TL;DR

• Infercom, a Luxembourg-based company, is establishing sovereign European AI inference hosted in Germany, built with SambaNova.
• The platform is hosted in an Equinix data center in Germany, and data does not cross national borders.
• It is designed to adhere to European standards, GDPR, and the EU AI Act.
• Infercom does not store any data for AI inference, which removes a category of risk for sensitive public sector and enterprise workloads.
• It runs on SambaRack, powered by SambaNova's Reconfigurable Dataflow Unit (RDU), at an average of 10 kW per rack, serving open-source models including OpenAI's gpt-oss and DeepSeek.

Background

AI sovereignty became a procurement requirement across Europe. Enterprises and public sector bodies now need to know where their inference runs, which law governs it, and who could compel access to it. For regulated workloads, running on foreign cloud infrastructure means surrendering data control and compliance posture in a single decision.

Infercom set out to give European organizations a compliant alternative: A sovereign inference platform hosted in Germany, designed for European standards, that keeps inference data inside the bloc and stores none of it.

Challenge: Compliant Inference that Stores Nothing

Infercom's platform had to satisfy several requirements at once.

• Data residency that survives audit. Inference has to happen inside national and bloc borders, with no cross-border processing.
• Regulatory alignment by design. The platform had to be built for European standards, GDPR, and the EU AI Act, not retrofitted to them.
• No retained inference data. Sensitive workloads needed the assurance that inference requests leave nothing behind.

Solution: SambaRack, Hosted in an Equinix Data Center

Infercom built on SambaRack, powered by SambaNova's RDU, hosted in an Equinix data center in Germany.

Data that Does Not Cross the Border

The platform is hosted in Germany and designed to adhere to European standards, GDPR, and the EU AI Act. Data does not cross national borders, so compliance posture is built into where and how the platform runs.

Zero Inference Data Stored

Infercom does not store any data for AI inference. Rather than documenting a data-handling policy, the platform removes the category of risk entirely, which matters most for sensitive public sector and enterprise workloads.

10 kW Per Rack

SambaRack runs large models at an average of 10 kW, so Infercom can add sovereign inference capacity efficiently. The platform serves the best open-source models, including OpenAI's gpt-oss and DeepSeek.

Why Infercom Chose SambaNova

1. Jurisdictional control. Inference is hosted in Germany and designed for GDPR and the EU AI Act, with no cross-border data.
2. Zero retained data. Infercom stores no data from inference requests at all.
3. Energy efficiency. An average of 10 kW per rack makes sovereign capacity viable inside existing facilities.
4. Open-source breadth. Current open-weight models, including gpt-oss and DeepSeek.

What Comes Next

Infercom is bringing sovereign European inference online, hosted in Germany and compliant by design. The pattern is set: inference inside the bloc, governed by European law, storing nothing. Read more on sovereign AI and national autonomy, or talk to our team about a sovereign deployment in your region.