--- title: "Launch & Operations – Sovereignty That Pays Off" description: "Step L in the ANVIL system: go-live on EU-sovereign infrastructure (Hetzner, Germany), then monitoring, incident response and scaling." lastUpdated: "2026-06-27" canonical: "https://anvilstack.eu/en/services/platform-development" --- > Sitemap for agents: https://anvilstack.eu/llms.txt # Launch & Operations – Sovereignty That Pays Off A product only creates value when it runs in production – under real traffic, under real law, in the hands of whoever operates it. The path from a signed-off staging environment into sovereign ongoing operations decides whether a finished system becomes a viable business. Take that path on a US hyperscaler and you trade speed for dependency: unpredictable egress costs, foreign law over your own data, a latent GDPR risk. **Launch & Operations, Step L in the ANVIL system,** instead takes a system live on EU-sovereign infrastructure and keeps it stable there – turning data sovereignty into an argument you can pass on to your own customers. - **2.1x** – cheaper: Hetzner dedicated (CCX33) vs. AWS On-Demand at the same configuration - **0** – CLOUD Act exposure on EU-sovereign infrastructure (BMI report) - **99.9%** – availability as an error-budget target for production-critical systems (Google SRE) - **27%** – of IaaS and PaaS cloud spend is considered wasted (Flexera 2024) > **Info:** **Sovereignty is not a compliance checkbox – it's a property of the product.** A platform that demonstrably runs under EU law in a German data center isn't merely compliant – it's a promise you can give your own customers. And because the same team that built the system also operates it, there's no break and no loss of knowledge between build and operation. ## No sovereign operations, no real scaling An MVP that only runs in a staging environment creates no value. And a production system that sits on US hyperscalers under foreign law creates dependency instead of scale. Two burdens accumulate out of sight: cost, because egress and oversized managed services only surface at the end of the month, and law, because a foreign jurisdiction holds sway over every personal record. **No sovereign operations → dependency instead of scale.** Step L removes both entirely: the launch runs on EU-sovereign infrastructure, and operations continue with monitoring, incident response and scaling. **The result:** the platform runs live, sovereign and under EU law. Your data stays in Europe, your costs stay predictable. Step L is the final step of the ANVIL system. Every path – whichever entry point it begins on – ends here: in a production-ready go-live and stable operations. ## Go-live on EU-sovereign infrastructure The launch runs on Hetzner in a German data center – reproducible through infrastructure-as-code and with zero downtime: - **Compute:** Containerized application on Hetzner (German data center), as a K3s cluster with health checks and rolling deployments where needed - **Database:** Self-hosted PostgreSQL on Hetzner with automated backups - **Infrastructure-as-code:** Terraform for versioned, reproducible infrastructure instead of manually assembled servers - **Zero-downtime launch:** Blue-green deployment, with staging sign-off before the DNS cutover - **GDPR-compliant configuration:** Data processing exclusively in the EU, encryption at rest and in transit, logging without PII Provisioning itself is AI-positive: automation and AI-assisted tooling produce reproducible Terraform modules in hours rather than days – while sign-off, hardening and the operational judgment come from a senior team. Speed in the build, experience at the point where a mistake would get expensive. US government access to data held by US cloud providers "cannot be reliably ruled out," according to an internal report by the German Federal Ministry of the Interior. ([Igor's Lab: BMI report on US data access (2025)](https://www.igorslab.de/en/us-laws-allow-access-to-european-cloud-data-internal-report-by-interior-ministry-confirms-risks-to-digital-sovereignty/)) On EU-sovereign infrastructure, this exposure disappears entirely. If you're coming from a US hyperscaler, the migration to EU-sovereign infrastructure is part of Step L. Commissioning typically takes two to five days; operations then run continuously. ## The real payoff of Step L is commercial, not technical. Three concrete levers turn sovereignty into a sales instrument: - **Guarantee instead of disclaimer:** You can contractually guarantee that customer data never leaves Europe – no third-country transfer, no CLOUD Act exposure. ([Igor's Lab: BMI report on US data access (2025)](https://www.igorslab.de/en/us-laws-allow-access-to-european-cloud-data-internal-report-by-interior-ministry-confirms-risks-to-digital-sovereignty/)) - **NIS2 supply chains:** Since December 