CuspAI Raises $450M as AI Moves From Language to Materials

The next major AI breakthrough may not be a better chatbot. It may be a material that makes chips faster, batteries more efficient, or carbon capture commercially viable.

Cambridge-based CuspAI has raised $450 million in Series B funding at a $2.6 billion valuation, giving the two-year-old materials intelligence company more than $670 million in total disclosed capital. The round was co-led by Kleiner Perkins and NEA, with significant participation from Jeff Bezos’ investment firm, Bezos Expeditions.

The UK government-backed Sovereign AI Fund also participated alongside Glade Brook Capital Partners, Lux Capital, AMD Ventures, Tru Arrow Partners, StepStone, Invest-NL, and existing investors including Temasek and Northzone.

Designing materials backwards from the outcome

Founded in 2024 by Chad Edwards and machine-learning researcher Max Welling, CuspAI applies generative and agentic AI to materials discovery. Instead of screening only known compounds, customers can define the performance they need—such as higher conductivity, lower carbon intensity, or stronger chemical capture—and use the platform to search for materials with those properties.

Its MIRA platform supports a wider discovery cycle that includes generative design, digital simulation, synthesis planning, and experimental validation. The objective is to compress research programs that can traditionally take years into faster, data-driven workflows.

The company is targeting areas where materials remain a central constraint, including semiconductors, batteries, clean energy, water treatment, carbon capture, and advanced manufacturing. Customers and partners cited publicly include Meta, ASML, Hyundai, and major industrial research organizations.

A foundry built around data, compute, and laboratories

The funding announcement coincides with the launch of CuspAI’s AI Materials Foundry, a global network designed to connect AI models with scientific datasets, computing infrastructure, industrial expertise, and physical laboratories.

More than 45 organizations have joined the initiative. NVIDIA is contributing accelerated computing capabilities, while Meta’s Fundamental AI Research team is providing open models for atomistic materials research. Industrial and research participants include Applied Materials, Samsung, Hyundai, Fujifilm, Merck, Oxford PV, the University of Cambridge, A*STAR, and multiple advanced-materials laboratories.

This network matters because materials discovery cannot be completed through software alone. AI may identify promising candidates, but those materials must still be synthesized, tested, refined, and manufactured reliably.

CuspAI’s strategic bet is that the winning platform will connect digital discovery with physical validation, rather than operating as an isolated scientific model.

Why the $2.6 billion valuation matters

CuspAI’s valuation has increased fivefold from the $520 million level reported around its $100 million financing in 2025. That jump reflects growing investor interest in AI systems capable of influencing the physical economy rather than only digital workflows.

The commercial opportunity is significant. New semiconductor architectures, energy systems, climate technologies, and industrial processes often depend on materials that do not yet exist at the required performance or cost.

The challenge is execution. Materials suggested by AI must move through laboratory validation, regulatory requirements, supply chains, and commercial manufacturing. Strong simulation results do not automatically translate into scalable products.

The market signal

CuspAI’s funding points to a wider shift in artificial intelligence investment.

The market is moving from models that generate text and images toward systems that can design molecules, materials, and industrial processes. These platforms may require longer development cycles than consumer AI products, but their economic impact could be considerably deeper.

CuspAI now has the capital, partnerships, and infrastructure to test whether AI can become a practical operating layer for materials research.

The next milestone will not be another impressive model demonstration. It will be a new material that reaches a factory, energy system, or semiconductor production line.
Source : Tech.eu