Resources

A service-by-service breakdown of what EU clouds actually ship today versus AWS — and where the sovereign stack opportunity sits for systems integrators. Section 1 The Gap in Numbers AWS Frankfurt and Dublin regions expose the same full catalog of 200+ managed services as any other AWS region. European IaaS-first providers range from near-zero PaaS (Hetzner) to a credible but incomplete catalog (OVHcloud, Scaleway). The gap is not closing fast — it is structural. Cloud Provider Comparison 200+ Services in AWS full catalog — same in EU regions ~45 Managed services at OVHcloud — the deepest EU catalog ~30 Scaleway's catalog — the most developer-complete ~8 Hetzner managed services — intentionally minimal IaaS What "managed service" means here A cloud API-driven service where the provider owns patching, failover, backup, scaling, and SLA enforcement. A VM running a Docker container of PostgreSQL is not a managed database — it is a self-hosted database that happens to live on a cloud VM. → This distinction is the core of every migration conversation. Section 2 Relational Databases: RDS & Aurora vs. EU Managed PostgreSQL and MySQL are the single biggest migration blocker in the EU cloud stack. AWS offers two separate products here — RDS (traditional managed) and Aurora (cloud-native distributed) — each with materially different architecture. EU providers mostly match RDS-tier; none match Aurora. AWS RDS PostgreSQL — the baseline Traditional managed PostgreSQL, the reference most EU clouds try to match Global Infrastructure Baseline Amazon RDS PostgreSQL SLA: 99.95% Multi-AZ ✓ Lifecycle Management: Automated engine patching and minor version upgrades. ✓ High Availability: Multi-AZ standby deployments with auto failover transitions under 35 seconds. ✓ Horizontal Scaling: Provision and scale up to 5 individual read replicas natively. ✓ Continuous Backup: Point-in-time recovery options covering a rolling 35-day retention window. ✓ Elastic Storage: Automatic dynamic volume expansion up to 64 TB using gp3 or io1 blocks. ✓ Observability Stack: Real-time engine health diagnostics via Performance Insights and Enhanced Monitoring hooks. ✓ Enterprise Security: Rigid isolation via IAM database authentication alongside native KMS encryption-at-rest. ✓ Traffic Management: Native RDS Proxy integration to orchestrate high-velocity application connection pooling. 🇫🇷 OVHcloud Plans: Essential (1 node, no SLA), Business (2 nodes, SLA 99.9%), Enterprise (3 nodes, SLA up to 99.99% Multi-AZ). Supported: PG 14–16, MySQL 8.x. Extensions: pgvector, PostGIS, TimescaleDB. 🇫🇷 OVHcloud Technical Constraints Backup retention: Configurable, plan-dependent matrix. No PITR on Essential tier clusters. No connection pooler managed by the platform — client-side PgBouncer proxy layer is strictly required. Storage adjustments: Volume scale upgrades are natively supported, but no downscaling volume paths are permitted once provisioned. Network Isolation: Native vRack private network integration is readily available on Business and higher tiers. 🇫🇷 Scaleway 🇫🇷 Scaleway Database Infrastructure Matrix 🖥️ (1) Classic Managed Database Traditional provisioned instances designed for predictable production workloads. Baseline Instance: PRO2-XXS tier ~€80/mo (2 vCPU, 8 GB RAM, 20 GB). High Availability: HA add-on configuration roughly doubles the base instance cost. Recovery window: Native PITR coverage supported up to 7 days. Available Regions: Paris, Amsterdam, Warsaw, Milan (New March 2026) ⚡ (2) Serverless SQL Dynamic engine tier featuring auto-scaling and automatic scale-to-zero operational billing. Compute Scale Bounds: Auto-scales dynamically across 0–15 vCPU paths (with 4 GB RAM allocated per vCPU). Inactivity Sleep: Automatically enters idle sleep states after 5 minutes of total traffic silence. Data Limits: Hard capacity thresholds at a 1 TB maximum storage ceiling and 1,000 concurrent connection limits. Recovery Gaps: No PITR supported on the serverless architecture tier. Query Engine Boundary: Hard maximum query size limit capped at 1,024 KB. Available Extensions (30+ total): pgvector PostGIS timescaledb 🇩🇪 Hetzner No managed PostgreSQL. Legacy konsoleH hosting offers MySQL/PG but with no cloud API, no autoscaling, no SLA. Engineers must deploy Patroni + pgBackRest via Autobase on Hetzner VMs — full ops responsibility. Connection pooling, failover, backups: all self-managed.  Maximum SI surface. 🇨🇭 Exoscale Plans: Hobbyist/Startup (SLA 99.95%, no standby, daily backup 2-day retention), Business/Premium (SLA 99.99%, standby nodes, PITR enabled). pgBouncer pre-configured on all plans. Forking (point-in-time clone) available on Business+. Read replicas via DBaaS integration API. Vertical/horizontal scaling without downtime. Powered by Aiven engine. All zones available (6 EU zones). 🇨🇭 Exoscale Database Tiers Engine Core: Powered by Aiven 🌱 Hobbyist / Startup Plans Availability Commitment: SLA 99.95% node uptime. Topology: Standalone deployment (no standby nodes provisioned). Backup Window: Standard daily backup path featuring a constrained 2-day data retention loop. 💼 Business / Premium Plans Availability Commitment: Elevated SLA 99.99% operational path. Topology: High Availability architecture featuring active standby replication nodes. Data Resilience: Point-in-time recovery (PITR) streaming enabled natively. Database Forking: Point-in-time cluster cloning workflows supported out of the box. ✓ Integrated Pooling: pgBouncer comes pre-configured across all deployment tiers. ✓ Read Replication: Read replicas scale out via the native cloud DBaaS integration API. ✓ Zero-Downtime Operations: Horizontal and vertical engine resizes roll out with no live traffic downtime. ✓ Geographic Footprint: Global availability across all 6 sovereign European zones. 🇩🇪 IONOS DBaaS PostgreSQL 14–16 (PG 16 added Jan 2026). Automated daily backups, 7-day retention with PITR. Replication modes: ASYNCHRONOUS or STRICTLY_SYNCHRONOUS (SYNCHRONOUS deprecated March 2026). API + DCD visual management. No read replicas on standard tier. Frankfurt region primary, additional EU locations planned. Grafana v12.3 Logging Service (March 2026 upgrade). 