Core Concepts

UCloud AI is a sovereign AI inference platform integrated into UCloud, giving research teams access to open-weight frontier large language models running on Danish infrastructure.

The AI platform is designed around important core pillars, including:

• Instant accessibility: AI inference is integrated directly into UCloud. Users can access the browser-based chat interface with their university login, without installing software or creating a separate account. They can also connect third-party applications through the OpenAI-compatible API using an inference API key. AI usage requires credits, requested through an application form in UCloud.

• Scalable inference: The inference server uses NVIDIA Dynamo, a framework for scalable AI inference, and runs on a dedicated Kubernetes cluster within the Bitten AI supercomputer in Sønderborg.

• Security and sovereignty: Security is a core pillar of UCloud AI. Inference requests are processed on national infrastructure in Denmark under Danish law. Your data is never used to train models, and research files can be used from within your UCloud project without leaving it to reach a model.

  UCloud is ISO/IEC 27001:2022 certified and approved for sensitive data.