Confidential computing extends to GPU workloads, protecting model weights and data in use — critical for on-prem frontier model deployment
NVIDIA has pioneered confidential computing for accelerated computing; Google uses it for Gemini on-prem with Dell, Apple for consumer services. This becomes essential as enterprises run continuous agent workloads on GPUs and need to protect IP and data during inference.
NVIDIA Confidential Computing Enables Zero-Trust On-Prem Deployment of Frontier Models
New confidential computing technology allows encrypted model weights to run on-prem with zero-trust for both model IP and enterprise data, unlocking enterprise adoption of frontier models that previously could only be offered as cloud SaaS.