🇩🇪 IONOS Managed PostgreSQL Engine 📦 Database Core & Storage Engine Support: DBaaS PostgreSQL versions 14–16 (PG 16 deployed Jan 2026). Data Resilience: Automated daily backups with a strict 7-day retention loop backed by active PITR streams. Control Plane: Dual management access paths via programmatic API and DCD visual portal. Scale Bottlenecks: No read replicas supported on the standard deployment tier. Geographic Status: Frankfurt region primary (additional EU data hub footprints planned). ⚙️ Replication & Telemetry Supported Topology Modes: ASYNCHRONOUS STRICTLY_SYNCHRONOUS ⚠️ SYNCHRONOUS pipeline path deprecated March 2026. Managed Logging Stack: Telemetry is managed by the integrated Grafana v12.3 Logging Service platform upgrade rolled out in March 2026. Infrastructure Note: Enforces high data safety parameters on the synchronous replication options. 🇫🇮 UpCloud Managed Databases powered by Aiven (same engine as Exoscale). 100% uptime SLA claimed across products. PostgreSQL, MySQL, Redis, OpenSearch available. 15 global data centers. CNCF-certified Managed Kubernetes. Observability: self-assembly via Grafana on Aiven + Prometheus — no native managed stack. 🇸🇪 evroc No managed databases as of June 2026. Platform launched July 2025 — IaaS (compute, storage, networking) only, plus early AI services (roc agent, evroc Think platform). SUSE Rancher certified Q1 2026. Flagship data centers (Stockholm, Mougins) completing 2026. DBaaS is a direct gap for SI partnership. ⚡AWS Aurora PostgreSQL — the cloud-native tier Distributed architecture, instant scale-to-zero, 15 read replicas — nothing like this exists natively in EU clouds ⚡ The Cloud-Native Distributed Storage Deficit Evaluating the architectural capability gap between AWS Aurora and European alternatives AWS Reference Target Aurora PostgreSQL Serverless ✓ Distributed storage auto-scales natively up to 128 TiB ✓ True scale-to-zero capabilities (resumes engine workflows in ~15s) ✓ Provision up to 15 read replicas featuring millisecond replication lag ✓ Storage automatically replicated across 3 Availability Zones (AZs) ✓ Global Database framework for automated cross-region replication ✓ Aurora ML: Direct native SQL pipelines to SageMaker/Comprehend ✓ Backtrack engine: Rewind database logs instantly without state restoration ✓ Instant failover under 30 seconds via true compute/storage split architecture Uptime Target: SLA: 99.99% 🇫🇷 OVHcloud No Aurora Equivalent The Enterprise plan matches the 99.99% SLA via a 3-AZ configuration, but relies on classical streaming node replication paths rather than a distributed log-structured storage subsystem. Lacks scale-to-zero capabilities, Backtrack rewinds, and cross-region global databases. 🇫🇷 Scaleway ~30% Feature Parity The Serverless SQL Database functions as the closest European analog, supporting auto-scaling parameters (0–15 vCPU/RAM) and idling out after 5 minutes of total traffic silence. • Restrained by standard Block Storage volumes (1 TB maximum limit) • No cross-region application replication networks • No point-in-time recovery (PITR) supported on the serverless engine tier ⚠️ Strategic Infrastructure Gap No architectural Aurora-equivalent catalog product exists at Hetzner, Exoscale, IONOS, UpCloud, or evroc. Developing a replicated 6-copy engine cluster volume requires multi-year baseline R&D investments. At existing funding rates, this deficit remains a permanent structural barrier. Architectural Reality Check Aurora Gap Summary The cloud-native PostgreSQL tier — distributed storage, instant auto-scale, 128 TiB cap, Global Database — has no EU equivalent and no credible roadmap. Teams that depend on Aurora Serverless for variable workloads have no migration path today. → This is the primary architectural reason data-heavy companies stay on AWS even when they want to move. Section 3 Kubernetes Orchestration Managed Kubernetes is the most competitive managed-service category in the EU cloud market. Most providers offer something. The differences are in depth: autoscaling, node provisioning, networking features, and SLA. ⎈ Managed Kubernetes Feature Parity Matrix Benchmarking AWS EKS features against production-grade European sovereign cloud platforms Feature Vector 🇺🇸 AWS EKS 🇫🇷 OVHcloud MKS 🇫🇷 Scaleway Kapsule 🇨🇭 Exoscale SKS 🇩🇪 IONOS K8s 🇫🇮 UpCloud K8s 🇩🇪 Hetzner (DIY) Management & availability Parameters Control Plane Cost $0.10 / hr Free Free Free Included Included No native Deploy Time ~15 min ~5–10 min ~3–5 min < 2 min ~5 min ~3 min Via CFKE: ~5m Cluster SLA 99.95% 99.9% 99.9% 99.95% 99.9% 100% (Claimed) CFKE: 99.95% Node Orchestration & Scaling Layers Node Autoscaling Karpenter / CAS CAS Only CAS Karpenter Native CAS CAS hetzner-k3s Multi-AZ Pools Yes (3-AZ) Yes Limited regions Yes (All zones) Geo-redundant Yes N/A (Single DC) Networking Architecture & Data Storage CNI Customization VPC CNI / Cilium Cilium Supported Cilium / Calico Custom Support Default Only Default Flannel/Cilium (DIY) Multi-Cloud K8s EKS Anywhere No Kosmos (Multi-cloud native) No OpenShift/Rancher No No RWX Volumes EFS (Native shared FS) File Storage (Min 150 GiB box) Via Block Only Via Object Storage NFS Manual Manual NFS Manual NFS Specialized Compute Nodes & Acceleration GPU Node Pools Yes (p/g instances) Yes (H100, A100 tiers) Yes (H100, L40S arrays) Yes (NVIDIA nodes) Limited Via GPU Servers GPU Bare Metal only Key gap: Shared filesystem (RWX). AWS EFS makes ReadWriteMany volumes trivial — hundreds of pods mount the same filesystem. In EU clouds this requires either paying for oversized File Storage (OVHcloud minimum 150 GiB) or building NFS manually. For stateful microservices, media pipelines, and CI/CD shared caches, this is a recurring operational cost that falls on the customer. Section 4 Observability & Monitoring CloudWatch + X-Ray + OpenSearch Dashboards is one integrated bill, zero infrastructure ops, and available globally. EU clouds offer fragments. The full observability stack — metrics, logs, traces, alerting — must be assembled manually from open-source components and operated as a separate workload. AWS Reference Baseline CloudWatch + X-Ray + OpenSearch The native operational baseline for managing telemetry across multi-region enterprise serverless and container architectures. ⚡ Zero Infrastructure No nodes, storage arrays, or aggregators to manually scale. SLA: 99.9% Per Service ✓ CloudWatch Metrics: 1s granularity logs with 15-month storage. ✓ CloudWatch Logs: Scalable ingestion with Insights query