Skip to main content

Confidential Computing

Confidential Computing is a hardware-based security feature that protects data in use, meaning while it is being processed in memory. It achieves this by running computations in encrypted, isolated environments known as Trusted Execution Environments (TEEs). This ensures that data remains secure not only at rest or in transit, but also during runtime.

Feature Not Yet Available (Under Development)

Confidential Computing is not currently available in Hyperstack. This feature is actively being developed and will be offered exclusively to contracted customers.

If this is a feature you are interested in, please contact our sales team to discuss your needs.

Table of Contents​

About Confidential Computing​

Confidential Computing uses Trusted Execution Environments (TEEs), which are hardware-enforced and isolated to protect code and data from unauthorized access—even from privileged system software.

This is especially important for AI workloads where input data may contain personally identifiable information (PII), sensitive business logic, or proprietary intellectual property. The NVIDIA H100 GPU was the first to introduce native support for Confidential Computing, with newer models like the H200 and B200 extending those capabilities.

For more details, see NVIDIA Confidential Computing documentation.

Confidential Computing Availability in Hyperstack​

Confidential Computing is not currently available in Hyperstack. This feature is under active development and will be made available exclusively to contracted customers with dedicated GPU nodes. It will not be available on-demand.

If this is a feature you are interested in, please contact our sales team.

While some GPU models in our infrastructure, such as the H100 PCIe, H100 SXM5, and H200, support Confidential Computing at the hardware level, Hyperstack does not yet offer end-to-end support across the software stack. As a result, the feature is not currently available to customers.

Why Confidential Computing Is Not Yet Available​

Supporting Confidential Computing requires integration across the entire technology stack. It is not something that can be enabled by simply changing a BIOS setting. Even if Secure Memory Encryption is turned on, additional platform-level requirements must be met:

  • Compatible hardware such as H100, H200, or B200 GPUs with matching CPUs and chipsets
  • BIOS and firmware support, including secure boot
  • Operating system and kernel support (not yet in stable Linux kernels)
  • Hypervisor and virtualization support
  • Applications designed for or compatible with TEEs

Without alignment across these layers, systems may appear ready for Confidential Computing but cannot securely process encrypted workloads. In addition, encrypted workloads, particularly in multi-node configurations, can experience up to a 30% performance slowdown due to encryption overhead.

Interim Security Recommendations​

Confidential Computing is on our development roadmap and will be available exclusively to users with dedicated GPU contracts. If your workload depends on this capability, we encourage you to contact our sales team.

Hyperstack encrypts all data stored at rest. If you have elevated security requirements, you can optionally encrypt your data before uploading it or while it's in transit.


Back to top