Scaleway: Europe's Sovereign Cloud Giant You Probably Haven't Deployed On Yet
Kim Boender
If you're a developer in Europe, or anywhere who cares about data sovereignty, GDPR compliance, sustainable infrastructure, or just having a genuinely full-stack cloud alternative to AWS and GCP, Scaleway deserves a serious look.
It's been quietly building since 1999 and today has grown into one of the most comprehensive cloud platforms on the continent: over 100 products spanning bare metal, virtual compute, GPU clusters, AI inference, Kubernetes, serverless, managed databases, quantum computing (yes, really), and much more. All of it hosted in European data centers. All of it under European law.
And yet, outside of the French tech scene where it's a household name, Scaleway flies under the radar for a lot of developers. Let's fix that.
The company behind the cloud
Scaleway is a subsidiary of the iliad Group, the French telecoms and technology conglomerate founded by Xavier Niel. iliad operates Free (the French ISP that disrupted the market with aggressive pricing in 2012), Iliad Italy, and Play in Poland. Scaleway is their cloud and infrastructure arm.
The company traces its roots back to 1999, when it launched under the Online.net brand as a web hosting provider. It evolved from shared hosting into dedicated server hosting (Dedibox), then into a full public cloud platform. The Scaleway brand was formally adopted in 2015.
This heritage matters. Unlike cloud startups that raised VC money and built on top of AWS, Scaleway owns its own physical infrastructure. It operates data centers in Paris, Amsterdam, and Warsaw, runs its own network, and controls the full stack from the physical layer up. That ownership gives it both the ability to price competitively on compute-heavy workloads and a genuine claim to data sovereignty.
Customers include Hugging Face, France Televisions, Mistral AI, and the French Ministry of Education. These are institutions that need French data sovereignty, not just prefer it.
Why Scaleway, and why now?
The European cloud market is having a moment. Regulations like GDPR, EU data localization requirements, and the broader conversation around digital sovereignty have made "where is your data hosted?" a genuine business and compliance question.
The US hyperscalers can comply with GDPR, but they're ultimately US companies subject to US law, including the CLOUD Act, which grants US government agencies the ability to compel data disclosure from American companies' servers regardless of where those servers physically are. For many European public institutions, healthcare companies, legal firms, and tech startups, that's not an acceptable trade-off.
Scaleway's answer is direct: it's a European company, owned by a European group, operating under French and EU law. Your data is in Europe and stays in Europe.
But sovereignty alone isn't a differentiator if the product isn't good. Scaleway has spent the past decade building a technically competitive stack, and in several areas, particularly AI infrastructure and quantum computing, it's actually ahead of its European counterparts.
The product stack
Scaleway's catalog has over 100 products. The structure is logical once you understand the layers.
Bare metal. Dedibox is their classic dedicated server line: affordable bare metal machines popular with European developers building hobby projects and production apps alike. Dedibox servers start cheap and have been a staple for years. Elastic Metal is the newer, cloud-native bare metal tier: fully dedicated physical servers with API-driven provisioning, per-hour billing, and tight integration with the rest of the Scaleway cloud ecosystem.
Virtual compute. Scaleway's Virtual Instances are organized into four tiers: Development Instances for testing and staging, General Purpose for most production web apps, Compute Optimized for CPU-bound workloads like video encoding and batch processing, and Memory Optimized for in-memory databases and caching layers.
GPU and AI. This is where things get genuinely interesting. Scaleway has positioned GPU and AI infrastructure as a strategic priority. GPU Instances range from the L4 up to the H100 PCIe (up to 8 GPUs per instance, reaching up to 1,513 TFLOPs in FP16). For serious AI training, there are AI Supercomputers powered by NVIDIA DGX H100 systems. Scaleway's flagship supercomputer, Nabu, has 1,016 NVIDIA H100 Tensor Core GPUs. Mistral AI used Nabu to train early versions of their models.
For inference, Scaleway offers Generative APIs for serverless inference of popular models like LLaMA and Mistral (pay per million tokens, no infrastructure to manage) and Managed Inference for dedicated GPU infrastructure. In 2025, Scaleway became the first European sovereign inference provider on Hugging Face, meaning you can run Hugging Face models on European infrastructure directly from the Hub.
Kubernetes and containers. Scaleway's Kubernetes Kapsule is a fully managed Kubernetes service, one of the first from a European cloud provider. Kubernetes Kosmos is a unique offering that lets you run Kubernetes pods across multiple cloud providers simultaneously, useful for disaster recovery or geographic distribution requirements.
Serverless. Scaleway was early here: they launched the first European general-purpose serverless compute platform in 2021. The serverless suite covers Functions (think AWS Lambda, but European), Containers for stateless containerized workloads, and Jobs for batch processing and scheduled tasks.
Storage. Object Storage is S3-compatible, a critical detail: most tools that work with AWS S3 will work with Scaleway's Object Storage with a quick endpoint change. Block Storage provides persistent, high-performance volumes for instances.
