Digital Sovereignty

The subject is everywhere in speeches, almost nowhere in decisions. The essay distinguishes three levels too often conflated. The hyperscaler cloud (AWS, Azure, GCP) plays out at continental scale, and plays out badly there: neither Gaia-X nor national players have produced a large-scale alternative. National public infrastructures and sovereign AI models are under active construction: the Swiss Government Cloud (a commitment credit of CHF 246.9 million, rollout 2025-2032, reserved for the public sector), Apertus, launched in September 2025 by EPFL/ETH/CSCS. Everyday use, finally, is where sovereignty is tested in material terms. The decisive distinction sets the strategy of substitution (doomed to fail against the hyperscalers) against the strategy of complement (which can succeed on ground the hyperscalers do not serve). For Valais, a large general-purpose cantonal data centre makes little sense; one or two targeted specialised nodes can: cantonal medical data, models trained on Walliserdeutsch and Franco-Provençal, archiving of intangible heritage listed by Unesco. The realistic wager combines migration to the SGC, partnerships with Swiss operators (Genedis, Infomaniak, Exoscale), and targeted projects justified by their specificity.