Supercomputers, Sovereignty, and the Future of AI Research

When European Commission President Ursula von der Leyen stood before the AI for Good Summit in Paris this February, her message was clear: Europe’s AI future will be grounded in trust, research, and public good. In a reaffirmation of Europe’s academic strengths and priorities in the AI space, she spoke of Europe’s unique strength in scientific cooperation, of public supercomputers dedicated to researchers, and of investing “not just in technology, but in purpose.”

For anyone working in academia or public-sector innovation, it was a timely reminder that their progress and challenges are being recognised where it matters. At PCG, our role is often a quiet one: helping research institutions navigate technical complexity and bring together the right tools and services to support their goals.

Whether it's guiding architectural decisions or ensuring compliance with frameworks like OCRE, we work alongside teams to turn ambition into practical, working solutions. One of the privileges of developing cloud solutions for the public sector is that we've seen firsthand how AI is reshaping research environments—not in abstract, but in practice.

Infrastructure, But Not As You Know It

When people hear the phrase “public infrastructure,” they often think roads, bridges, transport. But in the age of AI, another kind of infrastructure is quietly taking centre stage—one that supports data access, collaborative research, and computational experimentation at scale.

And this shift isn’t just happening behind closed doors in elite institutions. Thanks to initiatives like the OCRE Framework (Open Clouds for Research Environments) and the EU’s newly announced AI gigafactories, academic institutions across Europe are gaining access to the kind of computing power and tools that were once reserved for tech giants. From real-time robotics simulation to large-scale document analysis, AI is becoming part of the academic toolkit.

Making Research Discoverable – GRNET’s Conversational Search

One of the biggest challenges in academic publishing isn’t creating knowledge—it’s finding it. For Greece’s GRNET (National Infrastructures for Research and Technology), the goal was to unlock access to vast archives of research papers and public knowledge, making them searchable in natural, human terms.

Working together, we built a Retrieval-Augmented Generation (RAG) system that allows users to ask questions in plain language and receive intelligent, cited answers drawn from thousands of research documents. Under the hood, it uses a blend of AWS Bedrock, Amazon OpenSearch Service, Amazon S3, AWS Lambda, and Titan embeddings, all running seamlessly on AWS infrastructure.

As a foundational tool for academic discovery, the system delivers:

• Semantic search that understands intent
• Fast, conversational access to complex academic material
• A scalable model that can expand with new domains and languages

This isn’t just a better search engine—it’s a new kind of public infrastructure for knowledge. It also helps research institutions operate more efficiently, reduce the manual burden of information retrieval, and create new value from their existing data assets. Transforming Data into Insights: Generative AI Solutions.

AI Experiments at Scale – RCCL’s Robotic Simulations

Meanwhile, over at the National and Kapodistrian University of Athens, the Research Centre on Interactive Media, Smart Systems and Emerging Technologies (RCCL) needed something slightly different: a way to run large-scale AI experiments for robotics and cognitive systems without getting bogged down in hardware bottlenecks.

We helped RCCL design a flexible cloud-based research environment that could scale with the complexity of their simulations. Using Amazon SageMaker, Amazon EC2 Spot Instances, and Amazon FSx for Lustre, researchers could spin up GPU-heavy workloads as needed—without burning through budgets or waiting on local infrastructure.

This new environment has:

• Reduced the time from idea to experiment
• Lowered costs through elastic compute
• Enabled real-world modelling of human-AI interaction

From simulating robotic behaviours to exploring cognitive architectures, RCCL’s researchers now work in an environment built for agility and scale. The setup also delivers operational benefits—minimising idle hardware costs, simplifying experimentation workflows, and enabling research teams to deliver results faster with fewer infrastructure constraints.

Cloud Strategy with a Social Impact

These examples aren’t isolated wins—they’re part of a broader pattern. Across Europe, public institutions are beginning to treat AI not just as a research subject, but as a tool to solve practical problems, deliver organisational goals, and demonstrate measurable value from research and innovation efforts. And cloud platforms—when deployed thoughtfully—are enabling that shift.

Initiatives like OCRE make it possible for universities and research bodies to access pre-approved, secure cloud services without wrangling procurement headaches. EU investment in AI infrastructure means more researchers can use large language models, run simulations, and share data—ethically, securely, and at scale.

Although we're definitely not unbiased on the matter, it's definitely the case that working with an experienced partner can help ensure these capabilities are deployed effectively. At PCG, that means quietly supporting teams behind the scenes—helping them design cloud environments, integrate AI tools, and stay aligned with funding and compliance frameworks like OCRE. Above all, it's a collaborative effort—and one that’s helping to build a more connected, capable, and forward-looking research landscape across Europe.