Accessibility Adjustments

Use these optional tools to adjust reading and display preferences. These tools cannot resolve every accessibility barrier. Please contact the website owner if you need assistance.

  • Text adjustments
  • Content scaling 100%
  • Font size 100%
  • Line height 100%
  • Letter spacing 100%
  • Colour adjustments
  • Orientation adjustments

Scaleway outlines AMD and NVIDIA expansion in the Nordics

Listen to this article

Scaleway announced an AMD powered supercomputer deployment in Sweden on October 13. The company announcement specifies Instinct MI355X GPUs and describes the project as its first supercomputer deployment outside France.

A separate announcement from the same day sets out plans for a Copenhagen GPU Zone using NVIDIA HGX B300. Together, the projects would broaden the European cloud providerโ€™s Nordic footprint across two accelerator suppliers.

What Sweden is getting

Scaleway says the Swedish system will use AMDโ€™s ROCm software. Bull will integrate the cluster, Giga Computing will supply servers, HPE Networking will provide Ethernet networking and VAST Data will handle the data platform. Motul is supplying liquid cooling fluid.

AMDโ€™s MI355X specification lists 288 GB of HBM3E memory per GPU and peak memory bandwidth of 8 TB per second. The accelerator uses the CDNA 4 architecture and supports reduced precision formats including MXFP4 and MXFP6. AMD lists its original launch date as June 12, 2025.

Copenhagen takes the NVIDIA route

Scaleway describes GPU Zones as dedicated locations for AI training and inference, complementing its existing cloud Availability Zones.

NVIDIAโ€™s HGX platform documentation describes HGX B300 as an eight GPU system using Blackwell Ultra accelerators and fifth generation NVLink. Its published specification lists more than 2 TB of total memory and 14.4 TB per second of aggregate NVLink bandwidth.

AMDโ€™s memory figure describes one GPU, while NVIDIAโ€™s figure covers eight. Those component specifications alone cannot establish the total memory or application performance of either new cloud installation.

NVIDIAโ€™s HGX reference architecture uses NVLink within the baseboard and ConnectX-8 between systems. The Copenhagen release does not specify its networking configuration.

Planned scale and customer availability

Scalewayโ€™s broader AI Pulse announcement says Copenhagen will contain several thousand B300 GPUs and identifies Stockholm as the location for the AMD cluster. It gives no GPU count for Sweden.

Across GPU Zones in Paris, Copenhagen and Stockholm, Scaleway describes an eventual ambition of tens of thousands of GPUs and tens of megawatts of capacity. Those are portfolio goals rather than installed capacity confirmed at either new Nordic site.

The reviewed announcements provide no exact Copenhagen GPU count, customer prices or date when customers can book either new site. Denmark is explicitly presented as a planned opening. The Sweden release announces a deployment without establishing general customer availability.

Archival server racks at NOIRLab headquarters in Tucson, Arizona, shown as a general infrastructure illustration. Photo by NOIRLab/NSF/AURA/P. Horรกlek, shared in 2023 under Creative Commons Attribution 4.0. Converted to WebP.

Marcus Reid
Marcus Reid

Marcus Reid is focused on covering the money, rules, and institutional choices shaping AI. He runs from funding rounds and chip deals to regulation, lawsuits, leadership changes, and the business of building enormous computing systems. Marcus follows the incentives behind the announcement. Who pays, who gains leverage, and what changes for everyone else? The voice is direct, measured, and occasionally dry, especially when a grand promise arrives with very little detail.

Leave a Reply

Your email address will not be published. Required fields are marked *

Gravatar profile