Scaleway Announces Plans to Deploy AMD-Powered AI Supercomputer in Sweden
European cloud infrastructure expansion reaches Sweden as Scaleway announces plans to deploy an AI supercomputer in the country powered by AMD Instinct MI355X GPUs, Cloud Computing News reported. The facility marks the French cloud provider’s first supercomputer outside France, designed to handle demanding machine learning tasks for regional enterprises and startups.
-
The Tech TL;DR:
- Scaleway is building its first non-French AI supercomputer in Sweden, utilizing AMD Instinct MI355X accelerators.
- The hardware stack features CDNA 4 architecture, 288 GB of HBM3E memory per GPU, and a peak memory bandwidth of 8 TB/s.
- Partners including Bull, Giga Computing, HPE Networking, VAST Data, and Motul are supplying the integration, hardware, network, data platform, and liquid cooling systems.
Swedish Deployment Uses AMD MI350 Accelerators
The Swedish deployment relies on AMD’s MI350 series accelerators built specifically for large-scale training and inference workloads. Each AMD Instinct MI355X GPU operates on the CDNA 4 architecture, packing 288 GB of HBM3E memory with a peak memory bandwidth of 8 TB/s, while maintaining a typical board power draw of 1,400 watts. To manage execution pipelines, the infrastructure runs the open-source ROCm software stack as a direct alternative to the proprietary CUDA ecosystem.
Hardware integration requires a multi-vendor supply chain. French computing firm Bull handles installation, physical integration, and long-term hardware maintenance, while Giga Computing provides the base server nodes. Network traffic routes through HPE Networking Ethernet fabric, and VAST Data provisions the underlying distributed data platform. Motul supplies the specialized liquid cooling fluid necessary to dissipate the thermal output of the high-density computing clusters.
French 🇫🇷 cloud provider Scaleway is heading north ⛰️🇪🇺It's building its first AI supercomputer outside France in Sweden 🇸🇪, picked for its abundance of clean electricity. A second site in Copenhagen 🇩🇰 will run several thousand of Nvidia's latest chips.How big is that?…
# Example cURL request for verifying ROCm device status on provisioned nodes
curl -s --unix-socket /var/run/rocm-smi.sock http://localhost/devices/metrics | jq '.devices[] | {id: .id, power: .power_usage_watts, memory: .hbm_used_percent}'
Scaleway Partners with Tele2 to Expand in Sweden
Scaleway selected Sweden to tap into the country’s reliable access to low-carbon electricity and established technology talent pool. The deployment builds directly on an existing partnership with Swedish telecommunications operator Tele2 formed in May to scale local cloud offerings. Stefan Trampus, EVP of Tele2 B2B, stated in Scaleway’s announcement that the Swedish installation strengthens local conditions for domestic companies to scale artificial intelligence operations.
The project forms part of a broader infrastructure push that includes a planned Nvidia HGX B300 GPU Zone in Copenhagen, Denmark. Across Paris, Copenhagen, and Stockholm, Scaleway’s expanding footprint aims to aggregate tens of thousands of GPUs delivering tens of exaflops of capacity. This geographic scaling aligns with broader European Union objectives to boost regional computing autonomy through initiatives like the European Commission’s AI Gigafactories call for tenders, in which Scaleway and its parent organization, iliad Group, are participating via a French consortium.
| Location | Hardware Platform | Expansion Timeline |
|---|---|---|
| Paris, France | Mixed GPU Deployment | Active / Expanding |
| Copenhagen, Denmark | Nvidia HGX B300 Systems | Announced |
| Stockholm, Sweden | AMD Instinct MI355X | Announced |
| Frankfurt, Germany | Cloud Availability Zones | Scheduled for 2027 |
Unresolved Operational Metrics and Deployment Timelines
Key deployment details remain unconfirmed as integration planning continues. Scaleway has not disclosed the total number of GPUs allocated to the Swedish supercomputer, its aggregate computing capacity, the financial investment amount, or the exact operational launch date.
Disclaimer: The technical analyses and security protocols detailed in this article are for informational purposes only. Always consult with certified IT and cybersecurity professionals before altering enterprise networks or handling sensitive data.