NVIDIA Fleet Intelligence Generally Available: Free GPU Fleet Monitoring with Open-Source Agent
Tags Infrastructure · Hardware · Enterprise
NVIDIA Fleet Intelligence is now generally available as a free, agent-based managed service providing real-time telemetry, health monitoring, and cryptographic attestation for NVIDIA data center GPU fleets. The host agent is open source (NVIDIA/fleet-intelligence-agent on GitHub). It supports Vera Rubin, Blackwell, and Hopper GPU architectures, with attestation only on Vera Rubin and Blackwell. The service monitors power, temperature, performance, health, and configuration across GPU and CPU fleets in near real-time.
Technical significance
Fleet Intelligence addresses a growing pain point as GPU fleets scale to thousands of chips across multiple data centers. The open-source agent approach builds trust through auditability, and the cryptographic attestation capability is particularly relevant for multi-tenant GPU clouds where customers need assurance about GPU integrity. By offering this free, NVIDIA is building lock-in at the infrastructure management layer.