Why AI factories matter for Bigpicture (and why now is the moment to engage)

To understand what AI factories are, and why they matter for the future of Bigpicture, we spoke with Tommi Nyrönen (CSC – IT Center for Science, Finland), who works closely with the LUMI AI Factory initiative and is deeply involved in the technical foundations of Bigpicture.

What are AI factories

The term AI factory can sound abstract or even intimidating. Tommi is the first to acknowledge it. “It’s a marketing word,” he says. “But behind the label is something very practical: a one-stop shop where data and supercomputing come together so you can train powerful AI models.”

Unlike traditional HPC environments, AI factories are being built with a new objective:

• support industry and SMEs
• enable data-intensive AI development
• provide secure, high-performance environments
• reduce dependency on non-European cloud providers
• create a trusted European ecosystem for sensitive data

For Bigpicture, that combination is uniquely relevant.

The opportunity for Bigpicture: data meets compute

Bigpicture was never designed to become a compute provider. It is a data repository, the largest of its kind for digital pathology in Europe. But data alone is not enough. To unlock value, partners need ways to train models, run large-scale experiments and explore new computational approaches.

This is where AI factories come in. “We can build this incredible repository,” Tommi explains, “but to get to the applications, to the foundation models, or to the analysis tools, we need access to serious computing power. And we need environments that meet the same privacy and security standards Bigpicture requires.”

AI factories are being designed exactly for that purpose: high-performance compute + secure processing + European data governance. If aligned well, the collaboration becomes mutually reinforcing.

Bigpicture brings the high-value data and domain expertise. AI factories bring the compute and the technical environment to use it. Together, they can drive innovation in digital pathology that neither could achieve alone.

Why this matters now

Several developments make this moment particularly important:

1. The regulatory landscape is shifting fast

With the European Health Data Space adopted in March 2025, clearer requirements for secure processing environments are emerging. AI factories are now working to meet these specifications. “The question is whether these environments will satisfy GDPR and EHDS expectations,” Tommi says. “Once they do, they become trustworthy for data holders like Bigpicture.”

In parallel, the EU Data Union strategy has set new priorities for strengthening Europe’s data ecosystem. Its goals, scaling up access to data for AI, streamlining data rules, and reinforcing Europe’s position in international data flows, create an environment where trusted, high-quality datasets like Bigpicture’s become increasingly important. AI factories are expected to play a key role in operationalizing these objectives, making alignment between Bigpicture and AI factory initiatives both timely and strategic.

2. Europe wants digital sovereignty

US-based cloud providers offer powerful compute but create long-term dependency. AI factories are Europe’s answer, a strategic effort to develop sovereign infrastructure for AI development.

3. Bigpicture is reaching the phase where compute becomes essential

If Bigpicture wants to enable large-scale analysis, for example, training new foundation models from millions of slides, the consortium will need compute capacity far beyond what any individual partner can maintain. “This level of computation is simply not possible without a very big machine,” Tommi says. AI factories provide that machine.

Where we stand today

Several partners in Bigpicture already have ties to AI factory initiatives. CSC, for example, is deeply involved in the development of the LUMI AI Factory, one of the most powerful supercomputers in Europe.

This puts Bigpicture in a strong position: 1) We can follow developments early, as the technical environment is still being shaped. 2) We can provide feedback on data management requirements unique to digital pathology. 3) We can act as a real-world use case, helping AI factories adapt their environments to biomedical data needs.

“Bigpicture is a credible use case,” Tommi emphasizes. “It’s public–private, it’s multi-country, and the dataset is extremely valuable. This is exactly the kind of collaboration AI factories are meant for.”

A shared opportunity for Europe

AI factories may sound abstract or even intimidating at first, but for Bigpicture they represent a clear and timely opportunity. As Tommi explains, the synergy is straightforward: “We need them, and they need us. With the right alignment, both sides can benefit enormously.”

For Bigpicture, these collaborations could provide sustainable access to compute, offer a trusted European environment for processing sensitive medical data, and enable use cases that would otherwise be out of reach. While AI factories may or may not be cheaper than commercial cloud services, cost is only one part of the equation, the real value lies in having suitable environments where data, compute and governance are aligned. More importantly, they offer a pathway to realize the full ambition of the project: enabling large-scale applications in digital pathology and AI that rely on both high-quality data and powerful computation.

Tommi is optimistic: “If we can bring millions of slides together and enable the development of new, large-scale AI models, that would be a major achievement. Bigpicture has the data, AI factories provide the compute, and with expertise on both sides, together we can deliver something none of us could achieve alone.”

As the AI ecosystem rapidly evolves, this is Bigpicture’s moment to position itself not just as a repository, but as a cornerstone for data-driven innovation in pathology. That future begins by engaging now and helping shape the environments where tomorrow’s models will be built.