On June 11, 2025, Nvidia moved its GTC stage to the Dôme de Paris, holding GTC Paris alongside VivaTech, Europe’s largest tech event, with Jensen Huang as the headliner. The star of the keynote was not a consumer product but a slogan: filling Europe with AI factories.
Put back in context, the event worked as a strategic pivot. US chip export controls on China had cost Nvidia a major revenue pool, so Europe became the market it had to win. The whole day revolved around one term: sovereign AI. Here is what was announced, and what it means for developers and enterprises.
The Tenfold Commitment
Huang told the audience that Europe’s AI computing capacity will grow tenfold over the next two years. He told CNBC: “Europe has now awakened to the importance of these AI factories, the importance of this AI infrastructure.” He offered two anchors for scale: inference users grew a hundredfold in a couple of years, from 8 million to 800 million, and Nvidia’s partners can now produce 1,000 GB200 systems per week — “and this is just the beginning,” he said.
The official press release put a number on the wave: the sovereign AI deployments announced across Europe will bring more than 3,000 exaflops of Blackwell compute in total. Huang framed AI infrastructure the way earlier generations framed power and the internet: “Every industrial revolution begins with infrastructure. AI is the essential infrastructure of our time.” Marketing language, yes — but paired with country-specific commitments, it reads as an expansion plan with budgets and timelines rather than a vague vision.
Country Deployments and Partners
France was the headline: Mistral AI will build its own AI cloud platform, starting with 18,000 Grace Blackwell systems and expanding to multiple sites in 2026. Germany received what Nvidia calls the world’s first industrial AI cloud, built on DGX B200 and RTX PRO servers with 10,000 Blackwell GPUs, aimed at European manufacturers working on digital twins and robotics. In the UK, Nebius and Nscale will deploy 14,000 Blackwell GPUs in the first phase. Italy’s Domyn (formerly iGenius) is advancing its Colosseum supercomputer, while CINECA and the Barcelona Supercomputing Center are setting up AI technology centers in Italy and Spain.
Telecom carriers form the second leg of the sovereign AI push, with Fastweb in Italy, Orange in France, Swisscom in Switzerland, Telefónica in Spain, and Telenor in the Nordics named as partners working on carrier-grade AI applications and local-language models. The official release also carried quotes from President Macron, UK tech secretary Peter Kyle, and Italian minister Adolfo Urso — national endorsements are part of the sovereign AI pitch.
Sovereign AI and the Developer On-Ramp
Huang defined sovereign AI plainly: compute and models should live on European soil and answer in the local language, culture, and sensibility. Nvidia’s matching products were the Nemotron open models for localized fine-tuning and DGX Cloud Lepton, which connects Europe’s developers to Nvidia’s global GPU cloud provider network. Lepton also integrates with Hugging Face, so models on HF can be deployed directly through Nvidia’s NIM software — “one model architecture, one deployment, and you can run it anywhere.” CNBC noted that Perplexity will carry Nemotron-based multilingual models for European users. For European teams wary of depending on a single US cloud, this is the entry point to a multi-provider compute mix.
The Product Messages in the Keynote
Beyond the country announcements, the keynote restated Nvidia’s current product lines. The GB200 NVL72 is in full production, described as “one giant GPU” — a rack positioned as a thinking machine for reasoning workloads. DGX Spark and RTX PRO servers extend the same architecture down to workstation scale. Robotics and physical AI formed another thread: “Soon, everything that moves will be robotic. And the car is the next one,” Huang said, with the NVIDIA DRIVE autonomous-vehicle platform announced as in production. He also called quantum computing an inflection point, echoing the news that Denmark’s Gefion supercomputer now runs CUDA-Q.
What It Means for Developers and Enterprises
Translated into engineering terms, three things are worth keeping. First, visibility into supply improved: from national supercomputers to carrier clouds to DGX Cloud, Europe now has a full spectrum of compute, and quantum-hybrid workloads have a landing spot. Second, the direction is confirmed: from mid-2025, European governments treat compute as a policy matter, and Nvidia has already claimed the supplier seat. Third, keep the right to be skeptical — “tenfold in two years” is Huang’s target statement, not an achieved fact. Watching whether Mistral’s AI cloud, Germany’s industrial cloud, and the UK deployments ship on schedule will tell you more than the slogan does.
Sources
- Europe Builds AI Infrastructure With NVIDIA — NVIDIA Newsroom
- Nvidia makes big play for Europe with infrastructure deals — CNBC
- NVIDIA CEO Drops the Blueprint for Europe’s AI Boom — NVIDIA Blog
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.