GPU clouds / Nebius AI Cloud alternative

A Nebius AI Cloud alternative for the Linux agent around GPU work

Nebius AI Cloud combines GPU and CPU VMs with managed network boot disks, shared filesystems and selected host-local SSD configurations. Standard network storage can retain installed operating-system tools, while local SSD is explicitly ephemeral. Current GPU pricing changed in October 2026, so older rate assumptions need a date and product check.

Start buildingCompare capabilities ↓

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

When Nebius AI Cloud fits

Nebius fits GPU training, inference and related cloud infrastructure needing accelerator presets, networking, managed storage and cluster integrations. Its network boot disks and independent data volumes already support whole-system and dataset retention.

When to consider Sandbox0

Consider Sandbox0 for compatible Linux agents around GPU jobs, with a complete durable writable RootFS, automatic free-space-triggered online growth and supported files while compute stays paused. GPU execution requires separate suitable accelerator infrastructure.

Capabilities

Nebius AI Cloud and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeNebius AI CloudNot 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 filesNebius AI CloudManaged network boot/data disks provide retained storage; host-local SSD and non-replicated variants have separate failure boundaries.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSNebius AI CloudA network boot disk can preserve the installed OS and system configuration, excluding separate local and runtime mounts.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 stateNebius AI CloudPlatform stop/start works with retained durable disks and separate storage billing; RAM continuation is not established.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateNebius AI CloudDisk snapshots are point-in-time disk state, not documented process or GPU-memory checkpoints.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.

Nebius already supports explicit hot disk enlargement, with a documented restart for OS recognition and subsequent filesystem steps. That differs from Sandbox0 automatically growing an enabled writable RootFS as free space runs low without restarting agent processes.

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.

Independent disks and shared filesystems can preserve data without running the original VM. That storage model alone does not establish a direct whole-RootFS file API against a stopped guest. Sandbox0 supports paused RootFS operations under its filesystem-only and retained-memory rules.

See paused file operations →

Choose the durable boot disk separately from scratch

Network SSD disks can be VM boot or additional disks. Standard Network SSD uses erasure coding, while Non-replicated disks explicitly lack redundancy. Local SSD on supported B300 presets is lost on Compute stop, deletion or host failure. Do not generalize a local scratch lifecycle to the network boot disk, or assume every disk type has the same durability.

Source 1 ↗Source 2 ↗

Stop through the platform to end running compute charges

Stopped VM GPU, vCPU and RAM resources are not charged, but provisioned volumes remain billed independently. Linux shutdown or halt is not the supported stopping mechanism: Nebius treats it as failure, can reboot the VM and continues charging. Stop/start preserves the selected durable disks rather than establishing a saved-RAM resume contract.

Source 1 ↗Source 2 ↗

Hot disk resize still has a restart step

Nebius supports increasing a disk without stopping the resize operation itself, but its guide requires a VM restart afterward for the OS to recognize the new size. Partition and filesystem steps depend on the disk. Boot disks are limited to 30 TiB, and block size can further constrain capacity; shared filesystems have separate resize controls.

Source 1 ↗

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.

Nebius AI Cloud

Pricing reviewed 2026-10-19

Current H100 NVLink reference
$4.50 per GPU-hour
USD, before tax; the current H100 platform is in eu-north1.
Rate source ↗
Current B200 NVLink reference
$8.50 per GPU-hour
Check the specific B200 platform, region and complete preset.
Rate source ↗
Current H200 NVLink reference
$5.40 per GPU-hour
Rate source ↗
Standard Network SSD capacity
$0.071 per provisioned GiB per 730 hours
Volumes remain charged while compute is stopped.
Rate source ↗

These GPU prices use the current October 2026 on-demand rows, not pre-October rates or advertised minimum spot rates. Compute is metered per second. Dynamic spot pricing can change during use, as often as every 15 minutes; confirm the complete preset, regions, storage and other resources. USD applies except where Nebius specifies ILS for Israeli companies.

Check current Nebius AI Cloud 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 Nebius AI Cloud

  1. Separate durable network boot/data disks from local scratch and record disk redundancy, GPU preset, regional quotas and spot or reservation terms.
  2. Back up indispensable local data before Compute stop and port compatible Linux agent stages while retaining needed GPU services.
  3. Test durable system paths, disk-growth recognition, supported paused files and recovery, then compare the current dated complete-resource bill.

Common questions

Is Sandbox0 a drop-in Nebius AI Cloud 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 ↗docs.nebius.com/compute/storage/types.md ↗docs.nebius.com/compute/storage/local-disks.md ↗docs.nebius.com/compute/resources/pricing.md ↗docs.nebius.com/compute/virtual-machines/stop-start.md ↗docs.nebius.com/compute/storage/manage.md ↗docs.nebius.com/compute/storage/disk-snapshots.md ↗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 →