The sovereignty argument — and its limits

• Application-level sovereignty improves. The core analyst argument is that a model you run on your own servers is, in a narrow but meaningful sense, more sovereign than one rented from California . Open-weight models reduce dependency at the application tier because organisations deploy them on their own infrastructure and retain full control over data, tuning, and uptime .
• Reducing concentration risk from a single vendor. European leaders increasingly see excessive dependence on a small group of non-European providers — particularly U.S. cloud and AI firms — as a strategic vulnerability . Chinese open-weight models offer a diversification option that weakens the monopoly of any one external supplier .
• But ecosystem-level dependencies may shift, not disappear. EU policy analyses warn that relying on models and tooling developed principally in China can create a new form of broader ecosystem dependency . The European Commission's own Technological Sovereignty Package (June 2026) frames sovereignty as derisking and diversifying supply chains, not eliminating them — meaning that importing Chinese model weights is consistent with the strategy as long as it reduces concentration in any single external source .
• A geostrategic hedge. Some European analysts frame the choice as a buffer against an unpredictable Washington: "excellent and cheap Chinese models could reduce our dependence on an unpredictable US administration" . The argument is that Europe is better off with multiple external AI sources than wholly reliant on one.

Bottom line: The pragmatic case (cost + performance + control) is driving adoption today, while the sovereignty case rests on reducing U.S. hyperscaler dependency through local hosting. The open question is whether replacing one external dependency with another constitutes genuine sovereignty — most policy experts argue it is a net improvement in diversification, not a complete solution .

The sovereignty argument — and its limits

• Application-level sovereignty improves. The core analyst argument is that a model you run on your own servers is, in a narrow but meaningful sense, more sovereign than one rented from California . Open-weight models reduce dependency at the application tier because organisations deploy them on their own infrastructure and retain full control over data, tuning, and uptime .
• Reducing concentration risk from a single vendor. European leaders increasingly see excessive dependence on a small group of non-European providers — particularly U.S. cloud and AI firms — as a strategic vulnerability . Chinese open-weight models offer a diversification option that weakens the monopoly of any one external supplier .
• But ecosystem-level dependencies may shift, not disappear. EU policy analyses warn that relying on models and tooling developed principally in China can create a new form of broader ecosystem dependency . The European Commission's own Technological Sovereignty Package (June 2026) frames sovereignty as derisking and diversifying supply chains, not eliminating them — meaning that importing Chinese model weights is consistent with the strategy as long as it reduces concentration in any single external source .
• A geostrategic hedge. Some European analysts frame the choice as a buffer against an unpredictable Washington: "excellent and cheap Chinese models could reduce our dependence on an unpredictable US administration" . The argument is that Europe is better off with multiple external AI sources than wholly reliant on one.

Bottom line: The pragmatic case (cost + performance + control) is driving adoption today, while the sovereignty case rests on reducing U.S. hyperscaler dependency through local hosting. The open question is whether replacing one external dependency with another constitutes genuine sovereignty — most policy experts argue it is a net improvement in diversification, not a complete solution .