engine. ✓ X-Ray Tracing: End-to-end trace collection and automated service mapping. ✓ OpenSearch Service: Fully managed native log index cluster pipelines. ✓ Application Signals: Standardized out-of-the-box APM SLO indicators. ✓ Container Insights: Automated structural infrastructure metric collection. ✓ Incident Pipelines: Native alarming bound directly to SNS and PagerDuty. European Sovereignty Observability & Telemetry Matrix 🇫🇷 OVHcloud Graylog / Grafana GA 60% CW Parity Logs Data Platform (LDP): Provides Graylog-based log aggregation supporting a 1–10 year retention lifecycle alongside an OpenSearch API layer for SIEM integration hooks. Integrates cleanly with platform Service Logs. • Metrics Deficit: Zero native managed metrics endpoints; DevOps teams must manually provision, configure, and maintain Prometheus layers on MKS. • Tracing/APM: No managed distributed tracing or automated application instrumentation blocks are available. • Managed Dashboards: Native Grafana engine is generally available (GA), supporting multi-source dashboards, alerting frameworks, and up to a 99.99% SLA tier on Enterprise plans. 🇫🇷 Scaleway Cockpit / Loki 50% CW Parity Cockpit Architecture: Delivers integrated, out-of-the-box observability with managed Grafana frontends, log collections backed by a Loki data pipeline, and Prometheus-compatible metrics processing. Includes a baseline free tier. • Distributed Tracing: No managed OpenTelemetry Tempo engine available yet. • Alerting Pipeline: Restricted to basic numeric threshold alarms; lacks sophisticated incident management or advanced call routing maps. • Topological Limits: Highly optimized for single-account Scaleway-native applications. Sophisticated hybrid scenarios or multi-account networks require custom manual deployments of Alloy or Grafana Agent infrastructure. 🇩🇪 IONOS Grafana v12.3 GA 45% CW Parity Logging & Monitoring Suite: Centralized log aggregation platform leveraging an updated Grafana v12.3 display engine (Deployed March 2026) across both EU and US computing regions. Companion Monitoring Service handles active infrastructure tracking and platform health diagnostics. • Telemetry Deficits: No distributed microservices tracing or integrated APM performance systems are exposed. • Network Transparency: Built-in support for Flow Logs to enable packet-level network traffic telemetry and inspection. 🇨🇭 Exoscale Managed OpenSearch Provides a native **Managed OpenSearch** engine tier optimized for full-text processing, log search index streams, and vector capabilities. Grafana target dashboards are documented. The Operational Trade-off: No managed Prometheus or managed Loki engines exist on the platform. Engineering pipelines must run log searches through OpenSearch while manually standing up Prometheus setups on SKS clusters for standard metrics. Result: Two isolated storage components without a uniform query engine layer. 🇫🇮 UpCloud No Native Stack No native telemetry, tracing, or log storage components are exposed directly inside the provider's native catalog. Standard Blueprint Pattern: Deploy Grafana via third-party Aiven infrastructure paths on UpCloud VM networks + attach targets to Managed Kubernetes nodes using custom Helm charts containing prometheus-community pipelines + kube-state-metrics + node_exporter arrays. Full DIY stack integration required. 🇩🇪 Hetzner 100% Operations Burden Zero managed platform telemetry endpoints or observability infrastructure SLAs are offered. SRE Implementation Cost Profile: DevOps engineers must spin up a dedicated telemetry VM (minimum specification requirement: 4 GB RAM / 50 GB SSD storage) running a self-managed layer of Prometheus + Grafana + Loki + Alertmanager orchestrations using docker-compose. Application nodes require manual setups of node_exporter and Promtail hooks. Estimated Operational Cost: €8–15/mo in base infrastructure overhead paired with total engineering liability. The observability gap is the most consistently painful. Every EU cloud customer without a dedicated SRE team ends up either (a) running AWS CloudWatch for their EU workloads — defeating the sovereignty argument — or (b) running an unmanaged Prometheus stack that becomes a maintenance liability. A managed, EU-sovereign Prometheus + Grafana + Loki + Tempo service is the highest-demand, lowest-supply managed offering in the EU cloud market today. Section 5 Serverless / FaaS AWS Lambda processes trillions of requests per month globally. In the EU cloud market, only Scaleway offers a comparable function-as-a-service. For most providers, "serverless" means Serverless Containers — still always-running compute, just containerized. ⚡ Serverless & FaaS Feature Parity Matrix Evaluating AWS Lambda and Fargate capabilities against European sovereign compute offerings Capability Vector 🇺🇸 AWS Target 🇫🇷 Scaleway 🇫🇷 OVHcloud 🇨🇭 Exoscale 🇩🇪 IONOS 🇩🇪 Hetzner 🇪🇺 UpCloud / evroc Managed Platform Runtimes True FaaS (Scale to Zero) ✓ Lambda ✓ Functions ✗ ✗ ✗ ✗ ✗ Serverless Containers ✓ Fargate / App Runner ✓ Serverless Containers Partial CaaS (Always-On) ✗ ✗ ✗ ✗ Operational Parameters & Triggers Event Source Triggers S3, SQS, DynamoDB, API Gateway, +200 native Object Storage, NATS, HTTP router N/A N/A N/A N/A N/A Max Execution Time 15 minutes 15 minutes (GB-s resource billing) — — — — — Free Tier Baseline 1M requests / mo always free Free tier tier-paths available — — — — — SI Integration Architectural Path Self-Hosted Blueprint — — Fn Project (on Bare-Metal/VM) OpenFaaS (on SKS Cluster) Knative / OpenFaaS Knative / OpenFaaS Knative / OpenFaaS Why this matters architecturally: Event-driven teams use Lambda for glue logic — S3 triggers, SQS consumers, API Gateway backends. On EU clouds without FaaS, this means running 24/7 containers (€20–80/month) to handle workloads that AWS Lambda serves for cents. The cost delta compounds with every microservice boundary. A managed Knative runtime on EU Kubernetes — deployed and operated by an SI — directly eliminates this cost gap. Section 6 Streaming & Messaging AWS offers three distinct messaging patterns: MSK (Kafka for streaming), SQS (queue), SNS (pub/sub fanout). EU clouds are converging on Kafka as the primary streaming primitive; simple queue and pub/sub semantics are unevenly covered. 