Lambda

Lambda is a GPU cloud focused on NVIDIA hardware for AI training and inference, which it now brands the Superintelligence Cloud. Its self-serve product is on-demand GPU instances: pick a 1x, 2x, 4x or 8x configuration of GPUs such as the V100, A10, A6000, A100, H100, GH200 or B200, and the machine comes up in minutes with Lambda Stack (PyTorch, CUDA and drivers) preinstalled. Instances are billed per GPU per hour, charged by the minute, with no egress fees, and the console shows live GPU and VRAM utilisation for each instance. Everything can also be driven through the Lambda Cloud API and CLI.

For larger jobs, 1-Click Clusters provide 16 to 2,000+ interconnected HGX B200 or H100 GPUs, and managed Kubernetes and managed Slurm add orchestration, health checks and automatic node remediation with Grafana dashboards. Above that, Lambda sells dedicated superclusters and private cloud capacity, including NVIDIA GB300 NVL72 and Vera Rubin systems, through its sales team.

In October 2026 Lambda announced support for Confidential Bare Metal, an extension of NVIDIA Confidential Computing to dedicated bare-metal instances, alongside the NVIDIA OpenShell reference architecture for agent safety. It suits ML researchers and teams that want straightforward NVIDIA instances with a ready ML stack, and organisations that need multi-node training capacity.