Dedicated servers for teams who outgrew Hetzner self-service and never wanted AWS in the first place

Honest answer: for many teams, you would not. Hetzner is excellent at what it does, the hardware is solid, the price-to-performance is unbeatable in Europe. We win in three specific scenarios. First: when you need EU sovereignty stronger than "German company" — Hetzner is a strong German operator, but their group structure has less clean separation from US partnerships than ours, and our Austrian GmbH structure with EU-only sub-processors is sometimes the easier audit answer. Second: when you want operational support, not pure self-service — Hetzner does not offer managed operations and does not pretend to. Third: when you need multi-PoP coverage beyond Hetzner's footprint, particularly the UK or Panama options. If none of those scenarios apply, Hetzner is genuinely a fine choice and we will tell you that on a discovery call.

You give us monitoring hooks (or we set up Prometheus and Grafana for you), you tell us which alerts you want us to action, and we set up an on-call rotation. When an alert fires, an engineer investigates within the SLA window (15 minutes for critical, 1 hour for warning). For incidents that require action — failed disk, runaway process, network issue, OS crash — we execute the remediation. We do not touch your application code or your data; the boundary is "everything OS-level and below". For application-level incidents, we escalate to your team. The €199/month per node covers a typical workload. If your environment is unusually noisy (more than 5 high-priority incidents per month sustained), we may have a conversation about adjusting the contract.

Debian 12, Ubuntu 24.04 LTS, AlmaLinux 9, Rocky Linux 9, and RHEL 9 with valid Red Hat subscription that you bring. Windows Server is available on request but we do not actively recommend it for the workloads our customers typically run. FreeBSD and other niche OSes can be installed on request for an extra setup fee. The default is Debian 12 unless you specify otherwise during onboarding. Custom OS images can be provided via our deployment API or as ISO upload.

You have three options. First: bring your own backup solution (Restic, BorgBackup, Duplicacy, your existing offsite backup target) — most customers do this. Second: use our managed backup add-on at €0.02 per GB-month, which orchestrates daily snapshots to our EU-based object storage with point-in-time recovery up to 30 days. Third: integrate with cloud object storage of your choice (we recommend Backblaze B2 EU region for cost, or Wasabi EU region for predictable pricing). The managed backup add-on includes restore testing once per quarter and documented restore procedures. We are not a backup company — we just provide enough hooks to do this right.

Yes, that is a common pattern. KVM, Proxmox VE, VMware vSphere, and ESXi all run cleanly on our hardware. Customers who run their own hypervisors typically do so for one of three reasons: cost optimization across multiple workloads on a single physical box, license-driven requirements (Windows licensing in particular is much cheaper if you control the underlying hardware), or VM-density use cases that do not fit neatly into cloud pricing. Our hardware is virtualization-friendly: ECC RAM, AMD-V or Intel VT-x enabled in BIOS by default, IOMMU support for PCI passthrough.

Available on Custom tier with case-by-case pricing. We have stock of NVIDIA L4 (cost-effective inference, around €1,200/month per node), L40S (heavier inference and lighter training, around €2,400/month), and H100 (training, around €4,500-€5,500/month depending on configuration). We do not stock A100 anymore because L40S has displaced it for most workloads. For customers who need a single GPU node for evaluation, we can do month-to-month. For ongoing GPU capacity (training pipelines, persistent inference workloads), annual contracts come with better pricing.

Volumetric mitigation up to 100 Gbps is included by default on every node and handles the vast majority of attacks against typical web applications. For services that attract bigger attacks — public game servers, popular SaaS apps, anything political — we have arrangements with our upstream for additional capacity, and we offer a layer-7 mitigation add-on (€99/month per node) that handles application-layer attacks (HTTP floods, slow loris, request smuggling). The combination handles real-world attacks well, but if your threat model includes nation-state-level adversaries we are not the right vendor — that level requires specialized DDoS providers (Cloudflare Magic Transit, Akamai Prolexic) and we will tell you that.

Yes. We support cross-connects within our PoPs (we have presence at major colos in each city), VPN concentrator setups for IPsec or WireGuard tunnels back to your office, and dedicated layer-2 transport between our PoPs and any other datacenter that has a presence in the same exchange. The cross-connect option is most common for customers who already colocate elsewhere and want our boxes integrated into their existing network. Talk to us during onboarding for the specific arrangement.

