Scaleway vs STACKIT

Elastic Metal, Dedibox and Apple Silicon bare metal

Single-tenant physical servers without a shared hypervisor layer: Dedibox for a wide dedicated catalogue, Elastic Metal for bare metal integrated with Scaleway VPC, load balancing, storage and Kubernetes, plus Apple Silicon Mac mini hosts for native macOS/iOS CI. Best when you need hardware isolation, custom kernels, or GPU bare metal without VM overhead—product lines are still converging, so compare availability zone coverage per SKU.

EU-resident GPU instances for AI training and inference

On-demand GPU VMs with recent NVIDIA options (including L4, L40S, H100 PCIe/SXM and B300-SXM) aimed at LLM fine-tuning, inference and graphics/media acceleration, with data residency in European regions. Integrates with Kapsule via the NVIDIA GPU Operator; pay-as-you-go by the minute. GPU capacity and Savings Plan eligibility differ from general instances—validate stock and zone before committing multi-node jobs.

Kubernetes Kapsule and multi-cloud Kosmos

Managed Kubernetes control planes on Scaleway nodes (Kapsule) with autoscaling-oriented operations, plus Kosmos for hybrid/multi-cloud worker pools including non-Scaleway infrastructure. Suits teams standardising on portable Kubernetes rather than proprietary orchestrators; external Kosmos pools remain your responsibility to patch and size.

Multi-AZ European regions with VPC networking

Place compute and storage in Paris, Amsterdam and Warsaw multi-AZ regions (docs also reference Milan), using regional private networking, load balancers and related network services. Useful for EU latency and residency designs without a US region on the primary map—confirm each product’s AZ matrix in the product-availability guide before multi-region DR planning.

Managed storage and databases on open-ish APIs

S3-compatible Object Storage (including multi-AZ class options), block volumes, and managed database engines (PostgreSQL, MySQL, Redis, MongoDB, ClickHouse-oriented analytics) plus serverless containers/functions/jobs. Reduces day-2 ops for common stack pieces while keeping export paths closer to open standards than pure proprietary PaaS—always check engine versions, HA options and backup retention for production.

API, CLI and Terraform-first operations

Full console plus documented APIs, CLI and Infrastructure-as-Code workflows for provisioning instances, networks and managed services. Fits platform teams automating environments; IAM permissions and account Owner roles still need deliberate design for production orgs.

Compute Engine VMs and GPUs in group-owned EU regions

Provision Linux and Windows virtual machines, GPU-backed instances, automated server backup and OS update management from the STACKIT portal, CLI, API, or Terraform. Workloads land in European cloud regions such as EU01 (Germany) and EU02 (Austria) on Schwarz Group–operated infrastructure rather than rented hyperscaler bare metal. Best for teams that need predictable EU residency for general compute and AI/ML training or inference on GPU shapes—confirm available instance families per region before migration.

STACKIT Kubernetes Engine (SKE)

Managed, CNCF-compliant Kubernetes with a highly available control plane, automated Kubernetes/OS upgrades, repair functions, pod and node autoscaling, optional temporary cluster shutdown, and automation via Terraform, SKE API, and CLI. Suited to cloud-native apps, stateful workloads on block storage, and ML pipelines that must stay in European regions. Not a self-hosted kubeadm replacement—you consume a managed service with STACKIT-controlled plane components.

Managed databases, messaging, and observability

Fully managed PostgreSQL Flex, MongoDB Flex, SQL Server Flex, MariaDB, Redis, OpenSearch, and RabbitMQ plus LogMe, Logs, and Observability stacks reduce ops load for application teams. Flex models emphasize automated maintenance, backups, and scaling inside the STACKIT cloud. Ideal when you want PaaS data services under the same German operator as your VMs/K8s—validate HA topology, backup retention, and region pairing for each service.

Confidential computing and key/secrets control

Confidential Server and Confidential Kubernetes options aim to protect data in use with hardware-backed isolation; Secrets Manager and Key Management Service handle secret storage/rotation and cryptographic operations. Useful for regulated or multi-tenant sensitive workloads beyond disk encryption alone. Confirm attestation models, supported node types, and which compliance reports cover these products.

EU colocation with hybrid path to public cloud

Racks, cages, and private rooms in Neckarsulm, Ellhofen, and Ostermiething facilities—with remote hands, carrier connectivity, and hybrid designs that keep non-migratable hardware next to STACKIT public cloud. Positions STACKIT for gradual cloud adoption without a pure forklift. Colocation is quote-driven and site certifications can vary by facility; request site data sheets early.

Object, block, file, backup, and archiving storage

S3-compatible object storage, high-performance block volumes, NFS file storage, backup storage, and audit-oriented archiving for retention workloads. Supports application data, VM disks, and compliance archives inside the same European footprint as compute. Check durability/replication claims and cross-region options against your RPO/RTO rather than assuming hyperscaler multi-region defaults.