Technology · NEWS EXPLAINER
Scaleway plans AMD capacity in Sweden and Nvidia GPUs in Copenhagen
The Nordic expansion offers two hardware paths for European AI workloads. The announcements identify the platforms, but do not establish when customers can book the new capacity or what it will cost.
Scaleway announced two Nordic AI infrastructure projects at ai-PULSE in Paris on October 13, 2026: an AMD Instinct MI355X system in Sweden and an Nvidia HGX B300 GPU Zone in Copenhagen. These are announced deployments, not proof that either new cluster is already accepting customer workloads.
Two locations, two announced platforms
The Swedish project is Scaleway’s first supercomputer deployment of this kind outside France, according to its release. The system will use AMD Instinct MI355X accelerators and AMD’s ROCm software. Scaleway’s wider event overview locates the cluster in Stockholm. The Copenhagen announcement separately specifies Nvidia HGX B300 and describes a new GPU Zone for training and inference.
Scaleway calls the Swedish system Europe’s first commercial AMD AI supercomputer. That is the company’s characterization, not a claim that Europe has never used AMD for AI or supercomputing. Reuters notes that AMD already powers research systems such as Finland’s LUMI, and other cloud providers rent AMD capacity to European customers.
The hardware partners and regional rationale
For Sweden, Scaleway names Bull for cluster integration, installation and ongoing maintenance; Giga Computing for servers; HPE Networking for Ethernet; VAST Data for the data platform; and Motul for cooling fluid. The announcement describes their planned roles rather than reporting completed installation or measured system performance.
Scaleway points to low-carbon electricity and local technology ecosystems when explaining the Nordic locations. Its Copenhagen release also cites the cooler climate as a factor that can reduce cooling requirements. These are location and design rationales; the reviewed material does not establish a measured carbon footprint or a performance-per-watt comparison for the finished clusters.
A GPU Zone is not the same promise as three cloud zones
Scaleway defines GPU Zones as dedicated locations providing large-scale GPU capacity alongside existing cloud Availability Zones. Its wider overview separately plans three Availability Zones in Stockholm and three more in Frankfurt for 2027. Those regional cloud plans should not be read as an exact opening schedule for either newly announced GPU cluster.
The overview says the Copenhagen cluster will feature several thousand GPUs. It describes eventual capacity across Paris, Copenhagen and Stockholm in much larger aggregate terms, without allocating that total to each new site. DCD reports that the Swedish system’s GPU count and expected compute capacity were not provided. Do not turn the portfolio ambition into a per-site specification.
What a prospective customer still needs to confirm
The practical development is a planned expansion of hardware choice and location for European AI work. The reviewed announcements do not publish an exact go-live date, a reservation window or pricing for the new clusters. A customer planning a workload should confirm actual regional availability, contract capacity, software support and total cost before scheduling a migration.
The project announcements establish the selected hardware and partners, not a benchmark proving one platform is faster or cheaper for a particular model. Keep that distinction when comparing offers: a deployment plan, a bookable service and a successful workload test are separate milestones.
