- Jensen Huang said in Paris that Nvidia will announce $2 billion of European investments in the next week, adding to $8 billion put into the continent over the past year, on a tour also taking him to London and Berlin. He announced new funding for French startups Doctolib, Genesis AI, Pigment and Entalpic.
- One of those investments carries a revealing price. A secondary share sale earlier this year, in which employees and early investors sold about €300 million of stock, valued Doctolib at €3.6 billion, a 38% discount to the €5.8 billion it was worth after a 2022 equity raise.
- Huang pitched open-source models as the route for Europe to catch up, arguing they can be customised for languages and for sovereignty. Nvidia is a key partner to France’s Mistral AI, which designs open-weight models and trained its latest system on 4,000 Grace Blackwell chips.
- He arrives as the technology faces growing backlash over models going rogue and the environmental impact of power-hungry data centres, with Europe and the UK seeking a role in safeguarding against potential harms. Huang has said the AI industry does not need new laws and pushes back on calls to slow development.
What Happened?
Huang said Europe needs to get in the race, and has pledged hundreds of billions of dollars of investment worldwide to accelerate adoption of the technology and his company’s products. He wants to broaden AI use across the global economy and reduce dependence on the products of a small group of very large companies. Doctolib, founded in 2013, handles administrative work for doctors and has become a staple for European patients booking appointments. Pigment develops AI-powered business software, Entalpic applies AI to chemistry and materials research, and Genesis focuses on AI for robots, a market Huang has predicted will be the next large opportunity through humanoid robots. Separately, Anthropic chief executive Dario Amodei and OpenAI’s Sam Altman are considering how to pace development following several high-profile breaches caused by agents.
Why It Matters?
The Doctolib valuation is the most useful number in the article and it sits in the middle of it. A well-established European technology company changed hands this year at 38% below its 2022 mark, which is a concrete reading of where application-layer valuations in Europe actually sit. Set that against AI infrastructure, where Firmus asked $30 billion on $51 million of revenue and had its IPO refused twice in 48 hours. The companies building things with AI have already repriced; the companies building capacity for it largely have not. That gap is the clearest expression of where the enthusiasm is concentrated. The convergence on open models deserves scepticism from both directions. Alibaba chairman Joe Tsai told Europe last week that open source is its best route to independence from the United States and China, and Huang now says open models are how Europe catches up and secures sovereignty. Alibaba is among China’s largest publishers of open-weight models and Nvidia sells the chips every open model trains and runs on. Open models do not reduce accelerator demand, they increase it, because more organisations run their own infrastructure rather than calling an API. The sovereignty argument is genuine and it is also a sales argument. The divergence on safety is the structural point worth carrying. Huang says no new laws are needed and resists slowing down, while the chief executives of the leading model developers are publicly weighing how to pace development after agent breaches. Their incentives differ: a chip supplier sells regardless of which model prevails and bears little liability for model behaviour, whereas the labs carry both. When the hardware layer and the model layer disagree about safety, follow the liability. Note also that investing in startups that buy your accelerators is vendor financing in substance, which is legitimate and also circular, and $10 billion into Europe builds an ecosystem of purchasers.
What Next?
The $2 billion of announcements expected within the week will show which companies and which countries, and the London and Berlin stops are where further commitments are likely. Watch whether European governments commit public capital alongside, since sovereign compute capacity requires more than venture funding and the fiscal room is constrained with French spreads at euro-crisis levels. Mistral’s trajectory is the clearest test of whether European open models can compete, having trained on 4,000 Grace Blackwell chips. On regulation, the gap between Huang’s position and the labs’ is the thing to track, because European and UK rulemaking will be shaped by whichever argument prevails.
Affected Tickers and Coins: NVDA, ASML, NBIS
Source: Bloomberg
