TL;DR
Hetzner is roughly 3-4x cheaper than DigitalOcean and Linode (Akamai) for the same vCPU and RAM, and that gap has only widened through 2026. DigitalOcean still wins for teams that want managed Postgres, Spaces, and App Platform under one bill. Linode is the middle child: priced like DigitalOcean, engineered like a traditional VPS host, and useful when you need a US-region alternative that isn't AWS-adjacent. The real cost story, though, isn't the sticker price on a droplet or a CX instance. It's snapshots, floating IPs, load balancers, and egress overages, where a single bad month of traffic can wipe out a year of "cheap VPS" savings.
The headline numbers (2026 list pricing)
Let's anchor on the smallest production-grade instance each vendor sells, because that's where most self-hosted open-source apps actually live. A Ghost blog, a Plausible install, a Gitea server, a small Mastodon node — none of it needs more than 2 vCPU and 4 GB of RAM to start.
| Tier | Vendor | Spec | Price (mo) | Included egress |
|---|---|---|---|---|
| Entry | Hetzner CX22 | 2 vCPU (Intel), 4 GB, 40 GB NVMe | €4.59 | 20 TB |
| Entry | Hetzner CAX11 | 2 vCPU (Arm), 4 GB, 40 GB NVMe | €3.79 | 20 TB |
| Entry | DigitalOcean Basic | 1 vCPU, 2 GB, 50 GB SSD | $12 | 2 TB |
| Entry | DigitalOcean Premium AMD | 2 vCPU, 4 GB, 80 GB | $28 | 4 TB |
| Entry | Linode Nanode | 1 vCPU, 1 GB, 25 GB | $5 | 1 TB |
| Entry | Linode Shared 4GB | 2 vCPU, 4 GB, 80 GB | $24 | 4 TB |
Two things jump out. First, Hetzner's CAX11 (Ampere Arm) gives you 2 vCPU and 4 GB for under €4 — less than a Nanode with one quarter of the RAM. Second, the comparable DigitalOcean and Linode plans sit within a dollar or two of each other; this isn't a coincidence, it's competitive anchoring.
Cost per vCPU·hour, normalized
Divide it out and the gap is stark. Per vCPU per month (rounded, using equivalent 2 vCPU / 4 GB tiers):
- Hetzner CAX11 (Arm): ~€1.90 per vCPU
- Hetzner CX22 (Intel): ~€2.30 per vCPU
- Linode Shared 4 GB: ~$12 per vCPU
- DigitalOcean Premium AMD: ~$14 per vCPU
You are paying roughly six times more per vCPU on DigitalOcean than on Hetzner's Arm fleet. That premium buys you the US-East datacenters, the integrated managed services, and arguably better networking on the public internet. Whether that's worth a 500% markup depends entirely on whether you'll use those services.
Where the spec sheet lies: hidden line items
The advertised droplet price is the start of the bill, not the end. Here's what actually shows up on an invoice once you run a real workload.
Snapshots and backups
Snapshots are billed by storage consumed, and the rates are wildly different:
- Hetzner: €0.0119 per GB per month for snapshots; automated backups are an additional 20% of the server price (so a €4.59 CX22 costs about €0.92/mo for backups, full stop — refreshingly simple).
- DigitalOcean: $0.06 per GB per month for snapshots; backups cost 20% of the droplet price for weekly, or 30% for daily.
- Linode: Backups are a flat per-plan fee (~$2/mo on a Nanode, ~$5/mo on the 4 GB plan); snapshots are included in that subscription.
If you keep a 40 GB snapshot for a year, you'll pay ~€5.70 on Hetzner versus ~$29 on DigitalOcean. Multiply that across a fleet and snapshots become a real budget item, not a rounding error.
Reserved / floating IPs
- Hetzner: a primary IPv4 is €0.50/mo when assigned, and €0.60/mo when unassigned (yes, they charge more for parked IPs to discourage hoarding). IPv6 is free.
- DigitalOcean: a Reserved IP is free while attached, $4/mo when unattached.
- Linode: additional IPv4s require justification and are typically free when attached; reserved/floating IPs vary by region.
If you run blue/green deploys with detached floating IPs between rollouts, DigitalOcean's "free when attached" rule is genuinely nicer than Hetzner's flat charge.
Load balancers
This is where the gap gets uncomfortable.
| Vendor | Entry LB price | Notes |
|---|---|---|
| Hetzner LB11 | €5.39/mo | 10k connections, 20 TB traffic |
| DigitalOcean | $12/mo | 1 node, modest throughput; HA tier is $24+ |
| Linode NodeBalancer | $10/mo | Per-NB, decent throughput |
A Hetzner load balancer costs roughly the same as a CX22 instance. A DigitalOcean LB costs more than the droplets it fronts. For a small three-node setup, the LB alone can be 25-40% of the bill on DO/Linode.
Block storage
- Hetzner Volumes: €0.0440/GB/mo (~€4.40 per 100 GB).
- DigitalOcean Volumes: $0.10/GB/mo (~$10 per 100 GB), plus tiered IOPS.
- Linode Block Storage: $0.10/GB/mo, similar tier.
If your app is storage-heavy (Nextcloud, a Postgres warehouse, a Plex library), you'll see the divergence here as much as on compute.
Egress — the silent budget killer
This is the one that makes or breaks the comparison.
