The European Commission president’s sixth annual State of the European Union (SOTEU) speech in Strasbourg offered no shortage of geopolitical ambition. From “rebalancing” trade with China to opening a door to “associate membership” for Canada, it showed a Europe that wants to protect its sovereign democracy in an increasingly hostile and fragmented world. Yet perhaps her most consequential statement concerned AI—and a worrying complacency on the need to innovate at home.
“We do not need to develop the frontier technology to be the ones who draw the greatest value from it,” Ursula von der Leyen, the EU’s President, told members of the European Parliament.
There is sound economics behind this claim. Countries adopting new technologies can capture substantial benefits without bearing the full cost and risk of inventing them.
But AI is not an ordinary imported technology. It is becoming economic infrastructure—embedded in banks, logistics networks, research laboratories, defense systems, and public administrations. Whoever controls the models, computers, and chips can influence their price, availability, performance— and their conditions of use.
The EU president acknowledged this problem. She called for European capabilities, greater computing capacity, and more financing for domestic AI companies. Yet her distinction between developing AI and extracting value from it creates the wrong strategic hierarchy. Adoption without meaningful control could turn technological dependence into a constraint on the union’s freedom to act.
Europe has seen this dynamic before. Russia weaponized gas supplies, helping trigger one of the bloc’s worst energy crises in recent history. US tariff pressure was followed by EU commitments to purchase more US energy and invest in the US. China, meanwhile, seeks access to advanced European lithography technology as Brussels presses for fairer trade and to secure critical mineral supplies. Whether the dependency concerns goods, energy, or technology, control over access creates geopolitical leverage.
This is where adoption without continuous innovation becomes risky. If European companies build their operations around foreign models, the threat of losing access or facing delays in receiving upgrades could create a comparative disadvantage, weakening multiple business sectors simultaneously. The EU does not need to develop every frontier model, but it needs credible domestic alternatives. Innovation is therefore not a substitute for adoption. It is what prevents adoption from becoming a painful dependence.
The imbalance is already severe. European Central Bank Chief Christine Lagarde has noted that the US hosts some 75% of global AI computing capacity, compared with only 5% in the EU. According to Stanford’s 2026 AI Index, US-based institutions produced 59 notable AI models in 2025 and China produced 35, while Europe delivered only a handful.
The EU therefore needs more than adoption and diffusion. Raw numbers don’t tell the whole story. Europe may not need to win. But as President von der Leyen said, “We need to stay in the race.” EU leaders must create the conditions for European AI companies to progress from adopters and application developers into frontier competitors. Here, she is right to emphasize greater investment, easier access to risk capital, more computing, and stronger skills. Faster permitting, affordable energy, integrated capital markets, strategic procurement, and sustained upskilling remain essential.
We do not need to control every layer of the AI stack. But we need sufficient domestic capacity to preserve choice, support innovation, and withstand external pressure. Reducing one dependency (on Russian energy, say, or Chinese critical materials) while allowing another to deepen will not deliver “strategic autonomy.” It merely replaces one vulnerability with another, leaving the EU watching AI drive forward, from the back seat.
myTCB® Members get exclusive access to webcasts, publications, data and analysis, plus discounts to events.