CuspAI

CuspAI (Cusp AI Limited) is a Cambridge, UK-based company developing artificial intelligence systems for materials discovery. It builds chemistry "foundation models" that make predictions about materials in a manner analogous to how large language models predict language, and describes its technology as a search engine for the material world. Rather than relying on trial-and-error experimentation, customers specify the properties they need and the company uses generative AI, physics-based simulation and reinforcement learning to propose novel candidate materials, then simulates their properties and guides synthesis and experimental testing.

The company's platform, MIRA, coordinates the end-to-end discovery process from candidate design through simulation to experimental validation, including routing candidate compounds to laboratories judged suitable on criteria such as production capacity. MIRA is built on open-source components including Meta's UMA models, which simulate compounds at the atomic level to map electronic properties, and CuspAI's own kUPS toolkit, which simulates at the molecular level (useful for crystal structure work), improves GPU processing efficiency for mixed atomic/molecular datasets and assists in troubleshooting simulation errors. CuspAI also assembled training data from academic datasets for which it purchased exclusive AI training rights and from scientific journals. Its published open-source releases include kUPS and tojax (PyTorch-to-JAX translation), plus smaller tools such as tonno, slub, nanoargs and widom.

Application areas span semiconductors, batteries and energy storage, clean energy and hydrogen fuel cells, climate technologies including direct air capture of carbon dioxide, water purification and advanced manufacturing. In July 2026 the company launched the AI Materials Foundry, a consortium of more than 45 founding members spanning compute providers, laboratories, proprietary data holders and scientific expertise, with MIRA as the coordinating layer.

Founding story

The company was founded in 2024 by Chad Edwards and Max Welling. Edwards, who grew up partly in the Welsh village of Llanfyllin and was the first in his extended family to attend university, credits a chemistry teacher with sparking his interest in science; he took a PhD at Manchester University and worked at Google and at Quantinuum before CuspAI. Co-founder Max Welling is an academic AI researcher. An investor account describes the founding team as heavily cited in the field, combining researchers in AI, chemistry and engineering.

Business model

CuspAI pursues a dual approach: designing materials to order for large industrial partners under commercial agreements and partnerships, and developing materials for its own projects. Management has said the emphasis to date has been on making materials to partner specifications, chosen early on to work with partners that have clearly defined problems and a route to market. Access to MIRA-driven discovery is also offered through membership of the AI Materials Foundry, where members can request new compounds.

Commercial revenue derives from partner-funded materials design programmes in which CuspAI develops materials to customer specifications, complemented by participation in the AI Materials Foundry and by internally owned materials development projects. The sources do not disclose pricing, contract terms or revenue figures.

Traction

As of September 2025 the company had about 30 employees, hired in part from Google DeepMind and Meta, and said it had gone from a customer request for a material with specified properties to a novel material meeting those specifications within six months, versus a process it characterised as normally taking a decade, with an aspiration to reduce this to one to two months within two years. Partnerships include Hyundai Motor Group (sustainable energy applications) and Kemira (materials to filter PFAS from water), with semiconductor-industry partnerships under discussion. An investor account states the company works with NVIDIA, ASML and Hyundai and describes customer demand since the seed round as strong. The AI Materials Foundry launched in July 2026 with more than 45 member organisations.

Latest developments

In July 2026 CuspAI announced a $450m Series B at a $2.6bn valuation, co-led by Kleiner Perkins and NEA with a significant contribution from Bezos Expeditions, and simultaneously launched the AI Materials Foundry with more than 45 founding members including Nvidia (providing compute infrastructure), Meta's Fundamental AI Research team, Samsung, Hyundai Motor Group, Hitachi High-Tech and Merck. The capital is directed at expanding operations in the US, Europe and Asia-Pacific.