Rethinking Europe's AI Strategy: Leadership Not Enough

So Europe says it wants AI leadership. The Tony Blair Institute argues that this will boost competitiveness and security — and they're mostly right on the destination. The problem is the route map skips the toll booths: shared sovereignty, abandoned pet projects, and a real industrial plan instead of aspirational language.

Let’s start with the part the piece gets right. Treating AI as a strategic asset for competitiveness and security is overdue, not overblown. A strong AI stack can lift productivity, sharpen defense capabilities, and raise the quality of public services. The math doesn’t lie: when you sit closer to the core technology, you capture more of the value chain — design, integration, standards, and services.

But leadership is not a speech act.

You don’t “declare” technological leadership into existence any more than you declare a currency strong and make it so. Leadership is the outcome when three things line up: capital that can afford to be wrong, talent that concentrates, and states willing to focus on a few big bets instead of sprinkling money like confetti.

This is where the Blair Institute piece gets fuzzy. It treats leadership as a clean, positive-sum aspiration for Europe. It isn’t. If some European AI hubs actually scale to global relevance, others will lose out. Talent will move. National champions will be displaced. Right now, European governments talk about leadership as if every member state gets a trophy.

They won’t.

On the politics, the article underestimates how hard it is to turn “Europe” into a single tech market. Member states still guard labor laws, procurement rules, and tax regimes as instruments of domestic politics. That means a breakthrough model, lab, or application developed in one jurisdiction doesn’t automatically spread across the rest. You get pockets of excellence; you don’t get continental scale.

That’s not just an economic problem. The piece links leadership to security, but security requires more than having good tools somewhere inside your borders. It needs shared operational standards, compatible systems, and procurement that doesn’t fracture into 27 slightly different “nationally adapted” versions. You can’t buy interoperability with communiqués.

From my Goldman days, I remember how everyone claimed they wanted “scale,” right up until it meant closing local desks and killing pet products. Europe is in that same denial phase — the point where leaders say “single market” but still design rules to protect local players, local jobs, and local budgets.

The article is on firmer ground when it highlights feasibility, cost, and dependence on non-European actors. Those are the right stress points. But it still reads too much like a vision paper and not enough like a term sheet.

Take talent. High-end AI researchers and engineers cluster where they can work on frontier problems with serious compute, serious data, and serious upside. If Europe nudges resources without concentration, it will just create a few stronger magnets — say, in a handful of cities — while entrenching a wider periphery of “follower” regions. That’s politically uncomfortable, but pretending every capital can be a global AI hub is fantasy.

Capital is the same story. Public money that’s sliced by geography, sector, and politics produces dozens of small pilots with nice PDFs and no global footprint. We’ve seen this movie with other digital initiatives. If Europe wants leadership, it has to accept that meaningful sums will flow into projects that fail visibly — and that some member states will feel they are paying for someone else’s flagship.

Then there’s dependence. The Blair Institute is right to link AI leadership to strategic autonomy, but let’s be real: you can’t outsource key pieces of your stack and then claim autonomy with a straight face. If core compute, foundational models, or security tooling are controlled elsewhere, you’re effectively renting sovereignty. That might be a reasonable trade in some domains, but then call it what it is: managed dependence, not leadership.

The counter-argument here is familiar: central direction kills experimentation. A heavy-handed industrial strategy risks locking in today’s incumbents and suffocating exactly the startups everyone says they want. There’s truth there. Centralized planning of “the winners” is how you end up with bloated national champions that can’t compete outside their home turf.

But this is a false binary. You can centralize where it actually creates shared infrastructure and keep the edges chaotic. Think of it as a two-layer game: a coordinated push on foundational capabilities — compute access, sovereign and shared data infrastructure, trusted evaluation and verification mechanisms — plus a messy, local market for applications and business models. Public money either becomes a series of scattered grants that no one remembers, or a set of focused bets that create gravity wells around which private actors orbit.

The article hints at this but never gets specific. That’s the gap. Saying “leadership will boost competitiveness and security” is directionally fine; it just doesn’t help distinguish between a strategic asset and a vanity project. Building a flagship AI research center that quietly becomes a consultancy for local ministries? Vanity. Building widely used verification tools that global players adopt because they lower compliance and security frictions? Strategic.

History offers a useful reminder. When Airbus was stitched together, it wasn’t because Europe had a generic “aviation ambition.” It picked a segment where dependence was strategically uncomfortable, accepted cross-border governance headaches, and pushed enough capital and political will into the project that global players had to take it seriously. The internal fights were ugly, but the outcome was a genuine counterweight, not a press release.

Expect the AI version of that to be worse, not better: higher stakes, faster cycles, and more visible losers. If Europe follows the Blair Institute’s call without confronting those trade-offs, “AI leadership” will stay as a slogan. If it does confront them, the first real sign won’t be a strategy document — it will be a few governments taking heat for shifting money, talent, and control into places their voters didn’t vote for.