Hetzner vs DigitalOcean vs Vultr After Two 2026 Price Hikes: Which Cheap VPS Wins Each Tier of Your Fleet?

Hetzner raised cloud prices twice in 2026 — 30 to 37 percent in April, then again in June with some lines more than doubling — and it is still the cheapest of the three budget VPS shortlists by a factor of three or more. That sentence sounds like the end of the comparison. It is actually the beginning, because the headline per-core price was never what decided a fleet bill. The 2026 three-way pricing audit behind this post priced all three providers at identical specs and found the gap intact but the fine print doing the real differentiating: separately billed IPv4, a 20-to-1 transatlantic traffic split, and a managed-service menu so thin it only makes sense when your platform is the managed layer.

So here is the verdict first, node tier by node tier, for a self-hosted PaaS running its own Cluster API fleet:

List prices move — every figure below was checked against July 2026 audits — but the structure of the answer has survived both hikes. Price the box, then price the traffic, then price the IPv4, then ask who operates the database. In that order.

What the two 2026 rises actually changed​

The first rise landed April 1, 2026: 30 to 37 percent across the Hetzner cloud range, with The Register reporting figures up to 50 percent and Cybernews noting budget plans hit hardest, some surging past 200 percent. The stated cause was the hardware market — tight RAM, SSD, and GPU availability, with DRAM and NAND flash cost increases passed through to list prices. The entry CX22 (2 vCPU, 4 GB RAM) settled at €4.49 a month, and the CX23 moved from €3.79 to €5.49. Crucially, the adjustment applied to new orders only; existing contracts kept their original terms.

The second rise took effect June 15, 2026 at 8 AM CEST, covering new orders and instance rescales. This one was steeper on the pricier lines: the three-way July 2026 audit records AMD shared (CPX) plans more than doubling and dedicated-vCPU (CCX) lines roughly tripling, while the entry Intel CX line stayed the cheapest of the three providers at the same spec. Two rises in ten weeks, and the lead narrowed without flipping — which tells you how wide it was to begin with.

Context matters here. Mid-tier providers broadly raised prices through 2026 on hardware costs while hyperscalers with long-term supplier contracts stayed more insulated, so Hetzner's hikes are a market move, not a strategy change. Nobody repriced to fund a new managed-services catalogue; the menu is the same thin one it always was. That is the whole story of the next three sections: the box got more expensive, the fine print stayed the same shape, and the fine print is where fleets win or lose.

Same spec, three prices​

Take the workhorse size — 2 vCPU, 4 GB RAM, the node a small fleet actually buys in quantity. At July 2026 list prices, that spec costs around €5–6 a month on Hetzner's Intel CX line (40 GB SSD, 20 TB of EU traffic included), about $24 a month on DigitalOcean's Basic droplet (80 GB SSD), and about $24 a month on Vultr's High Performance plan (100 GB NVMe, 5 TB transfer). Same nominal CPU and RAM; Hetzner runs three to four times cheaper while the other two ship more and faster disk for the money.

Step up to 4 vCPU / 8 GB and the shape holds: DigitalOcean lists $48, Vultr prices its equivalent tier in the same neighborhood, and Hetzner stays a fraction of either. A June 2026 four-way price check put Hetzner's CX32 at $8.20 and AMD CPX21 (3 vCPU, 80 GB NVMe, 20 TB) at $8.90 against $24 for both DigitalOcean and Vultr at 4 GB — and early-2026 comparisons show the same 3–4x ratio before the April rise, which means the hikes compressed Hetzner's margin without changing the ranking.

The honest caveat runs the other direction: Hetzner's cheapest box also ships the least disk (40 GB vs 80–100 GB), slower Intel shared cores on the CX line, and no NVMe until you step to CPX. If your node tier is disk- or single-thread-bound rather than count-bound, the per-core ratio overstates the win. But fleet nodes — control planes, workers, build runners — are overwhelmingly count-bound commodity boxes, and at count-bound scale the ratio is the bill. Three control-plane nodes at ~€6 each is a rounding error; three at $24 each is a line item your finance thread will ask about.

The fine print that decides it​

Headline price gets you to the shortlist; four surcharge lines pick the winner.

IPv4. Hetzner bills IPv4 separately at about €0.60 a month per address — small, but it applies to every public node in the fleet, so add it to every row of your spreadsheet. DigitalOcean includes a public IPv4 with every droplet. Vultr's cheapest Regular plan is IPv6-only at $2.50, stepping to about $3.50 with IPv4. None of these change the ranking alone, but the Hetzner fee is the easiest line to forget and then rediscover across forty nodes.

Included transfer — the big one. Hetzner's EU regions include 20 TB of outbound traffic per plan, against a pooled allowance starting at 500 GB and scaling with plan size on DigitalOcean, and roughly 2–6 TB per plan on Vultr. Overage rates invert the rest of the pricing story: Hetzner charges about €1 per TB in EU and US regions, while DigitalOcean and Vultr both bill $0.01 per GB — roughly $10 per TB, ten times Hetzner's rate. For bandwidth-heavy tiers the transfer allowance is not a tiebreaker; as Better Stack's 2026 comparison puts it, on Hetzner EU this cost line effectively disappears.

The transatlantic split. Hetzner's US regions (Ashburn, Hillsboro) include only 1 TB of traffic per plan, not 20 TB — and the budget CX/CAX lines are EU-exclusive, so US nodes must be the post-June CPX/CCX lines at post-hike prices. Singapore is thinner still (from 0.5 TB, with overage at €7.40 per TB). Regional pricing trackers confirm the shape: same provider, two different products divided by an ocean. Any fleet math computed on EU terms must be recomputed from scratch for US nodes — this is the single most common error in Hetzner-versus-everything spreadsheets.

