Tech • AI • Robotics • Game
NVIDIA GTC Berlin 2026 | Live Keynote Pregame
8/10TL;DR
Europe’s AI push is increasingly centered on open models, sovereign infrastructure, and industrial deployment, with companies including Mistral AI, Dassault Systèmes, Deutsche Telekom, and Mercedes-Benz outlining large-scale plans from data centers to factory automation and autonomous driving.
KEY POINTS
Europe pushes for a full AI stack
European industry leaders framed the region’s next challenge as building the entire AI stack rather than relying only on research talent or manufacturing strength. The central argument was that Europe already has the industrial base, engineering depth, and scientific expertise, but now needs local compute, software, models, and deployment capabilities to turn that advantage into growth.
Open models are becoming a strategic priority
Mistral AI chief executive Arthur Mensch said open models are no longer merely catching up with closed systems, arguing that some now lead in areas such as economic engineering, cyber capabilities, and geospatial imaging. He described open models as critical for broad participation because they lower costs, allow customization, and give companies and governments more control over continuity and security.
Sovereign AI is defined as control, not isolation
Executives used sovereign AI to mean the ability to choose infrastructure, keep sensitive data close to deployment, and avoid dependence on a single external provider. Mensch said sovereignty applies not only to countries but also to companies and even individuals who want AI systems aligned with their own data, costs, and operational requirements.
Telecom-backed AI infrastructure is scaling in Germany
Deutsche Telekom chief executive Tim Höttges said the company’s first 10,000 GPUs in Munich sold out within six months, prompting plans for a larger gigafactory. He said the group is building sovereign AI infrastructure for enterprises that want classified or local services in Germany and Europe, and added that 400 megawatts of data-center capacity are already in place.
Mistral outlines a multi-year compute roadmap
Mensch said Mistral has invested heavily in its own infrastructure for both training and inference and is now operating about 50 megawatts of capacity. He added that the company plans significant expansion in 2027 and 2028, with a gigawatt roadmap for 2029 that is currently on schedule.
Industrial AI is moving from concept to operations
Dassault Systèmes chief executive Pascal Daloz said industrial AI must be grounded in physics, simulation, and manufacturing know-how, not just text generation. Dassault and NVIDIA are combining virtual twins, world models, and AI agents to improve design, testing, and operations, including the layout and commissioning of AI factories.
The biggest ROI comes from reducing downtime
In industrial settings, executives said the clearest returns come from predictive maintenance and process optimization. Mensch cited deployments where AI agents can forecast equipment failures and intervention windows, generating very large savings in sectors where service-level penalties and downtime can cost hundreds of millions.
AI is reshaping internal operations, not just products
Höttges said Deutsche Telekom is using AI across customer service, software development, marketing, and network management. He said the company is already seeing productivity gains of up to 60% in some areas and about 30% in service functions, while 26% of its 200,000 employees are now classified as AI-skilled.
Companies are adopting many models, not one
The discussion pointed to a future of multiple specialized models working together rather than a single universal system. Deutsche Telekom said it currently uses 38 models, while Mistral said roughly 30 internal AI-powered applications handle large token volumes, often combining large reasoning models with smaller fine-tuned systems for narrow tasks.
Autonomy and robotics are nearing commercial rollout
Mercedes-Benz said Level 2++ automated driving is planned for the German market before Christmas, with wider European expansion expected later. Executives also said humanoid robots are likely to enter homes in three to five years, while factory logistics deployments should arrive sooner, with broader manufacturing use beginning within about 24 months.
Healthcare leaders see AI extending survival and easing workloads
In medical applications, AI was described as improving early detection, diagnosis, treatment planning, and radiotherapy delivery. Industry executives argued that these tools can turn some deadly diseases into more manageable chronic conditions, while also giving clinicians time back by streamlining planning and operational tasks.
CONCLUSION
Europe’s AI race is shifting from ambition to execution, with infrastructure, data control, and industry-specific deployment emerging as the decisive battlegrounds. The outcome will depend less on headline models alone than on whether European companies can scale compute, integrate AI into operations, and retain control of the systems that matter most.
Ask a question