From Algorithms to Atoms, Part II: Doubling Down on CuspAI

Last year, we wrote about why we believed CuspAI would define the next era of materials discovery: a suite of proprietary generative models that combine novel materials generation, simulation, selection and physical validation to compress a decade of R&D into months. Since leading Cusp’s Series A in September 2025, the company has done what many peers struggle to do: validate the theory by making significant progress with customers in their research of synthesised and characterized materials.

Today, we're thrilled to share that NEA is co-leading CuspAI's $450M Series B, alongside Kleiner Perkins, and with participation from Bezos Expeditions and others, in one of the highest-conviction follow-on investments we've made. Here's why:

From Promise to Proof

When we first backed CuspAI, our thesis was built on team, data, model quality and early commercial partnerships. A few months later, our conviction has only strengthened. 

The team keeps getting stronger, with world-class hires and flawless execution. The data moat keeps widening, with CuspAI now leveraging the largest curated experimental materials datasets in the world. The company has made significant progress with customers: Kemira, a leading chemical company compressed years of R&D in six months, screening a search space of 300 trillion potential molecular structures  and delivering twenty validated novel candidates for testing and validation. CuspAI has announced partnerships with ASML, Hyundai and others. 

CuspAI is also proud to announce the launch of the AI Materials Foundry, which brings together a global network of 45+ organizations spanning data, labs, compute and scientific expertise for the design of new materials, all orchestrated by a single agentic platform. The Foundry provides the production infrastructure for industrial material discovery to accelerate discovery timelines across the whole value chain. At the core of this ecosystem lies MIRA, CuspAI’s proprietary generative platform which orchestrates the complete discovery loop—transitioning seamlessly from novel material generation and simulation to synthesis mapping and physical characterization. CuspAI then runs simulations on kUPS, its open source simulation toolkit. This agentic intelligence is powered by the most extensive curated repository of experimental materials data globally.

Through the Foundry program, partners can participate in a range of ways - including defining materials requirements, and having MIRA generate candidate structures, select the most promising ones, and orchestrate the design and production process - all the way through to identifying the right facility and tracking outcomes to feed back into the model. Within the program, members can also learn how to best implement CuspAI’s technology within their existing R&D infrastructure.  

Why We're Doubling Down

We continue to believe that CuspAI is leading the field of materials discovery and its capex-light, partnership-driven model will further solidify that position. We have been witnessing firsthand a clear pattern across customer engagement, a widening data moat, ever-improving models and a team executing impressively — which is exactly what gives us the confidence to lead this round.

We are proud to continue partnering with Chad, Max, and the entire CuspAI team as they build the intelligence layer for the physical world.