NVIDIA Eyes Software Revenue With a $2,000 per GPU Yearly Figure From Huang - TechPowerUp
NVIDIA CEO Jensen Huang has revealed the company aims to generate approximately $2,000 per GPU annually from software licensing as part of its AI business strategy.
During a Q&A session at GTC Berlin last week, Huang emphasized that software licensing will become an increasingly significant part of NVIDIA's business operations. The $2,000 figure represents the estimated annual cost to run all NVIDIA software that powers agents in AI factories, though Huang presented it as an example rather than a formal pricing structure.
Pricing Confusion
The announcement has created some confusion as NVIDIA already offers its AI Enterprise suite on a per-GPU basis. According to the company's licensing guide, the current pricing stands at $4,500 per GPU for a one-year subscription with support, or $18,000 per GPU for a five-year commitment. It remains unclear whether Huang's $2,000 figure encompasses the same software suite, a specialized agent-focused bundle, or represents a future licensing model.
OpenShell Architecture
For scale, every million GPUs licensed at $2,000 would bring in $2 billion a year, so it's easy to see why NVIDIA wants recurring software revenue on top of hardware it is already scrambling to expand.
At the same event, NVIDIA unveiled its OpenShell Reference Architecture, which integrates Vera Rubin systems for inference with Vera CPUs for running agents, and BlueField-4 DPUs utilizing NVIDIA Sentry to monitor agent activities. OpenShell functions as an open-source runtime that creates secure environments for agents, with Huang highlighting its importance for AI safety concerns.
Related
