As AI adoption grows across Europe, one question is becoming impossible to ignore: Where does the data go, who controls it, and how can companies stay compliant while still moving fast? In this video, SambaNova’s Vasanth joins Altug, managing director at Infercom, for a timely discussion on why sovereign AI matters so much in the European market. Their conversation gives viewers a clear look at the business, policy, and infrastructure forces shaping enterprise AI in the region, while also showing why this topic now matters far beyond Europe alone.
TL;DR
• Sovereignty means control, not just location. Knowing what data you are using, where it goes when you press enter, who can access it and at what level. GDPR and the EU AI Act turn that into both a compliance question and a trust question.
• Infercom is built European-first at every layer. The company was established in Luxembourg and runs SambaNova-powered inference from a data center in Munich, with hardware, software, platform and even business systems kept in Europe.
• Power per rack decides where you can actually deploy. Current GPU racks can draw 50 kW to 80 kW and need liquid cooling, which few European data centers support. The SambaRack systems Infercom uses peak around 20 kW on air cooling, so they fit existing facilities.
• It is positioned as an alternative, not a replacement. GPUs remain useful for training; the argument here is about real-time inference serving, where Infercom offers a European option alongside the hyperscalers.
What data sovereignty actually means
A major focus of the session is data sovereignty and why it has become a core requirement for European organizations. Altug explains that sovereign AI is ultimately about control: Knowing what data is being used, where it is stored, who can access it, and whether its use aligns with regulations such as GDPR and the EU AI Act. The discussion makes a strong case that this is not just a legal issue, but also a trust issue. For enterprises handling financial, health, or other sensitive data, the ability to keep AI infrastructure and data residency within Europe is becoming a critical part of responsible deployment.
How Infercom and SambaNova built a European-first stack
The video also highlights how Infercom and SambaNova are working together to deliver a European-first AI infrastructure model. Infercom has built its service in Luxembourg and deployed SambaNova-powered inference infrastructure in Germany, with the goal of offering a sovereign alternative to hyperscalers for AI inference workloads. The speakers explain that this approach is designed from the ground up to support compliance with every layer of the service, from infrastructure to software systems, centered on European hosting and governance. For businesses looking for practical paths to adopt AI without giving up control over their data, this is one of the most valuable parts of the conversation.
Why energy efficiency decides where AI can be deployed
Another standout theme is energy efficiency, which the webinar frames as one of the biggest barriers to scaling AI in Europe. Rather than treating compute as a purely technical issue, the speakers connect it directly to the realities of data center availability, power limits, cooling requirements, and long-term cost. They explain why efficient inference infrastructure can open up more deployment options across Europe, especially in environments where traditional GPU-based systems face power and cooling constraints. This makes the discussion especially useful for leaders thinking not just about AI capability, but about how to operationalize it at enterprise scale.
Where the market goes next
The session closes with a practical view of where the market is heading over the next few years. Instead of arguing that companies should adopt AI for its own sake, the speakers stress that enterprises should begin with real business problems, identify strong use cases, and then choose infrastructure that supports performance, compliance, and regional requirements. For anyone exploring enterprise AI in Europe, this video offers a thoughtful introduction to the future of sovereign AI, the value of local infrastructure, and the role partnerships like Infercom and SambaNova can play in building a more secure and sustainable AI ecosystem.
FAQs
Data sovereignty is control over your data: knowing what data you are putting into an AI system, where it goes once you submit it, who can access it and at what level of access. Without that visibility, organizations risk penalties under regulations such as GDPR and the EU AI Act, and risk losing customer trust. Accessibility matters too, because data held outside the EU may become unreachable if circumstances change.
The EU AI Act works as a guideline built around the risk level of your data. Health and medical information, financial account details and similar categories count as high risk. Companies are expected to understand what data they hold and its risk level, then put appropriate controls and measures in place to govern how that data is processed within AI systems. Without that understanding, an organization is exposed regardless of intent.
Altug frames it as an alternative rather than a competitor. GPUs remain useful for training and many other scenarios. The argument concerns real-time serving, where inference is the critical workload. Companies already invested in hyperscalers are encouraged to examine what level of sovereignty those arrangements actually provide, and to benchmark real-time serving performance against European options rather than switching wholesale.
It determines which data centers you can use at all. Traditional cloud racks draw roughly 4 kW to 6 kW. Early GPU deployments pushed that to 20 kW to 25 kW, and current generations reach 50 kW to 80 kW while also requiring liquid cooling, which relatively few European facilities support. The SambaRack systems Infercom deploys peak at around 20 kW and run on air cooling, which means capacity can be added in existing data centers across multiple European locations rather than a handful of the largest sites.