Confidential Computing on CPU and GPU Systems: How AI Data Centers Protect Data in Use
Confidential computing protects sensitive data and AI models in use within data centers.
Confidential computing addresses the challenge of protecting data while it's being processed in memory, a critical gap in traditional encryption methods. This technology utilizes Trusted Execution Environments (TEEs) on CPUs and, more recently, GPUs, to isolate sensitive computations and data from the underlying infrastructure. The integration of confidential computing directly onto GPUs, like NVIDIA's Hopper generation, is a significant advancement for AI data centers, ensuring that model weights and data remain protected during intensive computations.
AI summary of the linked source. Not reviewed by an editor. How this works
Run NVIDIA?
Get an email when a new story names NVIDIA, usually within the hour.
Free. Your list stays private and never appears in a subject line. One click stops it. How Vendor Watch worksPrivacy
The lesson from this incident
Full lessonTraditional encryption secures data at rest and in transit, but leaves data exposed in plaintext while being processed in memory — a critical vulnerability in AI data centers handling sensitive model weights and proprietary datasets. Confidential computing addresses this gap by leveraging Trusted Execution Environments (TEEs) to isolate sensitive computations from the host infrastructure, hypervisors, and even privileged administrators. The extension of TEEs to GPUs, such as NVIDIA's Hopper architecture, is particularly significant as AI workloads increasingly rely on GPU processing for large-scale model training and inference. Without these protections, any compromise of the underlying infrastructure — whether by a malicious insider, cloud provider breach, or supply chain attack — could expose the most sensitive data at its most vulnerable moment. Organizations running AI workloads must treat data-in-use protection as a first-class security requirement, not an afterthought.
What helps against this
How we list vendors- Encryption and key management
Stolen data stays unreadable, and the keys stay under your control.