Hetzner CX32 Pricing 2026: Deprecated Plan and CX33 Replacement

Quick Verdict

Hetzner Cloud CX32 is now a legacy plan. Hetzner lists CX32 among its deprecated Gen2 plans, so new buyers should choose CX33, the current 4-vCPU, 8GB, 80GB NVMe tier. Existing CX32 servers can continue to be useful; the important distinction is that CX32 is no longer the plan to recommend for a new deployment.

As of Hetzner’s pricing effective June 15, 2026, CX33 costs €8.49/month before VAT and before optional paid IPv4 in Germany and Finland. Read the CX32 vs CX33 comparison before resizing a running server.

The 8GB RAM allocation is useful for an n8n + database stack, multiple Node.js apps, or some quantized 7B models. If you only need one Node.js API or bot, compare the current CX23 before paying for CX33.

---

Hetzner CX32 Price in 2026: Legacy vs Current

Item	Legacy CX32	Current CX33
Order status	Deprecated	Available, subject to location capacity
Current EU monthly price	Not a new-order plan	€8.49 before VAT and IPv4
vCPU	4 shared vCPU	4 shared vCPU
RAM	8 GB	8 GB
Storage	80 GB NVMe	80 GB NVMe
Included traffic	20 TB	20 TB

The €8.49 figure comes from Hetzner’s official June 2026 price-adjustment notice. Hetzner also documents CX32 in its deprecated server plans. Prices vary by location, currency, tax status, and network options, so confirm the final amount in Hetzner Console.

---

What the 8GB Tier Can Run

The move from a 4GB plan to legacy CX32 or current CX33 is primarily about capacity, not guaranteed single-request speed.

Resource	Practical use
8 GB RAM	n8n + PostgreSQL, several small Node.js apps, or a quantized 7B model with limited headroom
4 shared vCPU	PM2 workers and bursty background jobs
80 GB NVMe	Application data, Docker images, logs, and a small database
20 TB traffic	More than most small API and automation stacks require

Hetzner says CX plans use shared resources and may run on Intel or AMD hardware; customers cannot choose the CPU model. Do not describe either CX32 or CX33 as guaranteed AMD EPYC compute. For sustained CPU-heavy work, compare CPX or dedicated CCX.

Should Existing CX32 Users Move?

Keep CX32 when the server is stable, correctly sized, and current account pricing is acceptable. Move when you are already rebuilding, changing regions, or rescaling—and only after checking the confirmation screen because Hetzner notes that some changes to legacy servers can activate current pricing.

The dedicated CX32 vs CX33 migration guide includes a backup checklist and a reproducible benchmark process.

Best Current Alternatives

Option	Choose it when
CX23	One Node.js API, bot, or small dashboard is the whole workload
CX33	You want the current 8GB cost-optimized Hetzner tier
CAX21	Your software supports Arm64 and availability fits your region
CPX	You need newer AMD shared compute and more consistent CPU performance
CCX	Sustained production CPU load requires dedicated resources

For provider-level tradeoffs, see the Hetzner vs DigitalOcean benchmarks.

Verdict

Do not order CX32 for a new deployment. Use CX33 as the current like-for-like 8GB tier. Existing CX32 users should migrate only for a concrete pricing, availability, rebuild, or capacity reason—not because CX33 has a newer name.

For setup, the Hetzner Ubuntu guide applies to current CX plans. For AI workloads, verify model memory needs in the Ollama VPS guide before assuming 8GB is sufficient.

FAQ

How much does Hetzner CX32 cost now?

CX32 is deprecated and has no current new-order price. CX33 is the replacement and costs €8.49/month before VAT and optional paid IPv4 in Germany and Finland under pricing effective June 15, 2026.

What specs did CX32 include?

Four shared vCPU, 8GB RAM, 80GB NVMe storage, and 20TB monthly transfer.

Is CX33 faster than CX32?

Not necessarily. Both are shared-resource plans, and CPU hardware allocation can vary. Benchmark the actual workload before migrating for performance.

Can an existing CX32 still host Node.js or Ollama?

Yes. Deprecation affects plan lifecycle, not whether a running server can execute workloads. Monitor memory and sustained CPU usage, and keep backups before any rescale.