TL;DR — Key Takeaways
- NVIDIA is expanding beyond GPUs: Its European strategy spans AI infrastructure, agent security, sovereign AI, robotics, automotive and telecom.
- Faster deployment and lower power consumption are priorities: NVIDIA claims DSX POD can shorten onsite installation to 14 days, while its energy software can reduce AI infrastructure power demand.
- Deployment timelines and ecosystem dependence remain considerations: Several systems are expected in 2027, and European customers must weigh greater control over AI deployment against reliance on NVIDIA’s technology.
NVIDIA CEO Jensen Huang outlined an ambitious expansion of the chipmaker’s European AI strategy at GTC Berlin, unveiling new infrastructure designs, agent security technology and AI partnerships. Overall, the announcements highlighted NVIDIA’s push beyond GPUs into the systems companies need to deploy AI at scale.
Rather than introduce a new chip, NVIDIA focused on reducing deployment costs, improving energy efficiency and supporting AI software in sectors ranging from automotive to telecom. The company has invested nearly $10 billion in Europe over the past year and now works with more than 12,000 European partners. It also announced 14 startup investments during Huang’s recent European tour.
Big picture: while NVIDIA is building an interconnected portfolio, some of the technology remains scheduled for future deployment, and NVIDIA’s performance claims will need to be validated in real world use cases. For European customers, the open question is whether they can expand AI adoption and maintain data sovereignty while avoiding dependence on a single tech giant.
AI Factories Aim for Faster Deployment
A major announcement was the Vera Rubin DSX POD, a modular AI hardware design that enables vendors to assemble and test computing systems before delivering them to customer sites. NVIDIA claims the approach can reduce onsite installation from about six months to just 14 days, provided the necessary infrastructure is up and running.
Seventeen tech partners, including Dell, HPE, Lenovo and Schneider Electric, are expected to offer DSX POD systems beginning in early 2027.
NVIDIA also introduced technology to support the vast energy requirements of AI data centers. The company claims its DSX MaxLPS software can deliver the same AI token output using up to 30% less power. In a trial involving the UK National Grid, NVIDIA’s DSX Flex technology reduced cluster power demand by up to 40% during simulated grid events.
Agent Security and Sovereign AI
NVIDIA also introduced OpenShell, a security architecture created to control how autonomous AI agents access enterprise applications and data. When deployed with NVIDIA’s BlueField-4 processors, the architecture can enforce security policies outside the AI agent itself, helping prevent unauthorized access. NVIDIA claims the architecture can increase agent capacity by up to 20%. More than 100 partners are participating, including major AI developers and cybersecurity vendors.
Another priority was sovereign AI, which enables organizations to operate AI models within infrastructure they control. NVIDIA expanded its open model offerings, including the Nemotron family, allowing enterprises to customize models using proprietary data while retaining control over deployment.
The approach addresses European concerns about data governance, although organizations using NVIDIA’s infrastructure remain dependent on its hardware and software ecosystem.
Robotics, Automotive, Telecom
Industrial AI was another major focus, particularly software that connects AI models with physical equipment. NVIDIA announced new robotics development tools and highlighted Agile ONE SN, a humanoid research platform developed with Germany’s Agile Robots. The system is expected in early 2027.
In automotive technology, Mercedes-Benz, Volkswagen, Volvo and Jaguar Land Rover announced plans for next-gen vehicles compatible with NVIDIA DRIVE Hyperion, a computing and sensor platform that supports the development of autonomous driving systems. NVIDIA also introduced the Halos SAFE framework to help automakers document and validate self-driving vehicle safety.
Telecom is another expansion opportunity. NVIDIA estimates that operators could support roughly 100 gigawatts of distributed AI computing capacity across 100,000 sites. European carriers are testing NVIDIA-powered radio access network technology, with broader commercial deployments planned for 2027.