Vienna is our primary PoP and the one we operate ourselves; the others are colocated with reputable third-party datacenter operators where we have racks. If Vienna has an incident, customers with multi-PoP setups (HA Cluster across PoPs, or workloads replicated to Frankfurt as DR) keep running with minimal impact. Customers running in Vienna only would experience an outage during the incident. We document incident history publicly on the status page; the worst event in the last 5 years was a 47-minute regional network issue in Q3 2025, no hardware loss. We are not promising no incidents will ever happen — we are promising they will be rare and we will be honest about them when they do.

Bare metal does not flex the way cloud VMs do — that is the trade-off. For workloads that need genuine elasticity, cloud is the right answer and we will tell you that. For workloads that are mostly steady with occasional bursts, the right pattern is dedicated baseline + cloud spillover for peaks. Many customers run their steady workloads with us and burst into AWS/GCP/Azure for peaks via a hybrid setup. We can help configure this during onboarding. If your workload is genuinely elastic with 10x daily peaks, you should probably stay on cloud and revisit when the steady baseline grows enough.

Yes, two options. First: month-to-month billing on every tier with no setup fee for the Single Node tier — you can run for 30 days and cancel if it does not work. Second: for the Performance Node tier, we offer a 14-day evaluation at half price for first-time customers, no commitment. The Custom tier requires more upfront discovery but we structure the first 90 days as a structured evaluation period with milestones. We do not offer free trials — the hardware cost is real and free trials in dedicated hosting tend to attract customers who never convert.

For HIPAA, we offer a BAA addendum to the standard DPA on Performance Node and above. For PCI DSS, our infrastructure is in scope as a "service provider" in your audit, and we provide the documentation needed for your QSA — control mappings and audit logs covering the controls relevant to PCI Section 12. We do not currently hold SOC 2 or ISO 27001, but we provide pre-certification control evidence on request. For specific compliance scenarios that are not covered by these standards (DORA, NIS2, sector-specific requirements), we have a compliance contact who works with your team during onboarding to design the appropriate contractual structure.

We have an Acceptable Use Policy and we enforce it. We do not host child sexual abuse material, terrorism content, malware command-and-control infrastructure, or anything illegal under Austrian or EU law. Beyond that, we are jurisdiction-aware: hosting decisions sometimes depend on which PoP your workload sits in. Content that is legal in Panama may not be legal in Frankfurt and vice versa. We work with customers to put workloads in the right jurisdiction. We do not host phishing infrastructure, scam operations, or cryptocurrency mining as a primary business model. We do host adult content (legal and consensual), cryptocurrency exchanges with proper KYC, gambling operators with valid licenses, and political content (across the spectrum) that is legal in the relevant jurisdiction. The AUP is on the legal page; read it before signing if there is any doubt.

Stages, almost always. Most customers migrate one workload at a time over 4 to 12 weeks. The typical pattern: start with a single non-critical workload (a CI runner, a staging environment, a development database) on the new infrastructure, run it for 2 to 4 weeks to validate the setup, then move a production workload, then another. By the time you cut over the last critical workload, you have months of operational experience with us and the migration risk is much lower than a big-bang switchover. For customers migrating off cloud, we offer a hybrid setup option where your old cloud infrastructure and our new dedicated infrastructure coexist in a single VPC via VPN tunnel during the transition. This costs nothing extra on our side; you just keep paying the cloud provider until you fully cut over.

Each server gets redundant high-capacity uplinks with DDoS protection at the edge, dedicated IPs, and the option on Enterprise to announce your own IP allocation. Datacenter locations are EU-resident plus North America and Panama, and you choose where your server sits during configuration. We keep the internal routing and addressing detail private, but we are happy to walk through the setup relevant to your use case under NDA.

Standard refresh cycle is 3 to 4 years from initial deployment. Hardware that has been in service longer typically gets retired or moved to lower-tier roles (development, internal use, lower-priority workloads). When a customer is on hardware approaching end-of-life, we contact them 90 days before the planned retirement and offer either a free migration to newer hardware in the same tier or, for customers running stable workloads on amortized hardware that does not need to move, the option to stay on the existing hardware at the same price for an extended period. We do not force migrations as a revenue tactic. Hardware refreshes are operationally expensive for us too and we coordinate them around your maintenance windows when possible.