BERLIN (Realist English). Major European industrial companies are stepping up pressure on Brussels, warning that tough regulation of artificial intelligence and a lack of “openness” in the rules of the game are forcing them to move multibillion-euro investments outside the European Union.
The most high-profile statement came from Siemens: its CEO Roland Busch directly threatened that the bulk of a planned €1 billion for industrial AI could go to the US and China.
The Essence of the Warning: ‘I Can’t Explain This to Shareholders’
Busch formulated the company’s position bluntly: “I cannot explain to my shareholders why I should invest in an environment where I am held back.”
According to him, the European AI Act and Data Act “miss the mark,” regulating industrial AI in the same way as consumer applications and imposing additional layers of oversight on industries that already comply with sector-specific rules.
“It is completely nonsensical to treat industrial and machine data the same as personal data,” he emphasized.
The easing measures proposed by the European Commission — a delay of rules for “high-risk” AI systems and simplified cybersecurity reporting — Busch considers insufficient. His position was openly supported by German Chancellor Friedrich Merz, who stated that the government would push for “freeing industrial AI from the current, overly restrictive straitjacket” of European rules.
Eric Schmidt’s Position: ‘Europe Will Either Pay or Use Chinese Models’
Former Google CEO Eric Schmidt formulated an even more radical thesis earlier this year. Speaking at the World Economic Forum in Davos, he warned that if Europe does not build its own open AI laboratories and does not solve the problem of high energy prices, it “will quickly become dependent on Chinese models.”
“If Europe is not prepared to spend big money on European models, it will ultimately use Chinese ones. Probably not the best outcome for Europe,” he said. Schmidt also pointed to a fundamental gap: the US relies largely on closed models, while China massively develops open ones.
Regulatory Context: Fines Up to 7% of Global Turnover
Companies’ fears are backed by concrete figures. The AI Act, updated in 2026, provides for fines of up to €35 million or 7% of global annual turnover for violations — amounts that, according to critics, may exceed revenue from the AI product itself.
Since August 2026, the regime for “high-risk” AI systems has been in force, applying extraterritorially to any funds operating in the EU. According to estimates by The Parliament Magazine, additional compliance costs could reduce AI investment by 20%.
The Alternative: Mistral as a Bet on ‘Sovereign Open AI’
Against the backdrop of pressure on Brussels, an alternative model is taking shape in France. Mistral AI raised €3 billion on September 8 at a valuation of over €21 billion — the largest private tech funding round in European history.
The round was led by Samsung Electronics, the Scaleup Europe fund managed by EQT, and existing investor PSG Equity; among new investors are Advent, BlackRock structures, and the Grand Duchy of Luxembourg. The key investor is ASML, underscoring the industrial rather than purely venture nature of the bet.
Mistral’s strategy is built on open models with the ability to deploy in the customer’s own environment — “sovereign AI” as a response to European governments’ and corporations’ demand for data control.
More than 125 global enterprises already use Mistral technologies, including Airbus, ASML, and HSBC. The company plans to increase its own computing capacity by about 100% over five years and build up to 1 GW of European capacity by 2030.
However, this model has a built-in paradox: Mistral’s infrastructure runs entirely on Nvidia chips. Sovereignty at the model level does not eliminate dependence at the hardware level.
Key Disagreement: Sovereignty or Competitiveness?
Within the European establishment, subtle disagreements remain in assessing priorities. The European Commission is preparing a “tech sovereignty package” aimed at strengthening the domestic AI industry and cloud infrastructure.
However, Busch sees this as a “danger of misprioritization”: delays in AI adoption — whether due to security considerations, bureaucracy, or the pursuit of “purely sovereign” infrastructure — will “massively slow economic growth.”
In his assessment, the American economy, thanks to the use of existing AI tools, acts like a “fast-flowing river,” while the European ecosystem resembles “stagnant water.”
The European AI paradox is becoming increasingly obvious: the more Brussels strives for technological sovereignty through regulation, the more actively large companies move investments to where the rules are softer.
Siemens, whose history spans more than 175 years, says directly: €500 million out of €1 billion could go outside the EU. Mistral demonstrates that a European alternative is possible — but it requires capital, time, and chips that Europe does not have in sufficient supply.
The question is not whether “openness” in AI is needed. The question is what Europe is prepared to offer in return: if the regulatory barrier remains high and infrastructure dependent, “sovereignty” risks remaining a declaration. For now, the numbers speak louder: 7.5% of global AI investments come to Europe. And that share could shrink even further.
