Every conversation about EU digital sovereignty eventually runs into the same wall: leadership wants workloads off AWS for compliance or political reasons, and the platform team says it can’t be done without rebuilding half the stack by hand. Both sides are right. The infrastructure layer — compute, storage, networking — is essentially commoditized across European providers. The managed services layer above it is not, and that gap is the real reason migrations stall.
At Gart Solutions, we just finished a service-by-service comparison of AWS’s managed catalog against seven European providers — OVHcloud, Scaleway, Hetzner, Exoscale, IONOS Cloud, UpCloud, and the newer sovereign entrant evroc. The full research is published at The EU Cloud Managed Services Gap: an AWS Capability Breakdown and Sovereign Stack Opportunity. This article walks through what we found, why it matters for platform teams planning a move, and where the opportunity sits for systems integrators willing to build the layer the providers haven’t.
The Headline Numbers
AWS’s Frankfurt and Dublin regions expose the same catalog as any US region — over 200 managed services, full stop. On the European side, the picture is uneven by design rather than by accident:
OVHcloud ships roughly 45 managed services — the deepest catalog of any EU-native provider.
Scaleway sits around 30, but it’s the most developer-complete: serverless SQL, true FaaS, and a NATS-based queue/pub-sub service in one platform.
Hetzner offers about 8, and that’s intentional. Hetzner has explicitly avoided building a database-as-a-service control plane, partly because of the engineering investment and partly because of AGPL licensing risk on integrated tooling.
The definition matters here, and it’s worth being precise about it: a managed service is one where the provider owns patching, failover, backup, scaling, and SLA enforcement through an API. A Docker container running PostgreSQL on a VM is not a managed database — it’s a self-hosted database that happens to live in the cloud. Most of the friction in EU migrations comes from teams conflating the two.
Where the Gap Actually Bites
Relational Databases: RDS is Matched, Aurora is Not
Managed PostgreSQL and MySQL are the single biggest migration blocker we see in practice. AWS splits this into two products with materially different architectures — RDS (traditional managed, Multi-AZ failover under 35 seconds, up to five read replicas) and Aurora (distributed storage, scale-to-zero, up to 128 TiB, 15 read replicas, sub-30-second failover).
EU providers compete reasonably well at the RDS tier. OVHcloud’s Enterprise plan reaches 99.99% SLA on three-AZ deployments. Exoscale and UpCloud both run on Aiven’s DBaaS engine, with pgBouncer pre-configured and point-in-time recovery on Business-tier plans. IONOS added PostgreSQL 16 in January 2026 with seven-day PITR. Scaleway’s Serverless SQL is the most interesting product in the category — auto-scaling 0–15 vCPU with true scale-to-zero billing — but it caps at 1 TB, has no PITR on the serverless tier, and covers roughly 30% of Aurora Serverless’s feature set.
Nobody has built Aurora’s distributed, six-copy storage architecture. That’s a multi-year engineering effort, and at current investment rates it’s not coming. For data-heavy teams that depend on Aurora Serverless for variable workloads, there is no native migration path today — which is the single most common reason we see companies keep a hybrid AWS footprint even after deciding, in principle, to move.
Kubernetes is the One Category That’s Actually Competitive
This is the bright spot. Every provider we evaluated except Hetzner ships managed Kubernetes natively, and the differences are mostly in depth rather than existence. Exoscale’s SKS deploys clusters in under two minutes with native Karpenter support. Scaleway’s Kapsule pairs with Kosmos, a multi-cloud Kubernetes federation layer with no direct AWS equivalent. OVHcloud and IONOS both offer free control planes.
The recurring gap inside Kubernetes is shared storage: AWS EFS makes ReadWriteMany volumes trivial, letting hundreds of pods mount the same filesystem. On most EU clouds that means either over-provisioning File Storage (OVHcloud’s minimum order is 150 GiB) or hand-rolling NFS. For stateful microservices, media pipelines, and CI/CD shared caches, that’s a recurring operational cost that lands squarely on the customer.
Observability: the Most Painful Gap of All
CloudWatch, X-Ray, and OpenSearch Dashboards form one integrated bill with effectively zero infrastructure to manage. Nothing in the EU market comes close. Scaleway’s Cockpit (managed Grafana, Loki-backed logs, Prometheus-compatible metrics) covers about half of CloudWatch’s capability but has no distributed tracing. OVHcloud’s combination of Logs Data Platform and Managed Grafana gets to roughly 60% — again, no tracing, no auto-instrumentation. Hetzner, Exoscale, and UpCloud offer no native stack at all; the documented pattern across all three is the same self-assembled Prometheus + Grafana + Loki + Alertmanager deployment, run on a dedicated VM, owned entirely by the customer’s engineering team.
