Sam Winter-Levy is a fellow at the Technology and International Affairs Program at the Carnegie Endowment for International Peace, where Anton Leicht is a visiting scholar.
The middle powers in the West — European countries and Canada — are increasingly hoping to chart a course through the artificial intelligence revolution independent of China and the United States.
At the Munich Security Conference last month, German Chancellor Friedrich Merz declared the rules-based order dead and laid out a road map for “a strong sovereign Europe.” Ursula von der Leyen, president of the European Commission, declared that “our digital sovereignty is our digital sovereignty.” French President Emmanuel Macron, echoing Canadian Prime Minister Mark Carney, called for “de-risking vis-à-vis all the big powers.” Days later, at an AI summit in New Delhi, the same impulse surfaced in explicitly technological terms. Delegates from over 100 countries discussed sovereign AI infrastructure and how to prevent the technology’s benefits from being captured by a handful of American firms. On paper, it was the most ambitious assertion of middle-power independence in years.
But all this hopeful and ambitious talk ignores a painful reality: It’s too late in the AI race for the middle powers to play catch-up. The leaders calling for autonomy in Munich and New Delhi are doing so at a moment in which the defining technology of the next decade is controlled by the very powers they seek independence from.
In other sectors, settling for developing homegrown second-best technology might be a viable strategy for preserving sovereignty. In AI, that’s a riskier bet. If the gap between cutting-edge and second-tier systems continues to widen, and especially if advanced AI accelerates scientific research and industrial innovation, access to the best systems could become decisive for economic growth. A Council of Economic Advisers report released earlier this year projects a growing divergence between the U.S. and the rest of the world, with America poised to reap the productivity gains of deploying advanced AI domestically.
The national security risks may be starker still. Middle powers face hostile states that will aggressively deploy leading AI systems; defending against them may require systems just as capable. Any country without access to the cutting edge could face economic and strategic marginalization.
Faced with this dilemma, middle powers may be tempted to rush for the frontier through a concerted joint effort — pooling resources to build sovereign AI systems. A recently announced Franco-Germaninitiative with support from the European Commission, for example, aims to funnel investment into frontier AI research. But the answer isn’t a sovereignty moon shot.
Similar joint projects have already failed to restore European sovereignty over strategic tech sectors like cloud computing and social media. AI will be harder still. France’s Mistral and Canada’s Cohere are arguably the best AI companies founded outside the great powers, but their models remain well behind the most capable systems in the world. Closing that gap demands concentrations of talent found only in Silicon Valley — and vast capital, energy and computational power, all built on advanced chips that only U.S.-dominated supply chains can provide.
Today’s AI infrastructure projects determine tomorrow’s catch-up capacity, and middle powers lag far behind. The European Union’s up to five planned AI “gigafactories” are slated to come online between 2026 and 2027; in 2024, Elon Musk’s xAI built a cluster of comparable size in 122 days. Europe’s most ambitious AI infrastructure projects are two to three years behind the curve, an eternity in AI development. Canada has committed roughly $2 billion to “sovereign AI compute” over five years; Microsoft will spend over $100 billion, more than 50 times as much, on its biggest data center in Wisconsin. Other middle powers face similar constraints: limited budgets, skeptical publics, strained energy grids.
A sovereignty moon shot could take years to fail, absorbing energy and resources while the technology gap widens irreversibly. In the end, these countries would be left with sub-frontier national champions — second-rate systems that cannot truly compete but get deployed anyway when local politics or regulation demands it. These champions have value as fallback options or alternatives for particularly sensitive applications, but they are not a substitute for access to the cutting edge.
So what are middle powers to do? Rather than chasing dreams of AI independence, they should double down on their strengths to get a better deal — ideally from Washington rather than Beijing. For Western middle powers, Chinese AI comes with dependencies more dangerous than anything the U.S. could impose. Beijing is no stranger to economic coercion, and Chinese AI systems carry additional risks around potential back doors and data security that these countries can ill afford.
Middle powers are far from powerless. Upstream, they hold critical minerals, semiconductor manufacturing equipment and data — inputs American labs are already scrambling to secure. Downstream, they offer deployment capacity: the manufacturing, robotics and applied research and development needed to turn AI capabilities into real-world value. Together, these bottleneck assets — the Netherlands’ extreme ultraviolet lithography machines, Germany’s optics systems, Canada’s critical minerals — give middle powers real leverage.
But leverage only works if you’re willing to use it. The U.S. has shown it will weaponize supply chains when it serves American interests; middle powers should make clear they have learned the lesson. Access to critical inputs should be contingent, not automatic, offered in exchange for durable guarantees on access to the newest AI models, restrictions on data extraction and commitments to build significant compute capacity on allied soil.
Elements of the Trump administration recognize this logic: The State Department’s Pax Silica initiative is an early attempt to draw countries in the semiconductor supply chain into AI-for-resources agreements. Middle powers should take the opening, but use it to extract real commitments.
As things stand, middle powers lack the technological base to sustain an independent course in the age of advanced AI. They can use this political moment in one of two ways: squander it on sovereignty theater, or convert their leverage into durable access to frontier systems on favorable terms. The second path requires accepting an uncomfortable truth: They have already lost the AI race. The only question now is what they can extract from the winners.
The post Middle powers need to get smart about AI competition appeared first on Washington Post.
