The brief

Analyst view

• Q2 2026 total revenue $582M, up 454% YoY; AI Cloud revenue $575M, up 514% YoY. On track for $3B–$3.4B full-year 2026.
• Meta committed $27B over five years ($12B dedicated, $15B on-demand capacity). Microsoft committed $19.4B over five years.
• Vera Rubin deployment ramping H2 2026. Targets 4 GW contracted power by year-end 2026. Trading as NBIS on Nasdaq.

Technical approach

As reported

Non-virtualized GPUs

Dedicated compute without hypervisor overhead

Raw GPU clusters with InfiniBand networking eliminate hypervisor overhead and enable tight coupling for large workloads.

Full AI stack

Training, inference, and MLOps unified

Managed inference, serverless compute, and 24/7 expert support cover the full AI lifecycle from training to production.

Rapid deployment

Clusters operational in minutes

Infrastructure scales from small experiments to global operations with minimal setup time, targeting elastic demand.

Scale-out architecture

Multi-location distributed capacity

Data centers across multiple continents deliver low-latency inference and redundancy for enterprises at global scale.

Show all 4