Lambda

What it is for

Lambda is a self-serve GPU instance cloud: pick an H100 / B200 / A100 box, SSH or open Jupyter, and pay per GPU-hour. Use it when the job is a published on-demand instance, not when you want the cheapest marketplace bid.

It is not open source. Billing is per GPU-hour (plus tax). 1× H100 SXM is $4.29/GPU/hr; 8× H100 SXM is $3.99/GPU/hr.

What makes it unique: the cover is Lambda’s official Instances picker — GPU count toggle (8× / 4× / 2× / 1×) and the same $ / GPU / hr you checkout.

Key features

• On-Demand Cloud instances — launch a published GPU VM from the console; SSH in, or click Launch on the row for the preinstalled JupyterLab (On-Demand; Instances).
• 1-Click Clusters — reserve a 16–512 GPU cluster (H100 / B200 SXM) on InfiniBand; compute nodes sit on an isolated private network and you jump through three public head nodes (1-Click Clusters; official launch video).
• Filesystems — attach a regional NFS store at launch (/lambda/nfs/<name>); it survives a terminated instance and can be reused on the next box or 1CC in that region (Filesystems).
• Firewall rules — inbound TCP / UDP from the console Firewall page; default open port is 22 (SSH) (On-Demand).
• Cloud API — create, restart, list, and terminate instances (and filesystems) without the console (Cloud API).

Who it's for

Best fit: research and training teams who want a known SKU and will pay the list GPU-hour.

Poor fit: people hunting the lowest interruptible bid (Vast.ai).

Pricing

Verified 2026-10-22 (Asia/Shanghai) from lambda.ai/pricing. Figures are per GPU per hour, plus sales tax / VAT / GST. Re-check the 8× / 1× toggle.

On-demand instances (examples)

• 8× H100 SXM (80 GB) — $3.99 / GPU / hr
• 1× H100 SXM — $4.29 / GPU / hr; 1× H100 PCIe — $3.29 / GPU / hr
• 8× A100 SXM 80 GB — $2.79 / GPU / hr
• 1× GH200 — $2.29 / GPU / hr
• 8× B200 SXM6 — $6.69 / GPU / hr; 1× B200 — $6.99 / GPU / hr
• 1× A6000 — $1.09 / GPU / hr; 8× V100 — $0.79 / GPU / hr
• 1× Quadro RTX 6000 — $0.69 / GPU / hr

1-Click Clusters (term, talk-to-sales) — 16× H100 $6.16 / GPU / hr; 16× B200 $9.86 / GPU / hr (2 weeks–1 year).

There is no public $0 instance. Contact sales for reserved capacity.

Limitations / costs

• On-demand is first-come. A 1× H100 row can be sold out.
• 1-Click Cluster dollars are quotes on a term, not a cancel-anytime Pod.
• Crypto mining is banned on ODC (docs warning).
• Not a marketplace. For bids, see Vast.ai.

Alternatives

• Runpod — per-second Pods; H100 PCIe $2.89/hr on the Oct 8, 2026 card.
• Vast.ai — live H100 SXM from $1.73/hr on the public feed.
• Modal — functions if you do not want a VM.

Official tutorial

Official Lambda channel video: How to launch a Lambda 1-Click Cluster (oEmbed author: Lambda).

Written path: Pricing plus On-Demand Cloud.

Website

https://lambda.ai

Source

Catalog row curated for AI Tools Nav. Cover is Lambda’s official Instances price picker. 8× H100 SXM $3.99/GPU/hr; 1× H100 SXM $4.29. Tutorial: 1-Click Cluster on @lambda-ai (oEmbed).

Last verified

2026-10-22 (Asia/Shanghai) editorial check.