Hetzner April 2026 price increase: what changed and what it costs

Hetzner announced on February 24 that prices across its cloud portfolio will increase on April 1, 2026. The changes apply to both new orders and existing subscriptions across all Hetzner regions: Germany, Finland, the US, and Singapore.

The headline figure was "up to 37%." That number is real, but it doesn't answer the more useful question: what does your bill look like after April 1? This post breaks down the actual changes by resource type and works through the numbers on a realistic setup.

---

What's changing on April 1

Cloud servers (EUR billing)

Most cloud server tiers will increase by 30-34% in EUR. A few specific examples:

Server type	Before	After	Change
CX23 (2 vCPU / 4 GB, shared)	€2.99/mo	€3.99/mo	+33%
CPX11 (2 vCPU / 2 GB, shared)	€4.49/mo	€5.99/mo	+33%
CAX11 (2 vCPU / 4 GB, Arm64)	€3.29/mo	€4.49/mo	+37%
CCX13 (2 vCPU / 8 GB, dedicated)	€11.99/mo	€15.99/mo	+33%
CCX63 (32 vCPU / 128 GB, dedicated)	€287.99/mo	€374.49/mo	+30%

Cloud servers (USD billing)

USD-denominated prices increase more steeply, 38% to over 40%, depending on tier:

Server type	Before	After	Change
CPX11	$4.99/mo	$6.99/mo	+40%
CCX13	$14.49/mo	$19.99/mo	+38%
CCX63	$332.99/mo	$460.49/mo	+38%

Volumes and snapshots

Block volume and snapshot pricing both increase 30% in EUR and 53% in USD.

For the full list of affected resource types, see Hetzner's official price adjustment documentation.

---

Why is the increase this steep

DRAM costs have increased up to 500% since September 2025 — a structural crisis, not a temporary spike.

Three manufacturers: Samsung, SK Hynix, and Micron control approximately 95% of global DRAM production. As AI hyperscalers placed enormous orders for High-Bandwidth Memory (the stacked memory inside Nvidia GPUs), those manufacturers reallocated wafer capacity away from standard DRAM. Every gigabyte of HBM consumes roughly three times the wafer capacity of a standard DDR5 module. The result: memory for ordinary cloud servers became scarce, while AI infrastructure soaked up the available production capacity. TrendForce revised its Q1 2026 DRAM contract price forecast to a 90-95% quarter-over-quarter increase, nearly doubling in three months.

This is not a purchasing decision that Hetzner or any other European provider can negotiate around. The suppliers have pricing power, and smaller buyers, including cloud infrastructure operators that aren't hyperscalers, are at the end of the queue.

OVHcloud faced the same costs and made a different call

OVHcloud, Europe's other large cloud operator, announced price increases of 5-10% for April-September 2026, citing the same hardware cost surge. The contrast with Hetzner's 30-37% is stark and is explained by the business model, not by differences in component prices.

OVH operates at higher margins and can absorb a share of the cost increase without passing it all downstream. Hetzner built its position on the opposite logic: the highest-value infrastructure in Europe, sustained by the thinnest margins. A provider with no buffer cannot hold the line when costs spike 500% at the component level. The price increase is a direct consequence of the business model that made Hetzner attractive in the first place.

Will prices come back down?

This is the question Hetzner's public statement does not answer. Based on supply chain projections, the realistic answer is: not before 2028. New memory fabrication facilities take three to four years from investment decision to volume production. OVH, which has direct visibility into hardware procurement costs, stated explicitly that memory prices are not expected to return to historical levels before 2028. That estimate is for the best case.

Treat post-April pricing as your new permanent baseline through 2026 and 2027. Plans that assume a reversion mid-year are likely to be wrong.

---

What it costs in practice

The headline percentage is easy to find. The harder part is translating it into actual euros. Here are two concrete setups.

Small production environment: two servers and a volume

A web application running two CCX13 dedicated vCPU servers with a 100 GB block volume for persistent storage:

Resource	Before Apr 1	After Apr 1
2× CCX13	€23.98/mo	€31.98/mo
100 GB volume	€4.40/mo	€5.72/mo
Total	€28.38/mo	€37.70/mo

Dev environment with accumulated snapshots

A single CX23 running a development or staging workload, with three old snapshots at 50 GB each — a common pattern when snapshots are taken before deployments and never cleaned up:

Resource	Before Apr 1	After Apr 1
1× CX23	€2.99/mo	€3.99/mo
3× 50 GB snapshots	€1.65/mo	€2.15/mo
Total	€4.64/mo	€6.14/mo

The pattern scales: a team with 20 such dev environments and accumulated snapshot debt sees a proportionally larger jump, entirely from resources that aren't adding value.

---

Where the increase hits hardest

Large block volumes. The percentage increase is uniform, but storage-heavy workloads absorb the most in absolute terms. A 5 TB block volume will cost €66 more per month after April 1, with no change in usage.

USD-billed accounts and non-EU regions. USD and Singapore pricing increased 38-53%, more than the 30-37% EUR equivalent. The primary reason is structural: Hetzner leases, rather than owns, its US and Singapore data center space. Hardware inflation compounds with facility leasing costs and local commercial terms, producing steeper increases than in the EU, where Hetzner owns its infrastructure.

There is a second layer for teams running on Hetzner EU (EUR-billed) while tracking costs in USD. The euro has appreciated roughly 20% against the dollar since the October 2022 low. Teams that adopted Hetzner at that time have already absorbed that drift as an invisible cost increase. The April price hike lands on top of it.

EUR/USD rate, July 2022 - January 2026

Accumulated snapshots. Hetzner bills snapshots based on compressed image size. Teams that take snapshots before every deployment and never define a retention policy have built up snapshot storage that compounds quietly. That accumulated debt just got 30% more expensive to carry.

Powered-off servers. Hetzner charges the full server rate for instances that are off; resources remain allocated even when the server isn't running. A stack of idle dev or staging servers you've been meaning to delete is now 30% more expensive to leave sitting there.

---

Four things worth doing

Audit orphaned resources. Unassigned IPs, detached volumes, old snapshots, and powered-off servers all carry a cost. Deleting them before April 1 avoids paying the increased rate on resources that aren't doing anything useful. The hcloud CLI makes this straightforward; the Hetzner resource audit guide covers the specific commands for each resource type.

Tighten snapshot retention. If your deployment pipeline creates snapshots without defining when to delete them, the post-increase price is a concrete reason to add a retention policy now. Snapshot storage is billed on compressed size and accumulates silently.

Re-check USD projections. If you budget in USD, the effective increase for your Hetzner line item is not the 30-37% widely reported. Depending on your resource mix, region, and when you last locked in USD cost baselines, it may be closer to 40-53%, or higher once exchange rate drift is factored in. Update your cost model before the April invoice lands.

Review backup configurations. Automatic backups add 20% to server costs; that 20% is now applied to a higher base price. If backups are enabled by default on development or staging servers and you don't actually need them there, April 1 is a reasonable time to disable them.

Any cost estimate, runbook, or infrastructure plan with Hetzner line items needs updating before April 1. CloudTally shows your Hetzner spend broken down by resource type and region, so you can see exactly which resources are driving your increase, and act before the invoice arrives.