What lands for me here is the shift from “AI as capability” to AI as infrastructure exposure. Framing it that way forces a different kind of governance question: not “how do we use AI?” but “which long‑lived dependencies, jurisdictions, and power grids are we quietly wiring our institutions into?” In most organisations, AI is still treated as an IT or innovation line item, so the real infrastructure choices get made piecemeal by vendors and regulators instead of boards. Your Xinjiang example makes it very hard to keep pretending this is a tools conversation rather than an infrastructure and sovereignty conversation.
Glad this observation resonates with you! And I like the way you framed Europe’s position as the “messy middle”, it captures the situation very well.
On your question about the path forward: I don’t think the answer is a clean break from US or Chinese tech stacks, but a deliberate strategy to build optionality and bargaining power. This means:
1. Architectural optionality: designing for true multi-cloud and multi-model interoperability, so European firms and states are never locked into a single vendor, API, or governance regime.
2. Infrastructure optionality: sustained investment in domestic and pan-European compute and energy, not necessarily to replace hyperscalers, but to ensure credible fallback capacity and price-setting power.
3. Standards and regulatory leverage: Europe has always been built on process knowledge: the ability to formalise, standardise, and industrialise complexity, turning building blocks developed elsewhere into systems that are more precise, reliable, and governable. In the AI era, this means shaping the layers above raw compute and models — the interfaces, safety regimes, validation methods, and integration of energy, data, and governance into something truly industrial-grade.
4. Institutional coordination: treating compute, models, and energy as strategic infrastructure (like ports or telecoms), with long-term public-private planning rather than fragmented procurement.
If the US leads in capital-driven model development and China in state-coordinated infrastructure, Europe’s strategic role is to become the end to end systems architect and standard-setter.
Another area that is not discussed as much but incredibly important is world model and robots. If Europe’s comparative advantage is turning complexity into governable systems, then robots are where that capability becomes decisive. Models and compute remain upstream dependencies, but robots sit at the point where AI meets labour, safety regulation, liability, energy usage, and physical infrastructure (that Europe still has an edge on).
What lands for me here is the shift from “AI as capability” to AI as infrastructure exposure. Framing it that way forces a different kind of governance question: not “how do we use AI?” but “which long‑lived dependencies, jurisdictions, and power grids are we quietly wiring our institutions into?” In most organisations, AI is still treated as an IT or innovation line item, so the real infrastructure choices get made piecemeal by vendors and regulators instead of boards. Your Xinjiang example makes it very hard to keep pretending this is a tools conversation rather than an infrastructure and sovereignty conversation.
Glad this observation resonates with you! And I like the way you framed Europe’s position as the “messy middle”, it captures the situation very well.
On your question about the path forward: I don’t think the answer is a clean break from US or Chinese tech stacks, but a deliberate strategy to build optionality and bargaining power. This means:
1. Architectural optionality: designing for true multi-cloud and multi-model interoperability, so European firms and states are never locked into a single vendor, API, or governance regime.
2. Infrastructure optionality: sustained investment in domestic and pan-European compute and energy, not necessarily to replace hyperscalers, but to ensure credible fallback capacity and price-setting power.
3. Standards and regulatory leverage: Europe has always been built on process knowledge: the ability to formalise, standardise, and industrialise complexity, turning building blocks developed elsewhere into systems that are more precise, reliable, and governable. In the AI era, this means shaping the layers above raw compute and models — the interfaces, safety regimes, validation methods, and integration of energy, data, and governance into something truly industrial-grade.
4. Institutional coordination: treating compute, models, and energy as strategic infrastructure (like ports or telecoms), with long-term public-private planning rather than fragmented procurement.
If the US leads in capital-driven model development and China in state-coordinated infrastructure, Europe’s strategic role is to become the end to end systems architect and standard-setter.
Another area that is not discussed as much but incredibly important is world model and robots. If Europe’s comparative advantage is turning complexity into governable systems, then robots are where that capability becomes decisive. Models and compute remain upstream dependencies, but robots sit at the point where AI meets labour, safety regulation, liability, energy usage, and physical infrastructure (that Europe still has an edge on).