🔄 Message Streaming & Event Bus Feature Matrix Analyzing messaging middleware, queuing architectures, and managed event brokers across provider catalogs Service Type 🇺🇸 AWS Target 🇫🇷 OVHcloud 🇫🇷 Scaleway 🇨🇭 Exoscale 🇩🇪 IONOS 🇩🇪 🇫🇮 Hetzner / UpCloud Managed Apache Kafka Ecosystem Managed Kafka ✓ MSK Fully managed, native scaling ✓ Yes + Kafka Connect & MirrorMaker ✓ Yes + Topics & Events ✓ Startup → Premium 3–30 node cluster ranges ✓ Managed Single-tenant architecture ✗ No native broker Kafka Serverless ✓ MSK Serverless On-demand storage & compute ✗ No ✗ No ✗ No ✗ No ✗ No Schema Registry ✓ Glue Schema Registry Native serialization governance ✗ No native ✗ No native Partial Via Aiven integration layer ✗ No ✗ No Point-to-Point Queuing & Pub/Sub Topologies Queue (SQS-Equivalent) ✓ SQS Near-unlimited structural scale ✗ No native queue ✓ NATS Queues ✗ No ✗ No ✗ No Pub/Sub (SNS-Equivalent) ✓ SNS High-throughput (15M+ msg/s) ✗ No ✓ NATS Topics ✗ No ✗ No ✗ No Dead-letter queues ✓ Native SQS DLQ Automated failure routing Kafka-level only Kafka-level only Kafka-level only Kafka-level only N/A Gap pattern: Kafka is well-covered across OVHcloud, Scaleway, Exoscale, and IONOS. The gap is in lightweight queuing — the SQS use case (task queues, job dispatch, async decoupling at moderate scale) has no native equivalent outside Scaleway's NATS. Teams building event-driven microservices on non-Scaleway EU clouds must self-host RabbitMQ or NATS — again, an operational burden that becomes an SI opportunity. Section 7 AI / ML Platform AWS Bedrock + SageMaker is a vertically integrated MLOps platform: experiment tracking, model registry, training pipelines, inference endpoints, and A/B deployment — all managed. EU clouds offer GPU compute and, in a few cases, inference APIs. The MLOps pipeline layer is absent everywhere. 🤖 Machine Learning & Sovereign AI Feature Matrix Mapping foundational models, MLOps orchestration pipelines, and data boundaries across sovereign cloud landscapes ML Capability 🇺🇸 AWS Target 🇫🇷 OVHcloud 🇫🇷 Scaleway 🇨🇭 Exoscale 🇩🇪 IONOS 🇪🇺 evroc Inference APIs & Foundational Models Managed LLM Inference ✓ Bedrock 30+ foundational models ✓ AI Deploy Custom models & weights ✓ Model-as-a-Service Managed Inference (OpenAI-spec) ✗ No native API Partial AI Model Hub (Berlin-only) Early Access evroc Think API Vector DB Managed ✓ Aurora pgvector + OpenSearch Vector Engine ✓ PostgreSQL pgvector ✓ pgvector + Serverless SQL targets ✓ pgvector + OpenSearch Enterprise ✓ pgvector On standard DBaaS ✗ No native DBaaS Lifecycle Orchestration & Frameworks Experiment Tracking ✓ SageMaker Experiments ✗ No ✗ No ✗ No ✗ No ✗ No Model Registry ✓ SageMaker Registry ✗ No ✗ No ✗ No ✗ No ✗ No Training Pipelines ✓ SageMaker Pipelines ✓ AI Training Automated GPU job runs ✓ AI Notebooks + Managed GPU clusters Partial Self-managed GPU instances ✗ No Early Access Sovereign GPU cluster A/B Model Deployment ✓ SageMaker Endpoints Native blue/green routing ✗ No ✗ No ✗ No ✗ No ✗ No Sovereign Jurisdiction & Compliance Data Boundary Boundary US CLOUD Act Extraterritorial warrants apply EU-Only Boundaries SecNumCloud Qualification EU-Only Sovereignty French data centers EU + Swiss Shield Strict federal laws Berlin Isolated German-owned stack Full EU Sovereignty Zero third-party transit The MLOps pipeline gap is total. Experiment tracking (MLflow), model registry, A/B deployment, and drift monitoring are absent as managed services across all EU clouds. Teams either self-host MLflow on K8s (engineering overhead) or remain on SageMaker/Vertex for their ML production systems while trying to move the rest of their stack to EU clouds — creating a persistent split architecture that never fully migrates. Section 8 Backup & Disaster Recovery AWS Backup provides a single control plane for coordinated backup of RDS, EBS, EFS, DynamoDB, and EC2 — with cross-region replication, compliance reports, and vault lock. EU clouds provide per-service backup (databases backup themselves, VMs have snapshots) with no coordination layer. 🛡️ Disaster Recovery & Business Continuity Feature Matrix Evaluating backup orchestration, Point-in-Time Recovery (PITR) parameters, and high-availability topologies DR Capability 🇺🇸 AWS Target 🇫🇷 OVHcloud 🇫🇷 Scaleway 🇨🇭 Exoscale 🇩🇪 IONOS 🇩🇪 🇫🇮 Hetzner Backup Orchestration & Replication Unified Backup Orchestration ✓ AWS Backup Cross-service centralized backup Per-service only Per-service only Per-service only Per-service only ✗ No native Cross-Region Replication ✓ Native Replication Built directly into AWS Backup Manual S3 sync automation required Manual Object Storage replication layer Manual User-managed scripts Manual User-orchestrated data transit DIY Setup Manual WAL-G shipping to S3 Database Point-in-Time Recovery (PITR) DB PITR Retention Window ✓ Up to 35 Days Native on RDS and Aurora Configurable Window Business & Enterprise plans 7 Days / None Classic: 7d | Serverless: No PITR PITR Supported Business & Premium tiers 7 Days Retention PITR enabled by default DIY Only Depends entirely on WAL-G config Governance & Immutable Storage Vault Lock / WORM ✓ AWS Backup Vault Lock Immutable write-once policies ✗ No ✗ No ✗ No ✗ No ✗ No Compliance Reports ✓ Backup Audit Manager Automated posture reporting ✗ No ✗ No ✗ No ✗ No ✗ No Active Availability & Multi-AZ Design Cross-AZ DB HA ✓ Multi-AZ Native Synchronous RDS / Aurora mirrors Plan-Dependent Included in Business / Enterprise HA Add-On Available (doubles baseline cost) Plan-Dependent Included in Business / Premium Managed Replication Async or strictly-sync setups DIY Clustering Self-managed Patroni architecture DORA Article 12 implication: Financial entities under DORA must document, test, and attest their ICT backup and recovery arrangements. "We back up our databases" is not sufficient — there must be a documented, tested, and auditable backup policy across the full stack. AWS Backup provides automated audit trails for this. On EU clouds without a coordinated backup service, this attestation burden falls entirely on the customer or their SI partner. This is a direct GART service opportunity: a Backup Orchestration & DR Attestation service sold to DORA-regulated entities. Section 9 Provider Deep-Dives What each provider actually ships, what it's missing, and where the SI surface area is largest. Sovereign Cloud Analysis 🇫🇷 OVHcloud Largest EU cloud provider by revenue. Deepest managed catalog. Industrial-scale bare metal paired with robust vRack private networking. 