Lambda

GPU capacity at several scales

Lambda builds and operates infrastructure for training and serving AI models. Its cloud documentation covers GPU instances, larger clusters, storage and the software stack needed to run workloads. The company dates its founding to 2012. [2][1]

On November 3, 2025, Lambda announced a multibillion-dollar Microsoft agreement covering tens of thousands of GPUs, including GB300 NVL72 systems. The deployment was a multi-year commitment; the announcement did not disclose a precise total contract value or establish that every contracted GPU was already installed. [1][3]

Lambda's contribution is the compute service and the work of assembling and operating its clusters. A contracted expansion and a currently usable cloud instance describe different stages of capacity, so the scale of this agreement should be read separately from the resources available to an individual customer. [1][2]