Switzerland's uphill climb to AI sovereignty (www.swissinfo.ch)

Switzerland has poured millions into domestic AI infrastructure — notably the “Alps” supercomputer (ranked eighth globally) and Apertus, Europe’s first fully public, open large language model trained on more than 1,000 languages and dialects — to reduce reliance on US and Chinese tech. Recent shocks (an Amazon cloud outage and security concerns about Microsoft in the Swiss army) and public unease around data privacy (close e-ID vote) have amplified calls for “sovereign” tech. Despite these projects, officials and researchers warn true sovereignty is elusive: Swiss public funding is modest (about CHF100M for Alps plus CHF20M through 2028 for the Swiss AI initiative), maintenance and energy costs add ~CHF10M/yr, and open-source models rely on voluntary global contributors rather than a guaranteed local support chain. The technical constraints are stark: Nvidia GPUs power roughly 90% of the world’s AI models and US export controls on advanced chips directly limit access, while ETH Zurich projects Switzerland will need 20–100× more compute within a decade. Chips, power, and data remain concentrated in a few countries, making full stack autonomy infeasible for a small state. As a result Swiss experts back cooperative solutions — a “federation of sovereign AIs” or an “Airbus for AI” consortium — arguing alliances and shared infrastructure are the pragmatic path to reduce dependency and retain cultural and operational control over AI.