99.99% Multi-AZ SLA 🛡️ SecNumCloud ✓ 🔒 ISO 27001 ✓ 📊 SOC 2 Type 2 ✓ 🌐 vRack Isolation 📦 Genuinely Managed Runtimes ✓ PostgreSQL & MySQL: Scalable cross-AZ architecture (Essential/Business/Enterprise plans) featuring up to 99.99% SLAs for 3-AZ active clustering. ✓ Managed MongoDB: Native document engine lifecycle support, including an accessible free-tier Discovery plan path. ✓ Managed Valkey: Direct, platform-supported production environments with rigorous SLA bounds for the open Redis successor. ✓ Apache Kafka Suite: Fully managed streaming architecture bound natively alongside Kafka Connect and MirrorMaker utilities. ✓ Managed OpenSearch: Scalable structured text processing paired with direct, out-of-the-box Grafana dashboard layers. ✓ Managed ClickHouse: Managed columns analytics execution (note: currently bound to limited regional availability). ✓ Managed Kubernetes (MKS): Free native control plane running high-performance Cilium CNI mesh infrastructure. ✓ Managed Grafana: Unified platform telemetry frontends providing seamless multi-source log, metric, and alerting charts. ✓ AI Deploy Stack: Turnkey ML model serving pipelines backed by raw enterprise-grade NVIDIA H100, A100, and L40S GPU bare metal. ✓ Logs Data Platform: Long-term data aggregation layer leveraging combined Graylog + OpenSearch engines with a 10-year retention footprint. ⚙️ Architectural Reality Check ⚠️ Q2 2026 Platform Renovations • Billing Reform (May 2026): Line items cleanly isolated for raw compute, local disk, and independent backup storage metrics mapped down to exact per-second billing fields. • Load Balancer Migrations (May 2026): Legacy IOLB layers deprecated completely. Engineering teams must map manual config migrations over to the modern load-balancer stack. • Keystone OpenStack Auth Patch (May 2026): Structural platform security changes temporarily disabled cross-project Trusts and Application Credentials. 🌐 Network Isolation (vRack): High-performance physical network backplane that explicitly maps DB and K8s internal nodes away from public internet exposure. ❌ Essential Tier Restrictions: Enforces a zero-SLA, non-HA architecture containing no private network integration targets. Highly restricted to sandboxed development pipelines only. 📦 File Storage Over-Provisioning: Persistent RWX storage volumes inside MKS demand a strict minimum initial order requirement of 150 GiB, generating artificial infrastructure bloat for smaller stateless services. Support Tier Operational Overhead TierCost Threshold StandardIncluded Free Premium$56 / mo fixed Business10% of bill (min $300) Enterprise30% of bill (min $5,850) Key gaps vs. AWS No serverless functions (FaaS). No unified backup orchestration. No distributed tracing (X-Ray equivalent). ClickHouse available only in select regions. RWX volumes require minimum 150 GiB File Storage purchase. No Aurora-equivalent (no distributed storage layer). Support enterprise tier starts at $5,850/month minimum. Developer-First Architecture 🇫🇷 Scaleway Closest alignment to the AWS philosophy. Features a highly unique combination of scale-to-zero serverless runtimes, native Apple Silicon bare metal, and multi-cloud Kubernetes federation engines. Scale-to-Zero Serverless SQL ✓ 📍 Milan Region (MIL-1 GA March 2026) 🍏 Apple M4 Pro in Cloud ⚡ Cockpit Grafana Stack 🚀 Differentiating Services Serverless SQL: On-demand PostgreSQL auto-scaling from 0 to 15 vCPUs. Features automated idle-to-zero suspension after 5 minutes of inactivity. Shipped natively with pgvector, PostGIS, and timescaledb extensions. Serverless Functions: True execution-isolated FaaS pipelines with event-driven triggers, supporting a 15-minute maximum runtime bound and precise GB-second billing curves. Serverless Containers: Cold-start-capable isolated runtime environments engineered for synchronous HTTP request-driven triggers. Kubernetes Ecosystem: Dual delivery paths via Kapsule (optimized single-cloud cluster structures) and Kosmos (advanced multi-cloud Kubernetes node pool federation). NATS Messaging Broker: Unified broker tier combining point-to-point queues and publish/subscribe topics. Acts as an efficient SQS and SNS equivalent within a single engine. Model-as-a-Service (MaaS): High-throughput hosted LLM inference endpoints optimized via clusters of NVIDIA H100 and L40S accelerator setups. Elastic Bare Metal: Compute flexibility stretching from standard Elastic Metal Aluminium/Titanium bare shapes to specialized bare-metal deployments of Apple Mac mini hardware (M1/M2/M4/M4 Pro). 📊 Technical Guardrails & Constraints 🛡️ Proactive Kernel Hardening Critical container security patching verified during the recent kernel vulnerability lifecycle (algif_aead, Apr–May 2026), which saw running Kapsule nodes securely patched within hours of zero-day disclosure. Serverless SQL Hard Limits Compute AllocationMax 15 vCPU / 60 GB RAM Storage Volume1 TB Block Limit Concurrency Bound1,000 Active Connections Query Payload Size1,024 KB Maximum 💡 Idle Mechanics: 5-minute timeout automatically cuts compute billing lines completely. Underlying block storage billing remains persistent. 📦 Classic Managed DB Profiles: Standard multi-tenant configurations benchmark at €80/mo for the base PRO2-XXS instance tier (2 vCPU, 8 GB RAM, 20 GB Storage) with a maximum 7-day Point-in-Time Recovery (PITR) ledger window. ⚠️ High Availability Overhead: Enabling active Multi-AZ replication doubles the monthly cost across the entire Classic Database tier. 