GPU clouds / Hetzner GPU servers alternative

A Hetzner GPU servers alternative for the surrounding Linux agent

Hetzner GEX servers provide dedicated physical hardware with local drives and root access. The GEX45 includes one RTX PRO 4000 Blackwell SFF GPU with 24 GB VRAM, 64 GB system RAM and two 512 GB NVMe drives. This is an allocated server a team operates, with a different cost and recovery model from a managed per-user agent sandbox.

Start buildingCompare capabilities ↓

Reviewed 2026-10-19 · Published by the Sandbox0 team

When Hetzner GPU servers fits

Hetzner GPU servers fit workloads needing a dedicated RTX GPU, full kernel control, European datacenters and sustained use of a fixed physical machine. Installed operating-system software can remain on that machine’s local boot disk.

When to consider Sandbox0

Consider Sandbox0 for compatible Linux agents that orchestrate GPU jobs and should keep a regionally published whole writable RootFS while compute pauses or is replaced. GPU execution itself still needs suitable accelerator infrastructure.

Capabilities

Hetzner GPU servers and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeHetzner GPU serversNot established in this reviewSandbox0ttl and hard_ttl default to 0 (disabled). Configure idle pause and deletion policies for the same persistent sandbox identity; an idle running sandbox still accrues compute charges.Read documentation →
Persistent filesHetzner GPU serversDedicated local drives keep files during the lease and normal power cycles; backups and server replacement are separate responsibilities.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSHetzner GPU serversThe customer-managed boot disk can retain installed system software and configuration, not just a workspace folder.View evidence ↗Sandbox0Persistence extends beyond /workspace to paths such as /usr/local, /etc, and /home. Runtime tmpfs and virtual filesystems are excluded.Read documentation →
Release compute, keep stateHetzner GPU serversPhysical power off/on remains billed and is not a managed saved-state sandbox pause/resume contract.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateHetzner GPU serversRoot access and physical storage do not establish a platform-provided process/RAM checkpoint.View evidence ↗Sandbox0Experimental memory checkpoints can preserve compatible process state. Use memory: true on both pause and resume on a compatible deployment. A filesystem checkpoint alone does not preserve running processes.Read documentation →

Online RootFS growth

On growth-enabled images, Sandbox0 automatically expands the writable RootFS as free space runs low, within a configured sparse logical ceiling up to 64 TiB. Agent processes keep running. Growth is enabled in the qualified Cloud region; confirm activation for other regions. Existing images and snapshots retain their captured storage geometry. The ceiling is logical capacity, not preallocated host disk.

The GEX45 has a fixed physical storage configuration. Growing or replacing its storage requires the appropriate server and filesystem operations; there is no published automatic online growth contract for a managed per-agent RootFS.

Files while compute is paused

Sandbox0's Files API reads persistent files without resuming compute. Filesystem-only pauses also permit edits; retained-memory pauses are read-only and reject mutations with 409. SSH, SFTP, file watching, and runtime-only mounts require running compute.

Out-of-band server management and rescue environments are separate operational tools. They do not establish a simple RootFS file API while the original guest is paused. Sandbox0 provides supported offline RootFS operations, with read-only access for retained-memory generations.

See paused file operations →

Local whole-system retention is real, with hardware boundaries

Root access and dedicated NVMe drives let the customer keep system packages, configuration and user files across normal power cycles during the lease. These local drives are not a managed object-backed checkpoint that automatically follows replacement compute. Backups and restoration of a failed or replaced server remain part of the operating plan.

Source 1 ↗

Powering off does not end the rental bill

Hetzner bills a product as long as it exists on the account, including when unused or turned off. Most server usage is hourly with partial hours rounded up and a monthly maximum. Cancel the rental to end charges; power management alone is not a compute-free pause contract.

Source 1 ↗

The GEX45 price covers the complete configuration

The published GEX45 plus primary IPv4 configuration is €214/month before VAT, plus €209 setup. The GEX line uses one RTX GPU per server rather than configurable multi-GPU H100 nodes. Hardware changes and additional storage follow the available server configurations and operator workflows.

Source 1 ↗Source 2 ↗Source 3 ↗

Pricing

Compare billing models, then price your workload

CPU allocation, memory, billing minimums, disk, plan fees, and supported workload differ. These meters are not equal-performance machine quotes. Confirm region, resource units, and included allowances before estimating a total.

Hetzner GPU servers

Pricing reviewed 2026-10-19

GEX45 with IPv4 monthly maximum
€214 per server-month, excluding VAT
Complete server; official USD reference is $249/month.
Rate source ↗
GEX45 with IPv4 hourly reference
€0.3429 per server-hour, excluding VAT
Partial hours round up; usage is capped at the monthly server price.
Rate source ↗
One-time GEX45 setup
€209, excluding VAT
Official USD setup reference is $249; optional licenses and add-ons differ.
Rate source ↗

References use the GEX45 plus primary IPv4 configuration, not the lower server-only API amount. The current product page places GEX45 in Helsinki; do not infer sellable location inventory from every region returned by the price API. This is one allocated physical server, and its fee continues while powered off.

Check current Hetzner GPU servers pricing ↗

Sandbox0

$0.015/GiB-hour compute (billed by the second at approximately $0.000004167/GiB-second), plus $0.02/GiB-month of retained incremental RootFS and memory checkpoint data. Shared base images and platform recovery retention are excluded; paused sandboxes accrue storage charges only.

Compute / GiB-hour
$0.015
Retained data / GiB-month
$0.02
Paused compute
$0.00

A 4 GiB agent, 100 running hours

$6.00 compute

4 GiB × 100 hours, billed by the second. Keeping 20 GiB of billable incremental data for a full billing month adds $0.40. CPU is derived from memory by the platform; choose a tier that fits your workload.

Understand storage billing →

Before migrating from Hetzner GPU servers

  1. Back up required boot-disk files and configuration, record RAID and GPU dependencies, and identify any long-term service commitments.
  2. Retain dedicated GPU stages where appropriate and port compatible Linux agent orchestration, credentials and network policies.
  3. Test system-path retention, replacement-compute recovery, online growth and supported paused files; cancel the original rental only after the migration is verified.

Common questions

Is Sandbox0 a drop-in Hetzner GPU servers replacement?

Evaluate API compatibility and product scope. These products serve different core workflows, so a change may require a different architecture.

Does keeping files also keep a running agent?

Filesystem state and volatile process state are separate. Sandbox0 requires explicit compatible memory checkpoints to preserve running processes; its default filesystem resume starts fresh processes.

Sources and review scope

The Boat team's Battleships directory is the discovery source. References below describe the product and plan discussed on this page. Check current availability, regional terms and billing units before migrating. Where the cited material does not establish a capability or price, the comparison identifies that uncertainty.

Battleships research record ↗www.hetzner.com/dedicated-rootserver/gex45/ ↗docs.hetzner.com/general/billing-and-account-management/billing-at-hetzner/billing-system-hetzner/ ↗website-price-api.hetzner.com/api/v1/products/ROBOT_1768%2BROBOT_1266 ↗www.hetzner.com/pressroom/hetzner-expands-its-gpu-portfolio-with-the-gex45/ ↗www.hetzner.com/dedicated-rootserver/matrix-gpu/ ↗Sandbox0 RootFS capacity →

Related comparisons

Azure GPU VMs (NC_A100_v4 reference) vs Sandbox0 →Cerebrium Serverless Compute vs Sandbox0 →Crusoe vs Sandbox0 →DigitalOcean GPU Droplets vs Sandbox0 →Fluidstack vs Sandbox0 →Google Compute Engine GPU VMs vs Sandbox0 →