Storage and data protection. Outgrow the included disk and all three sell block volumes per provisioned GB: Hetzner at about €0.057 per GB-month (raised from €0.044 in April 2026), DigitalOcean and Vultr at about $0.10 — so Hetzner's lower volume rate keeps it cheaper even though it includes less disk. Automatic backups cost 20 percent of the instance price on Hetzner and Vultr, and 20–30 percent on DigitalOcean depending on weekly or daily cadence; since the surcharge scales with instance cost it never flips the ranking. Snapshots diverge more: roughly €0.0143 per GB-month on Hetzner against $0.05 on Vultr and $0.06 per GiB on DigitalOcean — a 3–4x spread that matters exactly as much as your dev-cluster churn rate says it does.

Node by node: who wins each tier of a Cluster API fleet​

Now spend the numbers above on a concrete fleet: three control-plane nodes, a worker pool, build runners, and a dev cluster, all reconciled by Cluster API on provider VMs.

Control-plane VMs: Hetzner, by a walk. Three small EU nodes at ~€5–6 each plus ~€0.60 IPv4 land near €20 a month all-in. The equivalent on DigitalOcean or Vultr starts around $12 per node at smaller sizes and climbs past $70 for the trio at comparable spec — and control planes are the worst place to pay a premium, since they sit at low utilization doing etcd and API-server duty while transfer stays far under every allowance. The only reason to put control planes elsewhere is region: if the fleet must live where Hetzner has no datacenter, Vultr's 30-plus regions win by availability, not price.

Workers and build runners: Hetzner in the EU, recompute for the US. Workers inherit the 3–4x box advantage, and build runners add the transfer story: image pushes, artifact uploads, and tenant egress all draw on the allowance, and 20 TB at €1/TB overage versus pooled gigabytes at ~$10/TB is a structural moat. But move the tier to Ashburn or Hillsboro and both edges dull at once — 1 TB included, no CX line, post-June CPX prices. A US worker tier pushing real egress can easily land cheaper on DigitalOcean's pooled transfer or Vultr's per-plan 5 TB once Hetzner overage enters the picture. This is the sensitivity analysis the title promises: region is the variable that flips the table, so run both columns.

Dev and ephemeral clusters: Hetzner, on churn economics. Dev tiers live and die by spin-up cost, hourly granularity, and snapshot pricing for golden images — all three favor the cheapest box with the cheapest snapshots. The managed-service gap is also narrowest here: nobody needs managed Postgres for a cluster that lives a week. One operational note: Hetzner applies current list prices to rescales, so a dev node created before June and rescaled after inherits post-hike pricing. For ephemeral capacity that churns constantly this is already your reality; for long-lived dev boxes, prefer recreate-over-rescale.

US-latency nodes: DigitalOcean or Vultr, picked by need. With Hetzner US offering only pricier lines and 1 TB, the contest is between the other two: DigitalOcean when the tier benefits from the surrounding platform (Spaces for artifacts, a managed database next to the app, DOKS instead of self-run control planes), Vultr when the requirement is a specific region or NVMe on a small budget. Head-to-head 2026 testing gives Vultr the hardware edge at $24 (100 GB NVMe and 5 TB vs 80 GB SSD and 4 TB), while DigitalOcean bundles DDoS protection Vultr sells for $10 a month extra — for a three-node production tier that surcharge alone closes much of the headline gap, so price protection in, not on.

Where the managed-menu gap disappears (and where it doesn't)​

The standard knock on Hetzner — thin managed menu: raw compute, volumes, load balancers, networks, no deep catalogue of managed databases — is accurate and, for fleet nodes, mostly irrelevant. A self-hosted PaaS on Cluster API already operates its own control planes, its own build pipelines, its own Postgres (on volumes, with backups it controls), and its own observability. Paying DigitalOcean's premium for managed services the platform team has already built is paying twice. The audit's framing is exactly right: Hetzner sells raw compute, and a PaaS team is the managed layer.

The gap stops being irrelevant at two boundaries. First, managed data services the team has not built: if the choice is DigitalOcean's managed Postgres and Valkey versus a self-run database with no on-call rotation behind it, the $24-versus-€6 box comparison is measuring the wrong thing — the managed service buys back its premium in incidents that never happen. Second, object storage and serverless-adjacent SKUs (Spaces, App Platform, DOKS) that have no Hetzner equivalent at any price; Vultr's catalogue is wider than Hetzner's but shallower than DigitalOcean's, which is why the three-way map reads Hetzner–Vultr–DigitalOcean in ascending order of both menu and bill.

So the decision rule for 2026, after both hikes: default every self-managed fleet tier to Hetzner EU, recompute US tiers from the 1 TB baseline instead of assuming EU terms travel, and spend the DigitalOcean premium only where it replaces headcount — a managed database, object storage, or a control plane you would otherwise page yourself for. Region coverage beyond all three catalogues is Vultr's race to lose. Re-check list prices before signing anything; the team that raised twice in ten weeks has shown its willingness to move them.

Pricing a fleet's own nodes is exactly the decision a self-hosted PaaS should make boring: declare the tiers, let reconciliation hold them, and stop re-deriving the math every invoice. Bex.co is the open-source, AI-native Render alternative — push a git repo, get a running HTTPS service on machines you own. Star the repo on GitHub or deploy your first app today.