NVIDIA Expands Scientific Access in Europe, blogs.nvidia.com says

European Commission and NVIDIA Launch Accelerated Compute Initiative in Dublin

Compute bottlenecks and hardware scarcity across smaller labs and early-career science teams met a coordinated private-public response when NVIDIA expanded scientific access across Europe through open learning frameworks, sovereign cloud compute integrations, and institutional partnerships, blogs.nvidia.com reported from the AI in Science Summit in Dublin.

The Tech TL;DR:

• Infrastructure Expansion: NVIDIA is powering an infrastructure rollout of 35 new supercomputers across Europe, targeting over 3 million researchers.
• Open-Source AI Models: More than 250 open models, 250 datasets, and 1,400 software projects have been deployed to Hugging Face and GitHub to support agentic reasoning and domain-specific libraries.
• Sovereign Compute Integration: National AI factories in France, Spain, and Sweden are connecting startups and small- and medium-sized enterprises (SMEs) to local high-performance computing resources.

For early-stage researchers and computational laboratories, the friction of accelerated computing has historically centered on tooling assembly and infrastructure overhead. At the Dublin summit, the European Commission introduced the AI in Science Enablers Initiative, tying private-sector resources directly into RAISE (Resource for AI Science in Europe). Ekaterina Zaharieva, European Commissioner for Startups, Research and Innovation, noted that realizing Europe’s scientific AI ambitions requires uniting research communities with industry partners through concrete commitments of tools and expertise, as reported by blogs.nvidia.com.

Open Models, Datasets, and Specialized Software Toolchains

To bypass licensing constraints that typically restrict model fine-tuning, NVIDIA’s contribution opens with a repository release comprising over 250 open models, 250 datasets, and 1,400 open-source software projects hosted on GitHub and Hugging Face. These weights and datasets target high-throughput computational workloads spanning large language models, healthcare, robotics, climate science, physical AI, and material sciences.

Under the hood, the released framework features specialized development kits designed for domain-specific inference. Researchers can build agentic reasoning workflows using NVIDIA Nemotron, alongside specialized libraries such as the NVIDIA BioNeMo Agent Toolkit for molecular design, NVIDIA PhysicsNeMo libraries, NVIDIA Earth-2 atmospheric simulation models, and NVIDIA ALCHEMI NIM microservices.

# Example cURL request to invoke an NVIDIA NIM microservice for molecular simulation
curl -X POST "http://localhost:8000/v1/generate" 
     -H "Content-Type: application/json" 
     -d '{
       "model": "bionemo-agent-toolkit",
       "prompt": "Simulate folding kinetics for target protein sequence",
       "max_tokens": 512
     }'

NVIDIA Collaborates with European National AI Factories

To resolve this, NVIDIA is collaborating with infrastructure partners and national AI factories across the continent. In France, AI2F completed a four-month integration with the NVIDIA Inception program, producing measurable performance gains for participating startups. Spain’s BSC AI Factory, following a successful pilot phase, is scaling up its full accelerator program in preparation for broader replication across European national AI factories by 2027, according to blogs.nvidia.com.

Concurrently, the Sweden AI Factory—operating as one of Europe’s 19 EuroHPC-funded AI facilities—opened its high-performance computing infrastructure to Inception startups and SMEs focusing on life sciences, material research, and autonomous systems. Academic researchers and early-stage ventures can also access GPU infrastructure without direct user fees via EuroHPC supercomputing centers such as CINECA and JSC. These compute allocations are supported by technical office hours, hackathons, and joint engineering bootcamps organized alongside institutions like the Leibniz Supercomputing Centre and the High-Performance Computing Center Stuttgart.

Deep Learning Institute Launches Open-Source Courseware Hub

To address skills gaps in parallel computing architectures, the NVIDIA Deep Learning Institute launched a modular, open-source courseware hub. These curricula cover accelerated computing and deep learning fundamentals without enrollment fees or completion deadlines. Designed for integration into university syllabi, supercomputing bootcamps, and developer community workshops, the courseware establishes a direct learning bridge to advanced NVIDIA certification pathways.