At the VivaTech trade show in Paris, Germany’s space and digital minister Dorothee Bär delivered a blunt message to the United States: “Without us, it cannot be done.” She was referring to NASA’s reliance on European technology for its moon missions, a dependency that mirrors challenges in other high-tech sectors, including artificial intelligence hardware.

The European Service Module: NASA's Indispensable Partner

As reported by Politico, Bär pointed out that the European-built Service Module (ESM) is a critical component of the Orion spacecraft. The ESM provides propulsion, power, thermal control, and life support — functions without which the Artemis lunar missions could not proceed. This is not bravado; it's a factual statement of technological interdependence between allies.

Hidden Dependencies in the AI Supply Chain

Space exploration offers a lens through which to examine another domain where sovereignty is paramount: artificial intelligence. Organizations deploying LLMs on-premise must navigate a supply chain that is equally concentrated. Advanced GPUs essential for training and inference are produced by a narrow set of manufacturers, and leading-edge fabrication is geographically clustered. Software ecosystems remain proprietary, creating similar points of dependency.

Like the ESM, certain components in AI infrastructure have no easy substitutes. For European enterprises and public bodies, the push toward self-hosted solutions is not just about data privacy — it is about securing access to critical hardware and avoiding geopolitical disruptions. The recent history of export controls has shown that the flow of technology cannot be taken for granted.

On-Premise AI: Strategic Autonomy and Trade-Offs

Bär’s statement resonates beyond space policy. It reminds technologists that Europe is not merely a consumer but also a supplier of irreplaceable expertise. In the AI context, this could mean expanding European capabilities in accelerators, cooling systems, or inference software. Yet, for those planning on-prem deployments, the calculus involves balancing TCO with reliability of supply.

Choosing to self-host an LLM means regaining control over data pipelines and compliance, but it also locks organizations into a hardware ecosystem that may be subject to sudden policy changes. While cloud providers abstract away hardware procurement, they introduce their own dependencies. The lesson from space is clear: strategic autonomy requires a diversified approach. Getting to the moon — or deploying a resilient AI system — depends on managing mutual reliance, not eliminating it.