CoreWeave linked multiple Vera Rubin NVL72 racks into a single scale out system targeting agentic AI workloads. Each individual rack contains 72 Rubin GPUs and 36 Vera CPUs connected by NVLink 6 fabric, while the cluster uses Nvidia Spectrum-X Ethernet to connect hundreds of GPUs. The company previously validated a single rack in June and now utilizes its Mission Control software to handle hardware detection and power for larger clusters. The deployment includes a new Archive tier for object storage and a LOTA accelerator providing 7 GB/s throughput per GPU.

AI products and services revenue exceeded $400 million in annual recurring revenue for the second quarter, with managed inference booked ARR increasing from $1 million to more than $100 million. CoreWeave had 1.5 GW of active power in Q2 and 4.2 GW under contract, aiming for 8 GW by 2030. This expansion requires $35 billion to $39 billion in capital expenditures for 2026, and debt service costs range from $860 million to $940 million in Q3.