Christophe Fouquet’s criticism of the EU’s evolving Chips Act is less an attack on industrial policy than a challenge to where sovereignty can realistically be exercised in a global AI economy. As the UK and France shift towards sovereign compute, trusted hosting, and infrastructure control, Europe faces a harder question: not how to control supply chains, but where leverage still exists.
In sum – what to know:
ASML warning – the EU’s semiconductor controls are wrongheaded, says ASML chief; no region controls enough of the AI supply chain to direct outcomes through chip allocation alone, and Europe might do better to loosen up.
Local strategies – As lately, the UK and France are pursuing a more pragmatic (telco-aligned) version of sovereign AI, focused on compute infrastructure, data governance, trusted hosting, niche specialties, and national assets.
Space shadow – Even as SpaceX builds an orbital infra layer, sovereignty is not disappearing; it is becoming distributed across terrestrial, subsea, airborne, and space systems, making control of key interfaces crucial.
On the face of it, Christophe Fouquet’s warning to the EU about chip interventions reads like another Brussels takedown by the private tech sector. But it points to something more structural – that Europe is still trying to govern the new AI era with industrial-policy tools designed for a world where supply chains were linear, and maybe even controllable. That world no longer exists. And the gap between sledgehammer regional policy intent and amorphous global system architecture is widening so fast that ASML, Europe’s most valuable tech company, is firing shots across the bows: catch a sail, he sort-of says, or the whole ship’s going down.
The issue is this so-called ‘Chips Act 2.0’, an evolution of the Chips Act, which moves from subsidy-led capacity building to a framework for supply-chain coordination in times of shortage. The instinct is good. Semiconductors are the building blocks of modern tech, behind everything – from data centres to power grids to telecom networks to defence systems. But Fouquet’s argument, as reported in the FT, is not just that Europe lacks leverage in this base layer, but that it lacks control in every layer – of the whole interlinked AI supply chain. You can’t just tell your best EUV maker, say, to prioritize your best aerospace outfit because the White House says something silly.
Components come from everywhere – and, as it goes, the EU only accounts for one percent of EUV sales anyway. Which is the case with Dutch firm ASML, which is a clear example of a global choke point in advanced lithography. But its EU leverage does not translate into total supply-chain control; critical EU sectors also have to contend with externally-controlled and hyper-competitive memory, cloud, and packaging supplies, plus whatever else – all driven by global capital markets. Without a sovereign stake in every layer of the AI supply chain (which no country has), any attempts to steer chip flows are impotent. But we should not get carried away, even if the warning from ASML stands.
Yes, the new Chips Act looks like a sledgehammer mechanism to rubber-stamp ‘sovereignty’. But its objective is plain, even if it reaches into layers where control is diffuse, and Europe’s broader play is subtler, actually – shifting up the stack. Approaches by the French and UK governments, discussed here recently, are fairly representative of how big telcos are geared around sovereign AI – and of how ASML and the rest see it. They are quite pragmatic, seeking to apply jurisdictional controls around residency, trust, and access, and to make good on their national assets – lately, nuclear power (France) and supercomputers (UK) – and set their big industrial guns as corporate AI landlords.
The UK comparison is informative: its development strategy, outside of its landlord bid, is to strengthen its sovereign capability in parts of the AI supply chain where it has niche strengths, such as in chip design, GPU networking, R&D clusters, early-stage startups. Which might generate real companies and influence over time, and is the kind of stuff Fouquet advocates in the FT. But the UK is not mandating structural control of global chip supplies, like the EU. And it is worth noting how critical this lighter-touch sovereign AI role is – especially as SpaceX floats on the stock exchange with an extra-terrestrial full-stack infrastructure play, heaps of debt, and brand new shareholder riches.
Because on the face of it, SpaceX looks like an AWS of the skies, like it is going to hyper-scale the hyperscale model for the AI age – operating data centres and satellite networks 150 kilometres above cloud-level in low-Earth orbit, to beam AI data down into any jurisdiction. But that is not true. Yes, it gets around a bunch of familiar planning and connectivity bottlenecks, traded for a unique bunch of its own (latency, transport, maintenance, economics). But it is not a cloud compute layer in orbit. It is a parallel network, in the main, which integrates rockets and satellites, and fills a hole in rural coverage on the ground, and might some day offer niche segmented compute services workloads.
Sovereignty does not escape the system, however. Space-based networks still depend on ground stations, spectrum authorisations, regulatory frameworks, gateways and devices, Earthbound supply chains. They extend reach and resilience; they don’t eliminate jurisdictional controls. So sovereignty, itself, gets distributed too – across terrestrial, subsea, airborne, and orbital systems. Fouquet’s warning is that the EU is still looking to impose hierarchy and control in a system that is increasingly horizontal and layered. Sovereignty is not about controlling stacks or supply chains. It is about identifying where control is still meaningful, and building leverage at those points.
