Lambda GPU pricing

Lambda is an AI infrastructure provider offering self-serve GPU instances, 1-Click Clusters and larger private or supercluster deployments. It fits teams that want datacenter GPU capacity, preconfigured ML images and a path from single instances to large training clusters. On-Demand Cloud is billed by hourly usage in one-minute increments, while 1-Click Clusters are reserved and invoiced on their own terms.

Live Lambda prices (2026-07-08)

Cheapest listed per-GPU-hour rates verified at the source. Before storage, egress and networking.

Pricing model

Lambda's public billing docs split charges into On-Demand Cloud instances, 1-Click Clusters and filesystems. On-Demand Cloud prices instances by hourly usage but bills in one-minute increments, starting after launch and health checks and ending at termination. Running instances accrue charges whether or not they are actively used. 1-Click Clusters are priced per GPU per hour and billed in weekly increments under the reservation terms shown during approval.

Capacity tiers

Lambda exposes several capacity paths. On-Demand Cloud provides Linux GPU virtual machines tied to a region. 1-Click Clusters are production-ready clusters for larger jobs and are handled through a reservation process. Lambda's main site also describes Superclusters and private cloud for single-tenant, large-scale deployments. The public pricing page says reserved capacity is handled through sales, while self-serve Instances are first-come access.

GPU lineup

Lambda's On-Demand Cloud docs list current instance families including B200, GH200, H100 SXM, H100 PCIe, A100 SXM, A100 PCIe, A10, A6000, Tesla V100 and RTX 6000. The pricing page also describes 1-Click Clusters built around NVIDIA HGX B200 and H100 systems. Lambda's homepage highlights newer NVIDIA systems for larger deployments, including GB300, B300, B200 and H200 infrastructure.

Storage, egress and hidden costs

Lambda instances include a root volume whose size depends on the instance type. Persistent filesystems are separate regional stores that must be created in the same region and attached when the instance or cluster launches. Lambda's billing docs say filesystems bill per GiB used per month in one-hour increments, continue billing while they exist even if unmounted, and have no minimum storage period. The same page says Lambda does not charge for ingress or egress.

Reliability and support posture

Lambda does not publish a general public uptime SLA on the On-Demand Cloud docs checked for this profile. It does publish detailed 1-Click Cluster support docs: customers get a Customer Success Manager, Technical Account Manager, Machine Learning Expert and 24/7 Support Engineering through the ticketing portal. The enterprise page says Lambda backs deployments with compliance attestations and supports ML engineering, managed orchestration and production-oriented infrastructure.

Free credits and referral programs

Lambda publicly documents a research-credit offer for qualifying researchers on its research page, with an application link and possible website feature for selected work. I did not find a general self-serve free trial or referral program on Lambda's public pages checked for this profile. The billing page says refunds are issued as service credits for future Lambda Cloud usage, expire after the documented term and cannot be redeemed for cash. See Lambda research.

Who should pick Lambda

Pick Lambda if you want AI-focused datacenter capacity with a clear path from single GPU instances to multi-node clusters. It is strongest for teams that value preinstalled ML software, large local root volumes on many instance types, and cluster support for serious training workloads. It is less ideal if you need a published marketplace-style spot tier or a public general-purpose free trial. For production, confirm reserved terms, support entitlements and filesystem behavior before launching.

FAQ

Profile verified 2026-07-06. Sources: lambda.ai · lambda.ai · docs.lambda.ai · docs.lambda.ai · docs.lambda.ai · docs.lambda.ai · lambda.ai · lambda.ai