Scaleway needed a secure, scalable AI infrastructure that would let European customers train large models inside the EU while meeting data-sovereignty, privacy, and efficiency requirements.
Reported impact
Scaleway stood up its Nabu supercomputer in two months, gave European AI companies a sovereign alternative for large-model training, and paired that scale with GDPR-aligned hosting and lower-power, lower-water data-center operations.
Scaleway assembled its sovereign AI cloud around the Nabu supercomputer so European customers could train frontier models inside the EU without handing data control to non-European hyperscalers. The environment is built from 127 NVIDIA DGX H100 systems and 1,016 NVIDIA H100 Tensor Core GPUs, with NVIDIA NVLink handling intra-node bandwidth and Quantum-2 InfiniBand providing low-latency cluster communication. Hosted in Scaleway's DC5 data center in Paris, the platform gives customers a managed environment for training LLMs and other transformer workloads while staying aligned with GDPR and local data-sovereignty requirements. Scaleway used the same stack to support projects such as Mistral AI's Mixtral and Kyutai's Moshi, while its smaller Jero system extends the model to additional workloads. Energy-efficient data-center cooling reduces electricity and water usage, so the architecture pairs scale-out AI training with compliance and sustainability goals.
Customer model-training jobs are submitted into Scaleway's DGX-based cloud, GPUs and NVLink handle node-level training, Quantum-2 InfiniBand connects the cluster for distributed runs, and completed models remain inside EU-hosted infrastructure for downstream customer use.
DC5 uses advanced adiabatic cooling systems, reducing electricity consumption by 30-40% and using 9-10X less water than traditional data centers.
Its AI infrastructure complies with stringent data protection laws, like the EU's General Data Protection Regulation (GDPR).
Powered by 1,016 NVIDIA H100 Tensor Core GPUs and 127 NVIDIA DGX H100 systems, Nabu offers European businesses access to a turnkey solution capable of supporting the most demanding machine learning tasks.
The company built the Nabuchodonosor 2023 AI supercomputer-known as Nabu-in just two months.
Review how this reference was scored and trace each normalized claim back to its documented source.
Scoring methodology
Customer Reference Value measures usefulness as a customer reference. Evidence Confidence measures the quality and traceability of available evidence, not independent verification.
Traceability
Field coverage shows which facts each source supports. Dates are shown only when documented by the source.
LeadIQ says Scaleway has approximately 675 employees as of August 2026, and the Organization JSON-LD lists numberOfEmployees value 675.
Related
NVIDIA Sample Sep 22, 2026 Sluicebox Sluicebox needed to turn fragmented bills of materials, supplier records, PDFs, CAD files, and emissions inventories into timely product carbon assessments, but manual life-cycle-analysis work was slow, expensive, and… 51-80% lower inference costs Lucy · NVIDIA Nemotron 3 Ultra · carbon graph NVIDIA Sample Sep 15, 2026 Lambda Lambda needed to raise token throughput and improve AI-cloud economics, but fixed rack and facility power budgets were leaving GPU capacity stranded and limiting how many nodes could run safely at once. +17% Hybrid Training Throughput Dynamic Power Software · NVIDIA DSX AI Factory Platform · NVIDIA DSX MaxLPS NVIDIA Sample Sep 1, 2026 Onetera Onetera needed to help more than 90 municipalities turn fragmented regulations, zoning maps, permit rules, and legacy documents into fast resident-facing services without imposing heavy IT overhead or expensive manual… 1.7x increase in data throughput GPT OSS 20B · Government Intelligence Agent · LangGraph 