MLOps & Infrastructure · advanced · concept 225 of 311 · Updated Nov 4, 2026
Confidential Computing
← Numerical Precision & Mixed-Precision TrainingGradient Checkpointing and Memory-Efficient Training →
Confidential computing protects data while it is being processed, using hardware-isolated environments that even the cloud provider cannot read. For AI this means sensitive prompts, documents and model weights can stay protected during inference, not only in storage or in transit. It supports regulated uses of cloud AI, at some cost in performance and setup effort.
Key terms
Trusted execution environmentAttestationEncryption in useHardware isolationProtecting model weights
Learn these first
Where you meet it in the real world
AI on health and financial data, protecting proprietary models, multi-party analysis
Guides and articles
Confidential computing ↗
NVIDIA
What is confidential computing? ↗
Confidential Computing Consortium