Networking. The networking stack includes Virtual Private Cloud, Private Networks, Public Gateway, Load Balancer, Edge Services (CDN-like caching plus WAF), and Scaleway InterLink for private connectivity from on-premises infrastructure.
Managed data services. Managed Databases for PostgreSQL and MySQL, Managed Redis, Managed MongoDB, Data Warehouse for ClickHouse, and Messaging/Queuing services.
Quantum computing. In 2024, Scaleway became the first European Quantum as a Service platform combining both quantum emulation and real Quantum Processing Units. This is a partnership with Quandela, a French quantum computing startup that builds photonic quantum computers. For most developers, quantum is firmly in "things to watch" territory, but the fact that Scaleway is the only European cloud provider building this out is significant for the research community.
The sustainability story
All of Scaleway's data centers run on 100% renewable energy: specifically wind or hydro power backed by Guarantees of Origin, a European certification standard. This isn't offset credits, it's actual renewable energy sourcing.
The efficiency metrics are also published publicly. DC5 in Paris has a PUE of 1.25 (the industry average is around 1.6). AMS3 in Amsterdam achieves a PUE of 1.2, among the best in the industry. WAW2 in Warsaw reaches 1.24. Scaleway also has real-time PUE monitors for their data centers: public dashboards you can visit to see live efficiency data. That level of transparency is rare.
For teams that have sustainability commitments or need to demonstrate them to clients or investors, this is real, documented, auditable data.
Pricing
Scaleway prices in Euros, which matters for European companies dealing with EUR-based budgets and FX risk on AWS or GCP bills. Pricing is transparent and published for all products.
Development instances start at a few cents per hour. General Purpose instances (8 vCPUs, 32 GB RAM) run roughly 0.18-0.40 euros per hour. L4 GPU instances start around 0.50-0.60 euros per hour. H100 PCIe instances run at the high end, consistent with market rates.
One developer-friendly touch: Scaleway offers a free-tier sandbox with no time limit (unlike AWS's 12-month window) that includes Serverless Functions, Object Storage, and more. New accounts also receive 750 GB of multi-AZ Object Storage free for the first 90 days.
A track record of firsts
Scaleway has a consistent record of shipping things before other European cloud providers do. 2018: first European cloud provider with 3-AZ region architecture. 2020: first managed Kubernetes from a European cloud provider. 2021: first European general-purpose serverless compute platform. 2022: first European cloud provider with native IAM. 2023: first public cloud GPU supercomputer in Europe. 2024: first European Quantum as a Service platform. 2025: first European sovereign inference provider on Hugging Face.
This is a meaningful innovation track record for a company that isn't Google or Amazon.
How it compares
Vs. AWS/GCP/Azure: Scaleway can't match the breadth of services or global footprint. But for workloads that need to stay in Europe, it's often cheaper, simpler, and cleaner from a compliance standpoint. The S3-compatible storage means existing toolchains work with minimal changes.
Vs. Hetzner: Hetzner is excellent for simple, affordable compute and storage, and hard to beat on raw price for VMs. Scaleway beats it on breadth: Hetzner doesn't have Kubernetes, serverless, GPU clusters, managed databases, a quantum platform, or a full AI ecosystem. If you're going to grow, Scaleway is less likely to require a migration.
Vs. OVHcloud: Scaleway's closest direct competitor. OVH is bigger by revenue and has more global presence. Scaleway tends to have a better developer experience, cleaner UX, and is more aggressive on AI and quantum innovation.
Vs. Render/Railway: These are PaaS platforms, not IaaS clouds. If you just want to deploy an app without thinking about infrastructure, Railway or Render are simpler starting points. Scaleway is a full cloud for when you need the control, the raw compute, the AI infrastructure, or the European compliance story.
Getting started
Scaleway's onboarding is straightforward. Create an account at scaleway.com, verify your identity (standard cloud provider requirement), and you'll get access to the free tier immediately. The console is clean and well-organized, Terraform support is mature, and the documentation quality is generally good.
If you're migrating from AWS, the S3-compatible Object Storage is the easiest starting point. Point your SDK at Scaleway's endpoint and change your credentials. That's it.
The bottom line
Scaleway is impressive and quietly underrated outside of France. A company that started hosting websites in 1999 has built a platform that today can train large language models, run quantum circuits on real photonic QPUs, serve serverless functions, host Kubernetes clusters, and do it all on 100% renewable energy from European data centers under European law.
If you're building in Europe, or if your users, data, or regulators are in Europe, Scaleway should be on your shortlist. The pricing is competitive, the product breadth is real, the sustainability story is verifiable, and the sovereignty argument is genuine.
For AI teams in particular, the combination of on-demand H100 GPUs, supercomputer clusters, serverless inference, and Hugging Face integration makes Scaleway the most credible European alternative to AWS's AI infrastructure stack. The fact that Mistral AI trained on Scaleway's infrastructure is the endorsement that says more than any marketing copy.