Lambda Cloud
On-demand NVIDIA instances for training, fine-tuning and serving — 1 to 8 GPUs launched in minutes with self-serve access — billed by the minute with no egress charge, alongside the 1-Click Clusters and liquid-cooled superclusters Lambda sells for larger runs.
Find alternatives to Lambda Cloud
What it does
- GPU cloud
- Model training
- Model inference
Features 8
- Multi-GPU instancesInstances with 8x, 4x, 2x or 1x GPUs to fit workloads from proof of concept to production.
- Lambda Cloud APIAPI to create, stop and restart instances from the CLI, CI/CD or orchestration scripts.
- Lambda StackOptimized ML stack with essential tools like PyTorch and CUDA pre-installed on instances.
- Built-in observabilityLive, minute-by-minute monitoring of GPU, memory and network performance in the dashboard or API.
- Persistent storageKeeps datasets, checkpoints and outputs attached between sessions with no egress fees.
- Filesystem S3 AdapterLets you interact with Lambda filesystems using rclone, s5cmd, mc and other S3-compatible tools.
- Single sign-on (SSO)Lets users authenticate to Lambda Cloud through an organization's existing identity provider.
- Firewall rulesetsGlobal or per-instance firewall rulesets can restrict connections to trusted source IPs.
How it compares
-
Nebius AI Cloud
For our core AI infrastructure business, our key competitors are specialized cloud service providers focused on AI, including CoreWeave, Crusoe and Lambda.
Regulatory filing · 18 Sep 2026 -
Together GPU Clusters
Lambda, founded in 2012, is a cloud-based GPU company that specializes in GPU workstations, cloud services, and AI infrastructure.
Research report · 19 Sep 2026
Sources
- Pricing
-
Instance names and per-GPU-hour rates read from the pricing page 2026-09-04; no free tier appears on it.
- Description
-
Official documentation · 4 Sep 2026
The service page's own words: "Train, fine-tune, and serve models on 1 to 8 NVIDIA GPU instances" and "Launch NVIDIA HGX B200, H100, A100, or GH200 instances in minutes with self-serve, first-come access."
- Constraints
-
Derived from the security facts on this page, from the same source.
- Security
-
Lambda Trust Center compliance list: SOC 2 Type 2, ISO/IEC 27001:2022, ISO/IEC 27017:2015 (also ISO/IEC 27701 and ISO 22301, not in the vocabulary). Platform-wide page; training and retention are not stated.
- Sold within
-
"Create your Lambda Cloud account and launch NVIDIA GPU instances in minutes." The page offers "self-serve, first-come access" and a per-GPU-hour rate card under "Pay by the minute".
- Interfaces
-
Official documentation · 14 Oct 2026
Page: 'Automate with the Lambda Cloud API to create, stop, and restart instances from your CLI, CI/CD, or orchestration scripts'.
- Intended for
-
Official documentation · 18 Oct 2026
Page: 'For every mission Superintelligence Enterprise Government Startups and researchers'.
Also from Lambda
-
Lambda 1-Click Clusters Infrastructure service
Self-serve GPU clusters from 16 to 2,000+ NVIDIA B200 or H100 GPUs, provisioned without a sales cycle for training, fine-tuning and large inference runs.
-
Lambda On-Demand Instances Infrastructure service
Hourly NVIDIA GPU instances — H100, H200, B200 and A100 — launched in minutes with no egress fees.
-
Lambda Inference Model API
Hosted inference endpoints for open-weight models on Lambda's own GPU fleet, billed per token.
-
Lambda Superclusters Infrastructure service
Rents dedicated large-scale AI training and inference GPU clusters at supercomputer scale.
Something wrong here? Send a correction — quote this product id: lambda-cloud.