The Danish Centre for AI Innovation (DCAI), operator of Gefion, one of Europe’s most powerful AI supercomputers, has partnered with WEKA, the AI storage company, to launch a fully sovereign, end-to-end encrypted AI platform. The deployment unifies high-performance compute and storage into a single infrastructure, providing European organizations with hyperscaler-grade capabilities without compromising data sovereignty.
With an additional 140 petabytes of usable storage, including 80 PB of performance-optimized NVMe SSDs running on WEKA’s NeuralMesh™ software, Gefion now functions as a complete “AI factory,” enabling data ingestion, model training, and deployment within one trusted platform.
Unified Compute and Storage for AI Workflows
Until now, DCAI users primarily accessed compute power for model training. The new storage service lets customers manage entire AI workflows—from raw data to deployed models—without relying on external cloud services.
“Our mission is to give innovators the infrastructure they need to build the future,” said Nadia Carlsten, CEO of DCAI. “By adding storage to our platform, we’re enabling end-to-end AI workloads with the performance, sovereignty, and security our users expect.”
WEKA’s NeuralMesh architecture delivers sub-millisecond latency, eliminating traditional I/O bottlenecks that limit GPU utilization. The combination of Gefion’s NVIDIA GPU clusters and NeuralMesh creates a zero-copy, high-throughput data path, maximizing GPU efficiency while maintaining secure multi-tenant operations.
Efficiency, Sustainability, and Cost Benefits
The platform’s single-tier, software-defined architecture reduces infrastructure footprint, energy consumption, and operational complexity compared to traditional multi-tier systems. By removing the need for data replication or cloud egress, customers can lower total cost of ownership while advancing green AI initiatives.
Future integration with WEKA’s Augmented Memory Grid™ aims to accelerate time-to-first-token performance for large language models, enabling:
- Instant burst scaling between compute and storage resources.
- Consistent, high-performance multi-tenant workloads.
- Scaling from petabytes to exabytes without performance degradation.
Purpose-Built for Regulated Industries
The platform addresses critical European data sovereignty requirements, featuring:
- End-to-end encryption with zero operator access.
- Full compliance with EU data protection mandates.
- Independent privacy audits ensuring regulatory adherence.
“Europe’s AI innovators need infrastructure that can scale from petascale to exascale while maintaining sovereignty,” said Liran Zvibel, CEO and co-founder of WEKA. “Our collaboration provides the foundation to build AI factories capable of exabyte-scale growth with world-class performance and security.”
Carlsten added, “European organizations no longer have to choose between hyperscaler performance and data control. This platform eliminates the tradeoff, integrating NVIDIA supercomputing and WEKA storage into a single, sovereign AI solution.”
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
