GPU clouds / Lambda GPU Cloud alternative

A Lambda GPU Cloud alternative for persistent Linux agent orchestration

Lambda GPU Cloud provides GPU instances, reserved clusters and independent regional filesystems. Current On-Demand documentation supports launch, restart and termination, and explicitly says suspension is unavailable. A returning agent must distinguish data retained in a filesystem from the local system state of a terminated GPU machine.

Start buildingCompare capabilities ↓

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

When Lambda GPU Cloud fits

Lambda fits GPU training, inference and cluster workloads with dedicated accelerator shapes and specialized infrastructure. Its independent regional filesystems can store datasets and backed-up system state across instance lifetimes.

When to consider Sandbox0

Consider Sandbox0 for compatible Linux agents around GPU work, with a durable whole writable RootFS, compute pauses, automatic online growth and supported paused-file access. Keep training and accelerator-dependent execution on suitable GPU capacity.

Capabilities

Lambda GPU Cloud and Sandbox0: what persists?

Feature comparison with documentation and verification status
Environment and runtime lifetimeLambda GPU 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 filesLambda GPU CloudIndependent regional filesystems preserve datasets and backups separately from the instance lifecycle.View evidence ↗Sandbox0The writable RootFS survives successful checkpoints, including software, system configuration, dependencies, and user data.Read documentation →
Whole RootFSLambda GPU CloudThe supported cost-ending path terminates the instance; whole local OS retention for a later instance is not provided by the independent filesystem alone.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 stateLambda GPU CloudCurrent On-Demand instances can launch, restart or terminate; suspend is explicitly unavailable.View evidence ↗Sandbox0Pause compute between tasks; resume on available capacity. Default resume starts fresh processes from the saved filesystem.Read documentation →
Memory and process stateLambda GPU CloudBackups and container images do not establish a durable process or GPU-memory restore feature.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.

Lambda’s independent filesystems provide large regional data capacity, with region-specific limits. That capacity does not mean the local boot RootFS automatically expands or survives termination; Sandbox0’s automatic online growth targets the complete enabled writable 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.

Lambda already documents remote file operations through its regional Filesystem S3 Adapter where available. Sandbox0’s relevant distinction is supported access to the complete durable writable RootFS while the original sandbox compute stays paused, rather than access only to an independently provisioned data store.

See paused file operations →

A guest shutdown does not stop the bill

The current instance guide permits launch, restart and terminate, but no suspend operation. In-guest shutdown or poweroff can leave the instance in Alert status with billing continuing. On-demand billing begins after launch health checks and ends at termination, in one-minute increments, regardless of active use.

Source 1 ↗Source 2 ↗

Independent filesystems already support remote data access

Regional filesystems can attach to an instance or cluster at launch and are separate from the instance’s local storage. They cannot currently attach to an already running instance or transfer between regions. A Filesystem S3 Adapter supports remote file operations in selected regions; other regions can use an instance and rsync. These are useful persistent files, not an automatically retained whole guest RootFS.

Source 1 ↗

GPU count changes both shape and hourly pricing

The currently rendered eight-GPU table prices H100 SXM at $3.99 per GPU-hour and B200 SXM6 at $6.69 per GPU-hour. Multiply by eight for the whole configured node. Other GPU counts, PCIe versus SXM variants and reserved 1-Click Clusters use different prices and commitments.

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.

Lambda GPU Cloud

Pricing reviewed 2026-10-19

8-GPU H100 SXM shape
$3.99 per GPU-hour
Eight GPUs: $31.92 per complete node-hour, before applicable tax.
Rate source ↗
8-GPU B200 SXM6 shape
$6.69 per GPU-hour
Eight GPUs: $53.52 per complete node-hour, before applicable tax.
Rate source ↗
Independent filesystem storage
Current rate shown at filesystem creation
Metered by GiB used per month in one-hour increments; continues while the filesystem exists.
Rate source ↗

Prices exclude applicable sales tax, VAT or GST. Public filesystem billing uses $0.20/GiB-month only as an explicitly illustrative example, so it is not used here as a verified current rate. On-demand instance fees and reserved cluster commitments differ; terminating compute does not remove independent filesystem charges.

Check current Lambda GPU 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 Lambda GPU Cloud

  1. Inventory local OS changes and outputs separately from independent filesystems, GPU shape, region, reservation and tax terms.
  2. Back up indispensable instance data before termination; retain GPU jobs and port only compatible Linux agent orchestration with required tools and policy.
  3. Test returning system state, online growth and supported paused files, while comparing complete GPU-node costs and retained filesystem usage.

Common questions

Is Sandbox0 a drop-in Lambda GPU 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.lambda.ai/public-cloud/on-demand/creating-managing-instances/ ↗docs.lambda.ai/public-cloud/billing/ ↗docs.lambda.ai/public-cloud/filesystems/ ↗lambda.ai/pricing ↗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 →