🌐 Egress Surcharges: Data transfer is entirely free between native Scaleway components residing inside the same compute region. Public internet transit above baseline tier quotas incurs an additional surcharge of €0.01/GB. Key gaps vs. AWS Cockpit observability lacks distributed tracing (no Tempo equivalent). No managed Prometheus as a standalone service. No Aurora-equivalent (Serverless SQL is ~30% feature parity). PITR not available on Serverless SQL tier. No managed schema registry. No FaaS-native DLQ. MLOps pipeline absent — GPU + inference only. Infrastructure Sovereign Strategy 🇩🇪 Hetzner Cloud Recognized for the industry's premier price-to-performance ratio in the EU. Hetzner maintains an intentionally minimal PaaS footprint, providing maximum SI (Systems Integration) opportunity for customized infrastructure stacks. IaaS-Centric Minimal PaaS ✓ 🛡️ BSI C5 Type 2 (Verified March 2026) 🌱 EMAS Certified (Nov 2025) 🧩 Core Infrastructure Catalog Cloud Instances: Flexible CCX (Compute), CX (General), and CPX (Performance) lines delivering aggressive price-per-vCPU value. S3-Compatible Object Storage: Reliable, high-performance distributed blob storage for static media and backup archival. Load Balancers: L4 and L7 support; note the performance throughput ceiling (no auto-scaling utility). Networking & Storage: Floating IPs, Private VPC-style networking, hardware firewalls, and block volumes (ext4/xfs/zfs support). Bare Metal: Dedicated AX/SX/PX server lines accessible via deep-integration Robot API for enterprise-grade performance. Storage Box: Dedicated NFS/Samba targets optimized for long-term backup retention and off-instance staging. ⚙️ Integration & Modernization ⚠️ June 2026 Hardware Restructuring New dedicated server lineup adopts a "-1, -2, -3" suffix model. Includes "-1-Ltd" budget tier (refurbished hardware components). Note: Existing contract rates are locked, but any modification to server configuration triggers an immediate re-billing to the new market pricing structure. Third-Party Orchestration Layers • Managed Kubernetes: Supported via 3rd party ecosystems including CFKE (managed 99.95% SLA, free under 24 vCPUs), Syself Autopilot (GitOps lifecycle), and standard k3s community utility sets. • Managed Postgres: Deployed via Autobase (Patroni-backed HA cluster + pgBackRest on CCX23/33 nodes), managed via automated Ansible playbooks. Infrastructure Consistency Cloud Instances, Object Storage, Load Balancers, and Storage Box product lines remain structurally unchanged by the June 2026 hardware updates. Native gaps (by design) No managed databases. No managed Kubernetes. No observability. No FaaS. No messaging queues. No AI/ML. No backup orchestration. Hetzner has explicitly chosen not to build DBaaS — development of a resilient control plane would require large investment and impose AGPL open-source licensing risk for integrated tooling. Sovereignty & High-Velocity Compute 🇨🇭 Exoscale API-first infrastructure built on Swiss sovereignty standards. Combines a highly optimized, Aiven-powered DBaaS engine with the fastest Kubernetes cluster deployment velocity in the European market. < 2 Min SKS Cluster Provisioning 🛡️ Swiss Jurisdiction Sovereignty 📊 99.99% SLA (Business/Premium) 🤖 OpenAI-Compatible Inference ⚙️ Managed Infrastructure Highlights Scalable Kubernetes Service (SKS): Rapid cluster deploy featuring native Karpenter auto-scaling integration, isolated User Namespaces, Gateway API controls, and fine-grained Egress Filtering meshes. Aiven-Powered DBaaS: Production-grade hosting for PostgreSQL, MySQL, Valkey, OpenSearch, and Apache Kafka engines across a 6 EU + CH regional zone footprint. Performance DB Tweaks: Pre-configured, out-of-the-box pgBouncer connection pooling across all application plans to stabilize throughput. Vector Search for RAG: Native vector processing architectures leveraging coupled pgvector extensions and dedicated OpenSearch Vector Search nodes. Managed GPU Inference: OpenAI-compatible API layers serving open foundational LLMs directly from sovereign European GPU bare metal pools. 📊 Database Tiers & Guardrails ⚠️ Technical Administration Limits • Advanced Parameters: Certain deep cluster configurations—such as modifying specific Point-in-Time Recovery (PITR) ledger windows—are omitted from the web console and require direct CLI or API execution. • Database Forking: Live PITR-cloned database forks are restricted exclusively to the Business and Premium plans. PostgreSQL Service Tiers TierTopology / SLABackups Hobbyist1 Node / 99.95%2-Day Max Startup1 Node / 99.95%Limited Window BusinessActive Standby / 99.99%Native PITR ✓ PremiumUp to 9x Nodes / 99.99%Native PITR ✓ 💡 Storage Mechanics: Backups are completely encrypted at rest and piped to an independent Object Storage layer via plan-dependent schedules. Kafka horizontal scaling scales up to 30 distributed nodes on Premium tiers. Key gaps vs. AWS No FaaS (self-hosted Fn Project on VM is the documented workaround). No managed MongoDB. No schema registry for Kafka. No managed observability stack (Prometheus/Loki/Tempo). No MLOps pipeline — inference only. Smaller regional footprint vs. OVHcloud. Swiss jurisdiction useful for CH compliance, less relevant for GDPR-pure EU mandates. Enterprise & Public Sector Sovereign Cloud 🇩🇪 IONOS Cloud Engineered specifically for the strict compliance frameworks of enterprise and public sectors. Features unique visual topology designer interfaces alongside the deepest, most granular Identity and Access Management models in the EU market. DCD Canvas Visual Topology ✓ 🛡️ BSI C5 Type 1 Verified 📦 Red Hat OpenShift Supported ⚓ SUSE Rancher Native ⚙️ Managed Service Catalog Enterprise DBaaS Engine: Turnkey lifecycle deployments for PostgreSQL (versions 14–16), MariaDB, and MySQL-compatible applications. Managed MongoDB: Playground, Business, and Enterprise plans supporting v6.0 and v7.0 with automated patching and 7-day retention PITR backup streams. Kubernetes Power Trio: Native, managed provisioning matching upstream standard Kubernetes, Red Hat OpenShift, and SUSE Rancher platforms under one cloud framework. Private Container Registry: Secured Frankfurt-hosted registry featuring automated vulnerability scanning and time-bound temporary authentication tokens. Streaming, Logging & Monitoring: Single-tenant managed Apache Kafka clusters backed by integrated, centralized Logging (Grafana v12.3) and real-time health Monitoring metrics. 🔒 Enterprise Topologies & Security 🤖 Berlin-Isolated AI Model Hub Provides open-source foundation model inference hosted completely within isolated Berlin data center zones. Zero data leaks, zero cross-border token transit paths, and total compliance safety. Hierarchical Access Control (IAM) Structured 3-Tier Model: Contract Owner (Billing) ➔ Administrator (Technical Operations) ➔ User Groups (Delegated Project Access). 💡 Granular Security: Per-resource permission assignments mapped down to isolated Virtual Data Centers (VDCs), cluster shapes, volume snapshots, image repositories, and individual IP pools. 