At Rise of AI 2026 in Berlin, a panel on the Meta Stage discusses what European AI sovereignty means in practice. Moderator Mario Brandenburg opens by noting that Ronja Kemmer, originally announced from the Bundestag, called in sick and has been replaced by a representative from Nvidia's European public sector business. The other panelist is Tina Kluewer of AI Nation, a German government initiative bringing together regional AI startup support programs. The Nvidia representative frames sovereignty through a five-layer model (energy, chips, infrastructure, models, applications) and Nvidia's board-approved definition that sovereign AI is owned, operated and governed in-country. The panel debates how to measure sovereignty, its link to military security and quantum computing, why German AI startups struggle to scale despite a strong Mittelstand, whether GDPR and the coming AI Act help or hurt European competitiveness, and what Europe's context-rich manufacturing and healthcare data could contribute as a genuine advantage. The discussion closes with predictions about the panel topic in ten years.
Key takeaways
- Nvidia's board-approved definition frames sovereign AI as AI owned, operated and governed in-country across five layers: energy, chips, infrastructure, models and applications.
- Tina Kluewer argues German AI startups struggle to scale mainly due to weak late-stage funding, reluctant corporate buyers and few domestic acquisitions.
- GDPR compliance nearly killed Kluewer's own earlier startup, tying up half of her ten-person team.
- Kluewer expects the AI Act, if implemented as currently planned, to prevent any major new AI company from emerging out of Europe.
- Panelists point to Europe's context-rich manufacturing and healthcare data, not chips or large language models, as its strongest AI advantage.
Detailed notes
A late substitution and defining sovereignty
Moderator Mario Brandenburg opens by explaining that Ronja Kemmer, announced from the Bundestag, called in sick and has been replaced on the panel by a representative from Nvidia's European public sector business (her name is not clearly audible in the recording). The other panelist is Tina Kluewer, who describes AI Nation as a German government initiative that is bringing together several regional AI startup support programs, with early hubs in Munich and Berlin, into one national program for AI founders and startups. The Nvidia representative frames sovereignty using what she attributes to Nvidia CEO Jensen Huang as a 'five-layer cake': energy, chips, infrastructure, models and applications. She cites Nvidia's board-approved definition that sovereign AI is AI owned, operated and governed in-country, while stressing that no country is sovereign across all five layers and that sovereignty is really a set of strategic choices about what to build, what to buy and where to accept dependencies.
Measuring sovereignty
Kluewer argues sovereignty should be tracked through economic KPIs: the number of AI companies, AI-related jobs and education opportunities a country has. The Nvidia representative proposes a risk-based calculation instead: estimate the days of work or business that would be lost if a specific dependency, such as a local data-center provider, became unavailable, multiply by the likelihood of that failure, and compare the resulting risk exposure across different choices and layers to decide what is worth accepting or hedging against.
Sovereignty as a security and military question
Asked whether sovereignty can be separated from military and security concerns, Brandenburg frames it by suggesting digital sovereignty is essential for military sovereignty; Kluewer connects this to quantum computing, arguing that not having a stake in the 'global game' of quantum computing would create safety and military risks as the technology matures, though she says AI Nation does not focus specifically on quantum computing, only on some startups working at its interface with AI. The Nvidia representative adds that the point is also practical: adversaries already use AI and drone-based warfare, so lacking the same technological stack leaves a country vulnerable. She also describes Nvidia's own move from one chip two years ago to seven interconnected chips today and predicts a quantum processing unit will eventually become one more specialized chip in that system, without giving a timeline for mainstream adoption.
Why German AI startups struggle to scale
Kluewer says Germany produces many interesting AI startups but few grow into large players, pointing to insufficient late-stage venture funding, established companies' reluctance to buy software from startups, and a lack of mergers and acquisitions by German corporates to bring startup technology in-house. The Nvidia representative counters that scaling is hard everywhere, citing slow internal decision-making at large healthcare and life-science companies she worked with before moving into public sector, and suggests that Germany's many family-owned, privately held mid-sized companies could be an advantage because they can make longer-term technology decisions rather than thinking in quarters. Building on that point, Kluewer adds that an excessive focus on shareholder value can hinder innovation, describing the classic innovator's dilemma in which very successful companies ignore a new technology until a startup or competitor embraces it and gains a better product and market share within a few years, a view the Nvidia representative agrees with.
GDPR, the AI Act and who regulation protects
The Nvidia representative argues regulation can help smaller companies more than large ones because it defines clearly what is and is not allowed, reducing a legal risk that a small company cannot absorb the way a bigger one can; she also says she values Europe's data protection compared with markets where personal data trains models without consent. Kluewer pushes back from her own experience: GDPR nearly killed her startup when it took effect, since five of her ten employees had to work on compliance, and she believes the company would have reached the market faster without it. She says that if the AI Act is implemented as currently planned, she does not expect any major new AI offering or company to emerge from Europe. Both agree that past regulation, including GDPR, has not created the market opportunities for European products once expected, since most digital tools used in Europe are American.
Europe's future strengths and the role of values
Asked what Germany's next competitive 'product' could be, Kluewer suggests robotics, pharma and medicine, and networked, AI-assisted cars and highways. The Nvidia representative argues Europe's real advantage is its context-rich data from manufacturing and healthcare, including long, centralized public health records that systems like the US lack, plus mechanisms such as federated learning that could make that data usable for training while keeping it protected. On whether 'European values' are a genuine advantage or an excuse for weaker products, panelists agree a product cannot succeed on patriotism alone; the Nvidia representative adds that regulation is itself a manifestation of values about what a society decides is worth protecting, and that transparency about training data, not just open weights, could become a real differentiator. Closing the panel, the Nvidia representative predicts that in ten years radiologists will still be doing their jobs even if self-driving cars have arrived, while Brandenburg jokes the next panel topic could be the German autobahn's technological future.
Quotes
Sovereignty is not autonomy, that's a different concept.
Tina Kluewer
I basically built up my own company in the time the GDPR came up and it almost killed us, because we were ten people and five of them were dealing with GDPR.
Tina Kluewer
I would almost say it's more theoretical than actually doing the math, but I think you can calculate it.
Mario Brandenburg
Questions answered in this talk
What does 'sovereign AI' mean according to Nvidia?
According to the Nvidia representative on the Building European AI Sovereignty panel at Rise of AI 2026, Nvidia's board-approved definition is that sovereign AI is AI that is owned, operated and governed in-country, considered across five layers: energy, chips, infrastructure, models and applications.
Why did GDPR nearly kill Tina Kluewer's startup?
According to Tina Kluewer at Rise of AI 2026, GDPR compliance tied up five of her ten employees when it came into force, which she says significantly slowed the company down.
What is AI Nation?
According to Tina Kluewer at Rise of AI 2026, AI Nation is a German government initiative that unites regional AI startup support programs, including early hubs in Munich and Berlin, into one national program to support AI founders and startups.