Lambda

Lambda is a GPU cloud for training and running AI models. It rents out NVIDIA hardware, from single self serve instances to large reserved clusters, and describes itself as the Superintelligence Cloud. The job it does is to give AI teams compute without building their own data center. It serves machine learning engineers, research groups and enterprises. There are three ways to buy capacity. Instances are on demand GPUs you launch in minutes. 1 Click Clusters are production ready clusters from 16 to more than 2,000 NVIDIA B200 or H100 GPUs. Superclusters are larger private deployments, with GB300 NVL72 and VR200 NVL72 systems named on the site. The company states that it holds SOC 2 Type II compliance and offers single tenant options. Pricing is published per GPU per hour in US dollars, plus applicable sales tax. For on demand instances the 8 GPU configurations list B200 SXM6 at $6.69, H100 SXM at $3.99, A100 SXM 80 GB at $2.79, A100 SXM 40 GB at $1.99 and Tesla V100 at $0.79. Smaller configurations cost slightly more per GPU, for example an H100 SXM in a 1 GPU instance is $4.29, and other options include GH200 at $2.29, A10 at $1.29 and A6000 at $1.09. 1 Click Cluster terms of 2 weeks to 1 year start at $9.86 per GPU hour for B200 at 16 GPUs and $6.16 for H100, with lower rates at larger sizes. Longer reservations are quoted through sales. The home page contains an on page agent terminal widget that displays a manifest of the page. It is descriptive only, and nothing in it was followed.