With a shift from trial-and-error to predictive design, the Swiss National Research Competency Center MARVEL is scheduled to close on July 9, after 12 years of reshaping computational science and helping four materials startups raise more than $800 million in early-stage funding. Founded in 2016, the center combines quantum mechanical simulations, computational power and machine learning, an approach MARVEL director and EPFL professor Nicola Marzari explains draws from recent industry investments. Marzari points out that this reflects a broader movement toward designing materials on computers before physically creating them. Over its lifetime, MARVEL researchers have predicted and confirmed new quantum materials and strengthened the computational foundation for AI-driven materials discovery, an area currently receiving significant global interest.
Quantum materials discovery and predictive modeling
NCCR MARVEL’s impact is evidenced by the fact that four startups have raised more than $800 million in early-stage funding. For most of its history, materials discovery relied on iterative synthesis and testing, but MARVEL actively promoted a paradigm shift. Researchers integrated physics, chemistry, computer science, and experimental data to identify promising materials early and understand the principles governing their behavior. This approach has led to important advances, including the prediction and confirmation of new quantum materials that exhibit unusual electronic states that will be relevant for future technologies. These were not isolated incidents. This research translated a long-standing scientific puzzle into clearer design principles. Beyond scientific advances, MARVEL has significantly strengthened the computational infrastructure of materials science, developed advanced electronic structure methods, and incorporated them into open source code used around the world. This effort also developed machine learning techniques for molecules and materials, creating AI models that are now widely used. As machine learning has evolved from a promising add-on to a core research tool, MARVEL has helped pave the way for the current surge in interest in AI-powered materials discovery, fostering a Swiss digital ecosystem for materials science built around infrastructures such as AiiDA and Materials Cloud.
The team developed advanced electronic structure methods and incorporated many of these features into open source code used by researchers around the world.
MARVEL machine learning and simulation integration
Materials science is currently characterized by the convergence of computational power, quantum mechanical simulations, and increasingly sophisticated machine learning algorithms. While this combination has been largely theoretical in recent times, it is now rapidly becoming standard practice. Launched in 2016, MARVEL intentionally brings together these fields to accelerate and systematize materials research. This strategic focus influenced the scientific and economic environment. Four startups have raised more than $800 million in early-stage funding. “Some of the biggest internet companies, from Google DeepMind to Microsoft to Meta, have launched large-scale efforts in materials design and discovery over the past few years.” The MARVEL researchers didn’t just apply machine learning as an add-on. They pioneered ways to predict complex material properties such as spectra and electronic structure, laying the foundation for the current surge of AI-driven discoveries.
Beyond algorithms, MARVEL has developed open source code and a Swiss digital ecosystem, including the AiiDA and AiiDAlab infrastructure and Materials Cloud platform, to foster an open approach and encourage collaboration and validation. By emphasizing reproducibility and data sharing, we have created a national platform for computational materials research, allowing the global scientific community to easily run, compare, and build on simulations. The legacy of this work extends beyond specific discoveries to a fundamentally changed approach to materials innovation.
Some of the biggest internet companies, from Google DeepMind to Microsoft to Meta, have launched large-scale efforts in materials design and discovery in the past few years, reflecting an influx of startups into the field, with just four of them raising more than $800 million in early-stage funding.
Evolving computational infrastructure: AiiDA and material cloud
The transition to predictive materials design increasingly relies on robust computational infrastructure, and NCCR MARVEL is addressing that need by developing the AiiDA and Materials Cloud platforms. This focus on reproducibility and open science represents a major departure from the historical trial-and-error approach to materials discovery, where synthesis and testing dominated the process. MARVEL’s digital ecosystem is more than just a repository of data. Actively facilitate connections between simulation, data processing, experimentation, and in some cases operate automated experiment platforms. Its influence extends beyond academia, with MARVEL fostering relationships with industry partners in sectors ranging from energy to pharmaceuticals. This collaboration resulted in a transition from specialized software to practical tools that can be applied in industrial environments. This is a key factor reflected by the influx of startups into the sector, four of which have raised more than $800 million in early-stage funding. The culmination of these efforts will be celebrated in Lausanne on July 9, marking the official closing event of the 12-year NCCR MARVEL initiative and cementing its legacy as a pioneer in AI-driven materials science.
Social impact: energy and manufacturing materials
The impact of NCCR MARVEL extends beyond academic publications, clearly impacting the materials-based innovation landscape and attracting significant investment. The surge in startup funding shows that economic momentum correlates with MARVEL’s approach to materials discovery, with four companies raising more than $800 million in early-stage funding. This influx of capital signals growing confidence in computer-designed materials as viable commercial ventures, especially in areas that demand advanced performance properties. The effort’s focus on practical applications is evident in its research portfolio, which includes renewable energy, advanced batteries, and materials essential to high-performance manufacturing. Researchers were not only pursuing basic science. They actively researched materials for solar cells, water splitting, and solid ionic conductors to address issues directly relevant to society.
This translational focus, combined with a commitment to open science, has fostered collaboration with a diverse industry community, including companies such as BASF and Stellantis. “The main achievement has been the transition from software primarily built for professionals to practical tools that can be adopted in industrial settings.” This reflects a deliberate effort to bridge the gap between research and real-world implementation. MARVEL will end 12 years of operation with a closing event on July 9, but its legacy will be in establishing a digital ecosystem for materials science. Tools like AiiDA and Materials Cloud serve not only as research aids, but also as platforms designed with reproducibility, data sharing, and accessibility in mind, allowing computational workflows to be reliably integrated with automated experimental platforms.
