A key enabler of AI is infrastructure that is secure, accessible, scalable, and state-of-the-art. In the Netherlands, such compute capacity remains limited and fragmented, despite major initiatives. Power constraints in most regions limit hardware expansion. Institute-specific clusters fragment resources and duplicate effort. National facilities like the Dutch Supercomputer (Snellius) for research and the upcoming Netherlands AI Factory (NLAIF) are steps forward, yet as siloed systems these are not enough for the growing diversity of AI workloads.
Federation, i.e. sharing compute, data, and expertise across institutions, offers a path forward. However, effective federation for AI requires more than connectivity: it depends on orchestration, software portability, data access, and alignment between researchers, research software engineers, and infrastructure providers.
This workshop connects these three communities to jointly explore when federation enables AI in practice, and when it fails or becomes counterproductive. It produces a practitioner-oriented guide that identifies when federation is warranted for AI, what breaks in practice, and which workflow patterns are reusable. These outcomes will form a foundation for collaborative proposals on open, FAIR-aligned AI compute ecosystems.
The workshop will take place place from Monday 15th to Friday 19th of March, 2027.