- Hetzner: 20 TB included on most cloud instances; €1.00 per TB overage on EU/US locations (€1.20 in Singapore). That's not a typo — one euro per terabyte.
- DigitalOcean: 500 GB to 10 TB included depending on droplet size; $0.01 per GB ($10/TB) overage.
- Linode: 1-20 TB included by plan tier; $0.005 per GB ($5/TB) overage on most plans.
Hetzner overage is ten times cheaper per TB than DigitalOcean. Run a busy Mastodon instance pushing 5 TB/month over its allowance and you've added $50 to the DO bill versus €5 on Hetzner. For a video, file-sync, or media workload, this single line item often dominates the entire decision.
Networking and regional realities
Hetzner's cheap egress comes with caveats. Their network into the US is fine for European audiences hitting a US-East node, but latency from Falkenstein or Nuremberg to North American users is what you'd expect — 90-120 ms transatlantic. They opened Hillsboro and Ashburn in recent years, which closed the gap for US-targeted workloads, but capacity in those regions has been intermittently constrained.
DigitalOcean and Linode both have broader datacenter footprints (12+ regions each) and more mature private networking, including VPC peering across regions on DigitalOcean. If you need Sydney, São Paulo, or Mumbai with a single provider, Hetzner is not your answer.
There's also the Akamai factor: Linode is now part of Akamai, and the roadmap has tilted toward enterprise-edge integration. For a solo developer this is mostly invisible, but the platform feels less "indie VPS" and more "Akamai's compute tier" than it did three years ago.
Managed services: where DO earns its premium
If you only ever spin up Linux VMs, Hetzner wins on price every time. The picture changes when you start using managed offerings:
- Managed Postgres: DO starts around $15/mo for a 1 GB / 1 vCPU node with daily backups and PITR. Linode's managed DBs are similarly priced. Hetzner has no managed database product — you run Postgres yourself or use an external provider like Neon or Supabase.
- Object storage: DO Spaces is $5/mo for 250 GB plus 1 TB egress, then $0.02/GB. Linode Object Storage is comparable. Hetzner Object Storage launched recently with very competitive pricing (~€5.99/TB stored, generous egress), but tooling and ecosystem maturity is behind S3-compatible incumbents.
- Kubernetes: DOKS and LKE are free control planes; you pay only for worker nodes. Hetzner doesn't offer managed Kubernetes — you run k3s/k0s on raw instances, or use Hetzner Cloud as a backend for tools like Cluster API.
If your stack is "VM + Postgres + S3 + LB," DigitalOcean's all-in bill might land within 30% of a fully self-managed Hetzner equivalent, and you don't have to operate the database yourself. That's a real trade.
Which one wins, by workload
| Workload | Best pick | Reasoning |
|---|---|---|
| Static site / personal blog | Hetzner CAX11 | €3.79/mo, 20 TB egress. Done. |
| Self-hosted Ghost / Plausible / Umami | Hetzner CX22 | 4 GB RAM is plenty; backups under €1. |
| Mastodon / Pixelfed instance | Hetzner + S3-compat | Egress allowance is the killer feature. |
| Small SaaS with managed Postgres | DigitalOcean | One bill, PITR, no Postgres ops. |
| US-only B2B app, latency-sensitive | Linode or DO US regions | Broader US presence than Hetzner. |
| Nextcloud / file sync (heavy storage) | Hetzner Volumes | Block storage is less than half DO/Linode price. |
| Multi-region edge | DigitalOcean | Most regions, VPC peering, Spaces CDN. |
| GPU / ML inference | None of the above | Look at Lambda, RunPod, or Hetzner dedicated GPU. |
| Hobby / learning | Linode Nanode or Hetzner CAX11 | $5 vs €3.79 — pick your ecosystem. |
What to do this week
A pragmatic decision checklist before you migrate or commit:
- Estimate monthly egress honestly. Pull last month's bandwidth from your current host or CDN logs. If you're over 2 TB/mo, Hetzner's pricing alone justifies the move.
- List the managed services you actually use. If you're paying for managed Postgres, Redis, and object storage on DO, the cost gap to Hetzner narrows once you add equivalent third-party services (Neon, Upstash, Backblaze B2).
- Price the full stack, not just the VM. Add: instance + backups + 1 LB + 1 reserved IP + expected egress overage. Use the osscostcalc calculator on this site to plug in your real numbers across all three providers — the relative ranking often flips once load balancers and egress are included.
- Test latency from your actual user geography. Spin up a €4 Hetzner instance and a $6 droplet for a day, run real RUM measurements. Specs lie; networks don't.
- Decide on lock-in tolerance. Hetzner's API and ecosystem are good but small. DO's ecosystem (Terraform, App Platform, marketplace) is broader. If you're a one-person team, breadth matters less than monthly cost.
- Set a billing alert. Whichever you pick, enable a hard cap or alert. Egress surprises are the single most common "why is my bill 5x" story across all three providers.
The honest summary: in 2026, if price-per-resource is your dominant constraint and your users are in Europe or you can tolerate Hetzner's US regions, Hetzner is the obvious answer for raw compute. If you want a managed control plane and don't want to think about Postgres backups at 2 AM, DigitalOcean still earns its premium. Linode sits between them — pick it when you want a US-centric alternative with traditional VPS ergonomics.