2025, NIS2 has required roughly 29,500 German companies to maintain documented cybersecurity measures – including those of their service providers. A sovereign platform makes you a low-friction supplier. ([BSI: NIS2 Implementation in Germany](https://www.bsi.bund.de/DE/Themen/Regulierte-Wirtschaft/NIS-2-regulierte-Unternehmen/nis-2-regulierte-unternehmen_node.html)) - **Planning certainty against political risk:** The statutory basis for FISA Section 702 lapsed on 12 June 2026 and reauthorisation is still open. Should that lead to a Schrems III decision, competitors on US infrastructure would be under immediate pressure to act – a sovereignly operated system would not. ([Brennan Center: Section 702 FISA 2026 Resource Page](https://www.brennancenter.org/our-work/research-reports/section-702-foreign-intelligence-surveillance-act-fisa-2026-resource-page)) For more background, see our pages on the [CLOUD Act data risk](https://anvilstack.eu/en/compliance/cloud-act-data-risk.md) and the [GDPR vendor audit](https://anvilstack.eu/en/compliance/gdpr-vendor-audit.md). ## Predictable costs: Hetzner vs. AWS Sovereignty also lowers your operating costs and makes them predictable – with no egress surprises at the end of the month. The direct comparison of comparable configurations: | Criterion | Hetzner (DE) | AWS Frankfurt | | ------------------------------ | -------------------------- | ----------------------------------- | | 8 vCPU / 16 GB RAM | \~€16/month (CX43, shared) | \~€148/month (m6i.xlarge On-Demand) | | 8 vCPU / 32 GB RAM (dedicated) | \~€138/month (CCX33) | \~€296/month (m6i.2xlarge) | | 1 TB block storage | \~€57/month | \~€84/month (gp3) | | Traffic (20 TB/month) | Included | \~€1,584 (egress costs) | | Self-hosted PostgreSQL (16 GB) | \~€86/month (CCX23) | From \~€243/month (RDS managed) | | CLOUD Act exposure | None | Full | [Hetzner: Cloud pricing (as of June 2026)](https://www.hetzner.com/cloud/) [AWS: EC2 On-Demand Pricing (as of 2026)](https://aws.amazon.com/ec2/pricing/on-demand/) [AWS: RDS for PostgreSQL Pricing](https://aws.amazon.com/rds/postgresql/pricing/) The largest line in this table isn't the server, it's the traffic – and that's exactly where the surprises live. According to Flexera, organizations estimate that an average of 27% of their IaaS and PaaS cloud spend is wasted, for example on idle or oversized resources. ([Flexera: 2024 State of the Cloud Report](https://resources.flexera.com/web/pdf/Flexera-State-of-the-Cloud-Report-2024.pdf)) An infrastructure-as-code approach on Hetzner meets that waste with precise capacity planning and a transparent cost structure you can lay out for investors and customers. See the details in our [Hetzner vs. AWS comparison](https://anvilstack.eu/en/compare/hetzner-vs-aws.md). ## Ongoing operations: monitoring, incident response, scaling The go-live leads into an ongoing operations partnership – no handover vacuum, but a stably operated system: - **Monitoring** with OpenTelemetry, Grafana and Prometheus: health checks, distributed tracing and automated alerts on anomalies - **Incident response** with defined escalation tiers and response times - **Regular security updates** and dependency patches – the hardening from [Step I](https://anvilstack.eu/en/services/mvp-engineering.md) is preserved in operation - **Scaling and capacity planning** based on real usage data, handled by the same senior team that built the system Following Google's SRE principles, an error budget of 0.1% applies – 99.9% availability for production-critical systems. ([Google: SRE Book – Embracing Risk](https://sre.google/sre-book/embracing-risk/)) That's not a marketing promise but an operating discipline: it defines how much downtime is permissible over a period, making availability measurable rather than merely asserted. > **Warning:** **Build vs. buy:** An in-house platform team (2 senior engineers + DevOps + infrastructure) costs from €400,000/year in the DACH region (our buildup): market salaries ([gehalt.de: DevOps engineer salary, Germany](https://www.gehalt.de/beruf/devops-engineer)) plus employer contributions (\~23%), tooling and infrastructure. That fixed cost is carried by any company building operations in-house – whether the system is under load or not. An external senior team avoids that fixed staffing block and provides the same discipline as an operations partnership. ## In operation, the MVP grows into a platform The terms follow a clear ladder: the prototype is the starting point, the MVP from [Step I](https://anvilstack.eu/en/services/mvp-engineering.md) is the production-ready first version – and in live operation, that MVP grows into a platform. The system is designed from the outset to grow with you: from the first real customers to an established platform. A full-scale expansion to enterprise architecture – horizontal scaling with