Show HN: Can Europe train a frontier AI model on the compute it owns?

A European research project models whether existing public compute across EuroHPC supercomputers and national AI Factories could train a frontier-class AI model by 2028 through federated learning, potentially 5 years faster than waiting for new gigawatt datacenters to connect to the grid. The analysis concludes this is feasible as a stopgap solution, though it relies on assumptions about grid connection timelines and faces practical challenges around coordinating heterogeneous, shared computing resources.