real industrial challenges today.

"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," noted Ian Buck, Vice President of Hyperscale and HPC at NVIDIA.

Ian Buck, Vice President of Hyperscale and HPC, NVIDIA

The broader context matters here. The University of California system recently invested $19 million in an "AI Science at Scale" initiative, awarding grants to multi-campus teams working on AI-driven genomics, quantum materials discovery, geothermal energy, and integrated data platforms. This signals that the U.S. Department of Energy and major research institutions view AI-powered materials discovery as foundational to national competitiveness and scientific progress.

What CuspAI has built is essentially a production infrastructure for materials discovery. Instead of relying on the physical constraints of a single laboratory, the ecosystem approach creates what the company calls a "compounding intelligence loop." Breakthroughs achieved within the network have the potential to accelerate discovery timelines across the entire global value chain, meaning that a breakthrough in semiconductors could inform approaches to catalysis, and vice versa.

The 45 founding partners span industries that have historically worked in isolation: semiconductor manufacturers, chemical companies, tire makers, pharmaceutical firms, and government research agencies. By pooling data, compute, and expertise, they're betting that the bottleneck holding back industrial progress isn't creativity or engineering skill, but the speed at which new materials can be discovered, validated, and brought to scale. If they're right, the next decade of innovation in clean energy, semiconductors, and advanced manufacturing could look very different.