Europe has no shortage of reports, but a serious shortage of consequences. In September 2023, European Commission President Ursula von der Leyen asked Mario Draghi to examine competitiveness. His report, published on September 9, 2024, pulled no punches, lambasting a fading growth model weakened by poor productivity, a fragmented internal market, poor access to capital, expensive energy, an incoherent defence strategy and risky dependencies. The Commission priced the fix at an additional €750 billion to €800 billion of investment a year.
The Rhine Group is timely. Draghi and Patrick Collison, co-founder and chief executive of Stripe, have assembled business leaders, economists, policymakers and academics to turn that diagnosis into action. Launched on August 24, 2026 with some 55 members, the group holds its inaugural session on September 20-23. Brussels has a well-developed immune response to difficult reports: Acknowledge them, launch a programme of programmes and wait for interest to drain away. The Rhine Group makes that harder.
Europe creates, then watches others scale. Draghi’s diagnosis is painfully familiar to Europe’s technology founders. Europe produces excellent research and founders, but too often fails to supply the capital, customers and integrated market they need to scale. The Rhine Group’s founding statement notes that only four of the world’s 50 largest technology companies are European.
DeepMind is a case study: Founded in London in 2010, it pioneered modern AI and is one of the world’s leading frontier AI labs, but now belongs to Google, which bought it in 2014. Europe can create another DeepMind, but can it keep it?
Recent rapid developments in AI make this urgent. In 2026 alone, AI has materially advanced at least a dozen longstanding mathematical problems. It has disproved an 87-year-old conjecture in algebraic geometry in three dimensions and above, and made the largest advance in nearly four decades on a central bound tied to the Riemann hypothesis, one of mathematics’ most famous unsolved problems.
AI has also predicted hundreds of thousands of potentially stable materials and designed drugs that reached human trials. An algorithmically personalised melanoma therapy reduced recurrence or death by 49 per cent in five-year trial data from the phase 2b KEYNOTE-942 study.
Software agents now automate both coding and hacking. During OpenAI evaluations in July 2026, agents found ways to communicate, collaborate and escape controls, then penetrated OpenAI’s internal research infrastructure and Hugging Face’s systems, according to the company’s own account. When engineers rebuilt the compromised service, the agents re-established access and continued.
In retrospect, the “AI singularity” will look like a single dramatic event, but as we live through it it will feel more like dominoes falling one by one, as AI advances from assistant to autonomous expert in one field after another. The countries and companies that build and deploy the best systems will set the pace of innovation. Dependence on a handful of foreign suppliers leaves Europe negotiating for crumbs.
Europe’s response to three decades of digital disruption has focused on regulation. Increasingly, the technologies driving that disruption are built elsewhere, then adapted to European rules, delayed, withheld or deployed alongside an army of lawyers. European innovators face the same complexity before they have scale. Rules and trust matter, but regulation is no substitute for capability. It cannot create compute, electricity or frontier laboratories.
Sovereignty is agency, not isolation. It is often mistaken for self-sufficiency. Europe should neither accept inferior technology stamped with a blue flag nor cut itself off from the best intelligence. Sovereignty means credible European alternatives and freedom to choose between them and American or open-weight models. The Rhine Group’s international composition reflects that practical instinct.
Autonomy means using the best model for each job and switching at will. A model that is great at software engineering may lag in legal analysis, and the leaderboard may change next month. Model routing sends each task to the AI best suited for each job. This is both technical best practice and continuity risk management: Critical systems should not stop because one supplier fails, changes its terms or comes under external political pressure.
Model choice is only half of sovereignty. When employees work with AI, they are describing how they do their jobs, encoding their company’s intellectual capital, culture, values and institutional memory. Personal accounts and poorly governed AI products may retain that knowledge, use it for training, record it in logs and telemetry or expose it in a breach. Even where enterprise contracts prevent training, the resulting workflows can become trapped inside vendor platforms, leaving companies renting back their own operating knowledge. Businesses must own their own AI context layer, or risk having little left worth owning.
There are three things Europe must do.
First, create the conditions for scale. A European technology company should be able to expand across the 27 member states as easily as its American rivals expand into 50. Europe needs deeper growth capital, simpler rules and internal customers. We seed startups, but then they leave for America to find capital and customers. European savings must be mobilised to invest in Europe’s future, while our public and private institutions should be rewarded for reforming their procurement processes to increase purchases from European startups and technology companies.
Second, build the physical foundations. AI requires chips, data centres, grids and reliable power. Europe’s energy is now a sovereignty problem. Draghi’s report found European industrial electricity prices two to three times those in the United States. Europe must accelerate grids, generation, compute and associated supply chains while funding frontier laboratories at a scale proportionate to the prize.
Third, combine openness with control. Model neutrality and routing should be standard for critical AI. The Open Data Directive tried to make public data broadly accessible, but now the Data Union Strategy, adopted in November 2025, must ensure that publicly funded European data is used as fuel to drive European AI capability. European companies must own their own applied AI layer: The workflows, evaluations and context their teams create.
Regulation matters because consequential AI must be observable, accountable and kept under control. Trust is necessary before any responsible company deploys AI at scale, but complex and ambiguous burdens deter adoption and protect incumbents.
Europe must act at a European scale. Member states can build energy, skills and national champions, but capital markets, infrastructure dependency and bargaining power require continental scale. Together, we must embrace the best technology, while jointly building the capability and leverage to stay free to choose.
Draghi deserves strong support for describing the problem without flinching and proposing change at the scale required. The Rhine Group can sustain pressure, develop evidence and keep urgency in the debate. Implementation belongs to Europe’s institutions and governments, which commissioned the diagnosis and cannot claim they were not warned. Europe should build capability, use the best tools available and own what it learns.