🌐 Data Center Designer (DCD): A fully interactive visual canvas providing real-time drag-and-drop provisioning for complex layouts of VMs, storage blocks, load balancers, and LAN segments. ⛓️ Cross Connect & VPN: Establish seamless, private point-to-point tunnels across separate VDCs via single internal LAN mappings, combined with WireGuard or IPSec VPN Gateways. ⚡ Network & Edge Edge Layers: Layer 7 Managed Application Load Balancers matched with deep Network Flow Logs and global edge CDNs wrapped in integrated WAF + Layer 7 DDoS safeguards (GA Jan 2026). Key gaps vs. AWS No FaaS. No serverless database. No managed observability traces (APM absent). Logging + Monitoring cover ~40% of CloudWatch depth. No unified backup orchestration. No managed schema registry. AI Model Hub limited to Berlin region and open models only — no managed training pipeline. Read replicas not available on standard DBaaS tier. Mid-Market Infrastructure Excellence 🇫🇮 UpCloud High-performance European cloud architecture striking the optimal balance between cost transparency and enterprise reliability. Positioned strategically above bare-metal providers with a heavily managed production catalog. 100% SLA Uptime Backed ✓ 🛡️ ISO 27001 Certified 🦫 Aiven DBaaS Partnership ☸️ CNCF-Certified Kubernetes ⚙️ Managed Platform Ecosystem Managed Kubernetes: Fully automated, CNCF-certified engine with hands-off cluster lifecycle management and intelligent auto-scaling worker node groups. Aiven-Powered DBaaS: Premium managed database abstractions for PostgreSQL, MySQL, Redis, and OpenSearch—leveraging a production-proven enterprise backend layer. Object Storage & Data: Highly available, fully compliant S3-compatible object storage zones optimized for persistent backup retention and static assets. Software-Defined Networking: Instant provisioning of native Load Balancers, highly flexible Floating IPs, isolated Private Networks, and granular SDN Firewalls. 🚀 Market Positioning & SLAs ⚡ Hyper-Responsive Engineering Support Boasts an industry-leading average support response window of just 46 seconds, keeping infrastructure roadblocks to an absolute minimum. Strategic Mid-Market Positioning Offers a richer ecosystem of fully managed platform services than alternative providers like Hetzner, coupled with significantly more predictable, transparent flat pricing structures than OVHcloud. 🌍 Global Backbone Footprint: 15 enterprise-grade data centers spanning 4 continents, featuring a highly dense European presence (Frankfurt, London, Amsterdam, Helsinki, Warsaw, Stockholm). 💰 Financial Uptime Guarantee: The uncompromising 100% uptime SLA is actively backed by clear, transparent financial credits written directly into the service terms. 🧾 Zero Hidden Pricing Fees: Flat billing models ensure data egress, bandwidth allocations, and storage overhead remain fully predictable month over month. Key gaps vs. AWS No FaaS. No native managed observability — documented pattern is self-assembling Prometheus + Grafana via Aiven on K8s. No MongoDB managed. No Kafka managed. No AI/ML services. No backup orchestration. No CDN. Catalog is functional but narrow: K8s + DBaaS + Object Storage is roughly the full offering. Sovereign Hyperscale & AI-First Platform 🇸🇪 evroc Europe’s emerging sovereign hyperscaler. Engineered with full control-plane autonomy and zero reliance on non-EU software stacks, evroc is purpose-built for highly regulated workloads, defense-grade compliance, and dense AI processing. €50.6M Series A (EQT) ✓ 🇪🇺 100% EU Sovereignty Guaranteed 🦎 SUSE Rancher Certified (Q1 2026) 🛡️ Defense & Gov Grade 💻 Live Capabilities (As of June 2026) Core Infrastructure Stack: Production-ready, native IaaS layer delivering enterprise-grade compute, isolated object storage, and secure virtual networking. evroc Think Platform: Early-access sovereign AI infrastructure engineered exclusively for advanced model training, fine-tuning, and LLM deployments. evroc roc Agent: Newly announced (2026) enterprise-level AI agent runtime, empowering organizations to deploy context-aware AI safely inside European boundaries. Certified Kubernetes: Native container management powered directly by certified SUSE Linux Enterprise and SUSE Rancher Prime environments. 🌍 Topology & Ecosystem Context ⚡ Next-Gen Flagship Data Centers Flagship sites in Stockholm and Mougins (France) completing in 2026. Designed for **10,000+ GPUs**, featuring advanced direct liquid cooling and massive **150 kW per rack** density. Maximum SI Whitespace The Detecon + evroc partnership serves as the market reference model. Offers greenfield integration options for System Integrators looking to own the managed service layer. 📡 Current European Footprint: Operating out of active strategic partner data centers across Stockholm (2), Paris (2), and Frankfurt (2). 🎯 Target Sectors: Laser-focused on heavily regulated industries: Defense, Government, Healthcare, Financial Services, and Critical Infrastructure. 📈 2030 Aggressive Growth Map: Actively scaling toward a planned network of 10 fully proprietary, hyperscale sovereign data centers across Europe by 2030. Gaps (almost everything in PaaS) No managed databases. No managed Kubernetes (SUSE Rancher available but not managed-as-a-service). No observability stack. No FaaS. No messaging/queuing. No backup orchestration. evroc Think is the only managed service beyond IaaS — and it's early access. The entire managed layer is a gap. Section 10 The Sovereign Stack Opportunity The managed services gap is not closing from below — cloud providers are not investing fast enough in PaaS depth. It closes from above: SIs and MSPs who build the operating layer and deliver it as a service on top of EU IaaS. This is the GART position. Platform Ops & Telemetry Managed Observability Platform Deploy and operate a fully managed Prometheus, Grafana, Loki, and Tempo stack on sovereign EU IaaS. Engineered as a highly secure, multi-tenant service with per-environment SLA tiers. €800 – €2,500 Per Env / Month (Est.) ⚙️ Capabilities & Architecture • Full-Stack Telemetry: Seamless collection of metrics, logs, and distributed tracing. • Native Cloud Seeding: Deeply integrates with IONOS Logging Service and OVHcloud LDP out of the box. • SLA Tiers: Custom boundaries across Dev, Production, and mission-critical Enterprise environments. 