Directory
14 EU/EEA providers meeting the strict inclusion bar: EU-incorporated, EU-hosted, no foreign control. Sorted by sovereignty score.
German managed-cloud and hosting provider (plusserver GmbH, Cologne), founding member of Gaia-X. Runs pluscloud (OpenStack/VMware), dedicated servers and colocation from its own German data centers, plus "plusAI", a digitally sovereign AI/GPU platform aimed at SMEs.
Public IaaS cloud operated by T-Systems International GmbH, a Deutsche Telekom AG subsidiary, hosted in Deutsche Telekom's own German and Dutch data centers. Rebranded from "Open Telekom Cloud" to "T Cloud Public" in 2026; offers NVIDIA H100 GPU ECS instances ("p5" flavor family) alongside general-purpose IaaS.
Finland-based GPU cloud operator, rebranded from DataCrunch to Verda in November 2025. Runs bare-metal and cloud GPU instances from its own data centers in Finland and Iceland, powered by renewable/hydro energy.
B2B cloud/ICT arm of French telecom group Orange, operating the Cloud Avenue platform (including a SecNumCloud-qualified "Cloud Avenue SecNum" offer) with a GPU-as-a-Service IaaS option for generative-AI workloads hosted in Orange's own French data centers. Also co-owns Bleu, a separate Orange/Capgemini joint venture reselling Microsoft 365/Azure as a "cloud de confiance" — Bleu does not publicly disclose a GPU/AI compute product as of 2026-07-08 and is described here for context only.
Italian cloud and hosting company, family-owned since its 1994 founding near Bergamo, operating its own data centers in Italy plus additional European sites (France, Czech Republic, Poland), including GPU-equipped virtual private cloud and dedicated-server options.
Sovereign cloud brand of Schwarz Digits, part of Germany's Schwarz Group (parent of Lidl and Kaufland). Runs OpenStack-based IaaS/PaaS from its own data centers in Germany and Austria, including a Compute Engine GPU product (H100, A100, L40S), and is one of four winners of the European Commission's sovereign cloud framework contract.
100%-state-owned Luxembourg postal and telecom operator. Its digital services unit DEEP (legal entity Post Telecom S.A.) operates three Tier IV-certified data centers in Luxembourg and, since 2025, a sovereign-cloud partnership with OVHcloud whose published GPU instance line (H100/A100/ H200) is described as part of the underlying platform; DEEP does not publish its own GPU SKU pricing or confirmed regional availability as of 2026-07-08.
EuroHPC petascale supercomputer hosted by LuxProvide in Luxembourg's Bissen data center (run by LuxConnect), operational since 2021. The Accelerator Module uses 200 nodes of 4x NVIDIA A100 40GB GPUs (NVLink 3). Access is allocated through EuroHPC JU calls.
Europe's first exascale supercomputer, hosted by Forschungszentrum Jülich / Jülich Supercomputing Centre (JSC), Germany. The JUPITER Booster module uses ~24,000 NVIDIA GH200 Grace Hopper Superchips. Cornerstone of the JUPITER AI Factory (JAIF); access is allocated 50/50 between EuroHPC JU and Germany's Gauss Centre for Supercomputing.
EuroHPC petascale supercomputer hosted by IT4Innovations National Supercomputing Center in Ostrava, Czechia, operational since 2021. Its GPU partition provides 576 NVIDIA A100 40GB GPUs (4 per node, NVLink). The KarolAIna AI Factory upgrade builds on this system for the Czech national AI Factory.
EuroHPC pre-exascale supercomputer hosted by CINECA at the Bologna Tecnopolo, Italy. Its "Booster" module uses custom NVIDIA A100 64GB GPUs. Access is granted via EuroHPC JU and national competitive calls; the IT4LIA AI Factory (launched September 2025) builds on this infrastructure for AI-focused workloads.
EuroHPC pre-exascale supercomputer hosted by CSC - IT Center for Science in Kajaani, Finland, on behalf of a ten-country LUMI consortium. Its GPU partition (LUMI-G) uses AMD MI250X accelerators; access is granted through EuroHPC JU and national calls, not sold commercially. The LUMI AI Factory extends this capacity to AI startups, SMEs, and researchers.
EuroHPC pre-exascale supercomputer hosted by the Barcelona Supercomputing Center (BSC-CNS), Spain. Its Accelerated Partition uses NVIDIA Hopper- generation GPUs (1,120 nodes x 4 GPUs). A EUR 129M AI-upgrade project adds dedicated LLM-training and inference partitions for the BSC AI Factory.
The first EuroHPC supercomputer to become operational (April 2021), hosted by IZUM in Maribor, Slovenia. Its GPU nodes provide 240 NVIDIA A100 40GB PCIe GPUs. Slovenia has submitted a follow-on tender to expand Vega into a national AI Factory.