Scaleway Review 2026

Quick Verdict

Scaleway is the right choice when you want to build your entire AI stack on a single European provider. The GPU offering (H100 SXM, L40S) is solid, and more importantly, it integrates natively with Scaleway's mature cloud platform: managed Kubernetes, S3-compatible object storage, private networking, and load balancers. European startups and enterprises that already run on Scaleway for their non-GPU workloads can add GPU compute without switching providers or dealing with data movement between clouds. The EUR pricing eliminates USD volatility for EU-budgeted teams. The trade-off is cost: Scaleway is more expensive than US GPU specialists like RunPod or Lambda Labs.

What is Scaleway?

Scaleway is a French cloud provider founded in 1999 (originally as Online.net), making it one of the oldest and most established European cloud platforms. In recent years it has built out a GPU offering alongside its existing compute, storage, and Kubernetes services. The company operates two main regions: FR-PAR (Paris, France) and NL-AMS (Amsterdam, Netherlands), both fully within the EU.

The GPU lineup focuses on the current generation: H100 SXM (the highest-performance H100 variant with NVLink interconnect) and L40S (a versatile 48GB GPU well-suited to both inference and moderate training). Per-minute billing and 100GB free monthly egress are practical touches that reduce total cost for European teams.

Best For

• European startups on a single cloud — teams that want compute, storage, Kubernetes, and GPU from one provider and one invoice.
• GDPR-compliant AI inference — production inference APIs that need EU data residency with managed Kubernetes integration.
• Kubernetes-native AI deployments — Scaleway Kapsule integrates directly with GPU nodes for ML serving workloads.
• EU enterprise teams — organizations that need EUR contracts, EU invoicing, and EU support without compromising on GPU hardware quality.

Scaleway vs OVH — French Cloud Rivalry

OVH and Scaleway are the two dominant French cloud providers, and the comparison is instructive. OVH has more raw compute capacity and a longer enterprise track record, but its GPU lineup is older — V100 and T4 are still prominent in the catalog, with H100 availability limited outside committed contracts. Scaleway's GPU selection is more current (H100 SXM, L40S) and the interface is cleaner. OVH wins on total platform breadth (dedicated servers, CDN, domain registrar, bare metal). Scaleway wins on GPU hardware freshness and a more modern developer experience. For GPU-primary workloads, Scaleway is the stronger choice of the two.

Scaleway vs Hetzner — German vs French EU

Scaleway is both cheaper and more flexible here. It starts at €0.79/h on genuine cloud instances you can stop, while Hetzner's cheapest GPU server works out to €1.42/h and is a dedicated machine on a monthly contract that keeps billing whether you use it or not. Hetzner's catalogue is also two models deep, with no A100 and no H100, so anyone needing H100 SXM in the EU has no Hetzner option at all. Hetzner further lacks managed Kubernetes GPU nodes and the broader ecosystem Scaleway ships. Hetzner's case is narrow: dedicated German or Finnish hardware, 96 GB on a single card, and a workload that runs continuously enough to justify paying for the whole month.

Feature Tour

Scaleway's control panel is polished and well-maintained. GPU instances launch in under 5 minutes from the web console. The Terraform provider and API are well-documented — Scaleway clearly supports infrastructure-as-code workflows. CLI tooling (scw CLI) covers all resource types including GPU instances.

The Kubernetes integration (Kapsule) is particularly strong. You can create a Kubernetes cluster with GPU node pools, configure autoscaling, and deploy ML serving workloads with Helm charts — all within the Scaleway console. This removes the operational complexity of running your own GPU-capable Kubernetes cluster.

Object storage (compatible with the S3 API) is available in both regions and integrates directly with GPU instances for model weight storage. The free 100GB/month egress tier covers typical inference workloads without additional cost. For training workloads that move large datasets, egress costs should be factored into TCO calculations.

Support operates in French and English, with European business hours. Enterprise plans include dedicated technical account management and SLA guarantees. Community resources (documentation, Slack channel) are well-maintained.

Who Should Use Scaleway

Scaleway is ideal for European teams building cloud-native AI applications who want to stay within a single EU provider. If you're already using Scaleway for your non-GPU workloads, adding GPU compute is straightforward and keeps your architecture simple. If you need H100 SXM performance in Paris or Amsterdam specifically, Scaleway is one of very few providers that offer it.

Skip Scaleway if you primarily care about GPU price per hour (RunPod or Vast.ai are much cheaper), if you need H200 (not available at Scaleway), or if you need US datacenter locations.

Final Verdict

Scaleway earns a 4.0/5.0. It is not the cheapest or the most GPU-diverse option, but it is the most mature and integrated EU cloud platform with serious GPU capacity. For European teams building production AI applications who want EUR billing, EU data residency, and a complete cloud stack without mixing providers, it is a compelling choice. The H100 SXM availability in Paris and Amsterdam is a real differentiator in the EU GPU market.