🎯 Strategic Fit & Monetization • Infrastructure Base: Ideal for Hetzner (raw price/performance optimization) or OVHcloud setups. • The Gap Filled: Resolves the lack of native managed tracing services across almost all EU clouds. • High Stickiness: Embedded in day-to-day enterprise developer and triage workflows. Sovereign Database Operations Managed PostgreSQL Service Production-grade, fully operated Patroni-based High Availability (HA) clusters. Includes professional day-2 database administration, automated failover, and bulletproof disaster recovery. €600 – €3,000 Per Cluster / Month (Est.) ⚙️ Capabilities & Architecture • Enterprise HA: Patroni-orchestrated clusters with PgBouncer connection pooling built-in. • Advanced Backup: pgBackRest execution offering native Point-in-Time Recovery (PITR). • Cross-Region Safeties: Automated, encrypted backup pipelines pushing directly to S3-compatible storage. 🎯 Strategic Fit & Monetization • The Gap Filled: Directly closes the "RDS Gap" for 80% of legacy workloads migrating off AWS/Azure. • Target Base: High-growth Hetzner, UpCloud, and emerging evroc enterprise customers. • Retention Lock: Unrivaled customer retention—no organization wants to maintain raw Patroni topology manually. Serverless Modernization FaaS Runtime on EU Kubernetes An abstraction layer delivering cloud-native, AWS Lambda-compatible function deployments natively across European managed Kubernetes clusters. €300 – €1,500+ Base + Usage / Month ⚙️ Capabilities & Architecture • Engine Mechanics: Built on proven Knative or OpenFaaS runtime models. • Event Mesh Connectors: Out-of-the-box streaming hooks for S3-compatible Object Storage, NATS, and Apache Kafka. • Billing Pass-Through: Clean utility monetization mapped cleanly to per-invocation metrics. 🎯 Strategic Fit & Monetization • Target Base: Runs cleanly on Exoscale SKS, OVHcloud MKS, or IONOS Kubernetes instances. • The Gap Filled: Completely breaks dependency on AWS Lambda, removing the main technical roadblock keeping teams tethered to a hybrid US/EU cloud architecture. Mission-Critical Risk Assurance Sovereign DR Orchestration (DORA Compliant) Turnkey Business Continuity and Disaster Recovery orchestration engineered directly to satisfy rigorous audit demands under the EU Digital Operational Resilience Act (DORA). €2,000 – €8,000 Per Entity / Month (Est.) ⚙️ Capabilities & Architecture • Multi-Zone Sync: Cross-service infrastructure backup policies coupled with automated cross-region object storage replication topology. • Continuous Attestation: Rigorous RTO/RPO mapping logs backed by quarterly live-fire DR drills producing signed, audit-ready compliance attestations. 🎯 Strategic Fit & Monetization • Target Base: Regulated FinTech, banking entities, insurance firms, and payment processors. • The Gap Filled: No single EU cloud native engine ships with cross-provider or cross-region automated DR compliance logging. Guarantees near-zero churn parameters. Sovereign Intelligence Systems Sovereign EU MLOps Platform Enterprise MLflow deployment integrated onto localized GPU bare metal, paired with a sovereign model registry, automated A/B traffic splits, and an OpenAI-compliant API interface. €1,500 – €5,000 Base + Compute / Month ⚙️ Capabilities & Architecture • Sovereign Inference: Maps execution seamlessly into Exoscale GPU instances, IONOS AI Model Hub structures, or Scaleway MaaS backends. • Drop-In Gateway: Fully OpenAI-compatible schema interface for rapid code migration. 🎯 Strategic Fit & Monetization • Target Base: Active production AI teams forced out of AWS/Azure due to runaway inference budgets, data transit costs, or strict US CLOUD Act exposure risks. • The Gap Filled: Acts as a secure drop-in alternative replacing Azure OpenAI or AWS SageMaker architectures. Continuous Governance & IaC Compliance-as-Code (IaC) Subscription A premium subscription offering continuously updated, hard-engineered Terraform blueprints and Open Policy Agent (OPA) bundles mapped strictly to evolving European regulatory baselines. €3,000 – €12,000 Per Year / Subscription ⚙️ Capabilities & Architecture • Pre-Baked Modules: Securely isolated network patterns, CI/CD gates, and private subnets mapped to NIS2, DORA, and BSI C5 frameworks. • Guardrail Bundles: Native OPA/Gatekeeper policy assertions to block non-compliant deployments in CI pipelines. 🎯 Strategic Fit & Monetization • Cloud Ecosystems: Maintained across OVHcloud, Hetzner, IONOS, and Scaleway topologies. • High Switching Barriers: High-margin annual recurring product; since the customer's entire core IaC depends on these pipelines, switching costs remain exceptionally high. SI/MSP Economics Three revenue streams compound: managed service fees from customers, cloud consumption commissions from providers, and professional services for initial migration engagements. Managed service margin: 60–75% gross margin on operated services (platform costs are mostly engineering time, not cloud spend) Provider commission: 10–20% of attributed customer cloud consumption — EU clouds actively pay SIs to drive revenue they can't capture alone Migration PS: €50K–€300K per enterprise migration engagement — one-time but creates the managed service relationship Regulatory deadlines create non-discretionary spend: DORA Article 17 (Jan 2025), NIS2 enforcement active — customers must act Why First-Mover Wins EU cloud providers need enterprise customers. Enterprise customers need managed services. The SI who builds the operating layer first captures the account — not the provider. The provider's growth depends on the SI. EU cloud providers lack enterprise sales motion — they need SI partners to reach regulated buyers who will not do self-service procurement AWS CLOUD Act problem is structural — no path to SecNumCloud, no exemption possible. The ceiling is permanent evroc, UpCloud, and emerging providers need SI ecosystems to reach production scale — early partners get preferred pricing and co-sell rights Managed layer IP is portable: GART's Patroni runbooks, Knative config, and compliance modules work across any IaaS — not locked to one provider Sovereign Managed Infrastructure Gart Solutions closes the gap. On sovereign ground. We build and operate the managed services layer that EU cloud providers haven't shipped yet — so your workloads move to sovereign infrastructure without giving up the capabilities your teams depend on. Talk to Gart Solutions