K3s, read replicas, a dedicated team – is **explicitly not a fixed "Phase 6"**. It's an optional next step decided from within operations, based on real load data, if and when your growth justifies it. Which expansion stage makes sense, and when, is shown not by gut feel but by actual user behavior under load. ## Who is Step L for? Every project begins with the [analysis (Step A)](https://anvilstack.eu/en/services/prototype-audit.md) as the first step. Depending on what already exists, clients come in at different points – but all four entry paths ultimately run through Step L: - **Idea** → entry at A, then the full path through to a sovereign launch - **Concept/design** → entry at validation - **Prototype** → entry at [implementation & hardening](https://anvilstack.eu/en/services/mvp-engineering.md), the typical core case - **Platform (broken or insecure)** → direct entry at I/L: this is where Step L applies most directly, often as a migration to EU-sovereign infrastructure followed by operations In every case the result is the same: a live, sovereign platform under EU law. Want to take your system live sovereignly and know it's operated stably? Send us a short project description – we'll get back to you within 24 hours. The intro call is no-obligation. [Book a no-obligation intro conversation](https://anvilstack.eu/en/contact) ## Frequently asked questions ### How does a zero-downtime go-live work? Through a blue-green deployment: the new version runs fully provisioned alongside the old one, is signed off in a production-like staging environment, and only then does a controlled DNS cutover point traffic at the EU-sovereign platform. If something goes wrong, the way back is immediately available. Users notice nothing of the switch. ### How much cheaper is Hetzner compared to AWS? At a comparable configuration, Hetzner is roughly 2.1x cheaper than AWS On-Demand. A dedicated server with 8 vCPU / 32 GB RAM costs \~€138/month on Hetzner vs. \~€296/month on AWS. On top of that comes traffic: 20 TB/month is included with Hetzner, while on AWS it costs \~€1,584 as egress. The decisive difference, though, is predictability, not just the sum. ### Do I need EU infrastructure for GDPR compliance? If personal data is processed on US cloud infrastructure (AWS, Azure, GCP), there is a GDPR risk regardless of where the servers sit. The CLOUD Act obliges US providers to hand over data to US authorities. Operating on EU-sovereign infrastructure in a German data center removes that legal risk at the root. ### What does ongoing operations actually mean? Monitoring with automated alerts, incident response with defined response times, regular security updates, architecture reviews and capacity planning based on real usage data. Following Google's SRE principles, an error budget of 0.1% applies – 99.9% availability for production-critical systems. Not a support contract with a block of hours, but a stably operated system with clear SLAs. ### Does an MVP grow into a full-fledged platform in operation? Yes. The MVP is built to grow with you – from the first real customers to an established platform. The further full-scale expansion to enterprise architecture (horizontal scaling, a dedicated team) is not a fixed 'Phase 6,' but an optional next step decided from within ongoing operations, based on real load data – when growth justifies it. ### Is sovereignty really a selling point? Yes. A platform that demonstrably runs under EU law in a German data center is something you can pass on to your own customers as a guarantee – especially in regulated industries, the public sector and NIS2-obligated supply chains. Data sovereignty thus turns from a compliance checkbox into a measurable competitive advantage. ## Sources - [Hetzner: Cloud pricing (as of June 2026)](https://www.hetzner.com/cloud/) - [AWS: EC2 On-Demand Pricing (as of 2026)](https://aws.amazon.com/ec2/pricing/on-demand/) - [AWS: RDS for PostgreSQL Pricing](https://aws.amazon.com/rds/postgresql/pricing/) - [Igor's Lab: BMI report on US data access (2025)](https://www.igorslab.de/en/us-laws-allow-access-to-european-cloud-data-internal-report-by-interior-ministry-confirms-risks-to-digital-sovereignty/) - [Brennan Center: Section 702 FISA 2026 Resource Page](https://www.brennancenter.org/our-work/research-reports/section-702-foreign-intelligence-surveillance-act-fisa-2026-resource-page) - [BSI: NIS2 Implementation in Germany](https://www.bsi.bund.de/DE/Themen/Regulierte-Wirtschaft/NIS-2-regulierte-Unternehmen/nis-2-regulierte-unternehmen_node.html) - [Flexera: 2024 State of the Cloud Report](https://resources.flexera.com/web/pdf/Flexera-State-of-the-Cloud-Report-2024.pdf) - [Google: SRE Book – Embracing Risk](https://sre.google/sre-book/embracing-risk/) - [gehalt.de: DevOps engineer salary, Germany](https://www.gehalt.de/beruf/devops-engineer)