In our work helping teams plan EU migrations, this is consistently the gap that derails the sovereignty argument. Without a dedicated SRE function, customers end up either running CloudWatch against their EU workloads — which defeats the point — or maintaining an unmanaged observability stack that quietly becomes a second full-time job. A managed, EU-sovereign Prometheus/Grafana/Loki/Tempo offering is, by a wide margin, the highest-demand, lowest-supply category in this market.
Serverless, Messaging, and the AI/ML Pipeline
Three more categories worth flagging briefly:
FaaS. Outside Scaleway Functions, true scale-to-zero serverless compute essentially doesn’t exist in the EU cloud market. “Serverless” elsewhere usually means serverless containers — still always-running compute, just containerized. For event-driven teams used to gluing services together with Lambda, the EU alternative is running 24/7 containers at €20–80/month per function, a cost that compounds with every microservice boundary.
Messaging. Managed Kafka is well covered — OVHcloud, Scaleway, Exoscale, and IONOS all ship it. What’s missing is lightweight queuing: the SQS use case of simple async task dispatch. Outside Scaleway’s NATS implementation, there’s no equivalent, and teams end up self-hosting RabbitMQ or NATS to fill the gap.
AI/ML platform. This is the most total gap of the seven categories we looked at. GPU compute and inference APIs exist across most providers — OVHcloud’s AI Deploy, Scaleway’s Model-as-a-Service, IONOS’s Berlin-isolated AI Model Hub. But the MLOps pipeline layer — experiment tracking, model registry, A/B deployment, drift monitoring — is absent everywhere. Teams either self-host MLflow on Kubernetes or keep their ML production systems on SageMaker or Vertex while migrating everything else, creating a split architecture that never fully resolves.
Backup Orchestration and the DORA Problem
AWS Backup provides a single control plane across RDS, EBS, EFS, DynamoDB, and EC2, with cross-region replication, compliance reporting, and vault lock. Every EU provider we evaluated does backup per-service — databases back themselves up, VMs get snapshots — with no coordination layer tying it together.
This isn’t just an operational inconvenience. Under DORA Article 12, financial entities are required to document, test, and attest their ICT backup and recovery arrangements; “we back up our databases” doesn’t satisfy that requirement on its own. Without a coordinated backup service, the attestation burden falls entirely on the customer or their integration partner — which is exactly the kind of gap that turns into a standing managed-service relationship rather than a one-off project.
What This Means If You’re Planning a Migration
A few practical takeaways for platform and infrastructure teams evaluating a move:
Map workloads to the gap, not the provider. If you’re running Aurora Serverless, a true Lambda-style FaaS pattern, or an MLOps pipeline on SageMaker, plan for either a hybrid architecture or a build project — there’s no drop-in replacement today.
Kubernetes migrations are the least risky. Every major EU provider except Hetzner has a credible managed offering, and the feature gaps (mainly RWX storage and advanced autoscaling) are solvable.
Budget for observability as its own line item. Whatever provider you choose, assume you’re either operating a self-hosted Prometheus/Grafana/Loki stack or buying that capability from a third party. It is not coming bundled.
DORA and NIS2 deadlines change the calculus. If you’re a regulated entity, the backup orchestration and compliance-attestation gap isn’t optional — it needs a plan regardless of which provider you pick.
Why This is a Systems Integrator Opportunity, Not Just a Vendor Problem
The gap isn’t closing from the provider side fast enough to matter on a typical migration timeline. It closes from above — from integrators and managed service providers who build the operating layer and deliver it on top of commodity EU IaaS. That’s the thesis underlying the research: a managed observability platform, a Patroni-based managed PostgreSQL service, a Knative-based FaaS runtime, DORA-compliant DR orchestration, and an EU MLOps platform are all buildable today on existing infrastructure, and none of them compete with what the underlying cloud providers are shipping natively.
There’s also a structural reason this works in the integrator’s favor rather than the provider’s: European cloud providers generally lack the enterprise sales motion to reach regulated buyers who won’t do self-service procurement, and AWS’s CLOUD Act exposure has no clean fix — there’s no path to SecNumCloud-equivalent status for a US hyperscaler. Providers like evroc, still ramping toward production scale, need integrator ecosystems to reach enterprise customers at all. Whoever builds the managed layer first tends to keep the account relationship, not the IaaS vendor underneath it.
The full breakdown — provider-by-provider deep dives, pricing references, SLA tables, and the sovereign-stack opportunity sizing — is available in the complete report: The EU Cloud Managed Services Gap: an AWS Capability Breakdown and Sovereign Stack Opportunity.
