NVIDIA backs Europe’s AI-in-science push with open models, training resources, and startup compute pathways

NVIDIA announced a package of new commitments meant to broaden access to advanced AI tools for Europe’s scientific community, tying its efforts to an initiative launched by the European Commission at the AI in Science Summit in Dublin.

According to NVIDIA, the gap between “breakthrough technology” and day-to-day scientific progress is often less about model quality and more about what comes before research can start. For smaller labs, early-career scientists, and science-focused startups, the hardest work can include assembling research tooling, locating sufficient computing capacity, and developing the expertise to apply accelerated systems effectively.

The European Commission launched the AI in Science Enablers Initiative, presented as a shared framework connected to RAISE, the Resource for AI Science in Europe. NVIDIA said the initiative is intended to bring private-sector support closer to Europe’s research community, with Ekaterina Zaharieva, European Commissioner for Startups, Research and Innovation, arguing the ambition for AI in science depends on combining research strengths with industry and other partners through concrete commitments that give researchers access to tools, expertise, and resources.

NVIDIA’s pledge begins with what it describes as an open-source foundation. The company says it has released more than 250 open models and more than 250 open datasets, alongside over 1,400 open-source software projects covering areas including large language models, healthcare, robotics, climate science, physical AI, autonomous vehicles, and scientific research. NVIDIA says those models and datasets are available through Hugging Face and GitHub.

On top of that open ecosystem, the company points to NVIDIA Nemotron, described as a family of architectures with open weights, open source models, training data, and recipes. It also cites scientific software intended to accelerate work across biology, chemistry, molecular design, weather and climate, energy, engineering, physics, and materials science, including NVIDIA BioNeMo Agent Toolkit, NVIDIA PhysicsNeMo libraries, NVIDIA Earth-2 models, and NVIDIA ALCHEMI NIM microservices.

To address a second bottleneck, NVIDIA said its Deep Learning Institute has introduced a new open learning courseware hub. NVIDIA described the materials as freely downloadable and openly licensed, without enrollment, fees, or deadlines, and said the modules are modular so they can be adapted for university curricula, supercomputing center bootcamps, or developer community events. NVIDIA also characterized the hub as a bridge to its broader training and certification programs.

Access to compute and operational know-how is central to the company’s approach for smaller organizations. NVIDIA said it is collaborating with cloud infrastructure partners and European national AI factories, including sites in France, Spain, and Sweden, to create “routes” for startups and SMEs to obtain sovereign European compute and services. The company also highlighted its Inception program for startups, which it says provides compute resources and technical support to reduce the barrier to entry for teams seeking high-performance infrastructure alongside technological education.

NVIDIA said it is already working with several European AI factories in ways meant to make that startup access replicable. It cited AI2F in France as the first AI factory to collaborate with Inception for four months, adding that the company is reporting “strong results.” It also said Spain’s BSC AI Factory ran a pilot and is rolling out its first full accelerator program, with other European national AI factories showing interest in replicating a similar model in 2027. In Sweden, NVIDIA pointed to a collaboration with the Sweden AI Factory, one of Europe’s 19 EuroHPC-funded AI factories, saying it plans to open infrastructure and expertise to Inception startups and SMEs across life and materials sciences, autonomous systems, and gaming.

The company also pointed to free access at EuroHPC supercomputing centers such as CINECA and JSC, saying these facilities provide academia and startups with access to NVIDIA GPU infrastructure and services for open research. NVIDIA said it complements that access with practical support including office hours, bootcamps, and hackathons aimed at helping researchers fine-tune models on their own datasets. As an example, NVIDIA said it is co-organizing an AI hackathon in Europe this month with Austrian Scientific Computing, the Leibniz Supercomputing Centre, the High-Performance Computing Center Stuttgart, and the Cyprus Institute, for Pharos-CY.

NVIDIA’s broader claim is that the effort is aligned with a major buildout of AI infrastructure in Europe. The company said it is powering what it called Europe’s most ambitious expansion of AI infrastructure to date, with 35 supercomputers currently in development across the continent, describing it as the largest single-year buildout of supercomputing capacity Europe has seen. NVIDIA said that these systems would equip over 3 million researchers with next-generation infrastructure, accelerating discovery in fields ranging from climate science and drug development to materials research. What NVIDIA does not specify, at least in its announcement, are the exact allocation sizes, selection criteria, or service-level details for compute access, nor does it provide timelines for when each initiative will become fully operational across each country.