NVIDIA Adds Confidential Bare Metal for AI Models on Third-Party Servers

NVIDIA introduced Confidential Bare Metal on Oct. 21, giving AI model developers a way to run sensitive workloads on dedicated third-party servers while retaining control of their models and data.

The option uses hardware-based isolation, checks that verify the computing environment and access restrictions to protect models and data while workloads run.

Google Cloud is adding Confidential Bare Metal support to its A4X Max instances, which use NVIDIA GB300 NVL72 systems. Cohere plans to integrate the option into Cohere Model Vault.

NVIDIA is also expanding confidential computing support to Grace Blackwell NVL72, HGX Rubin NVL8 and DGX Rubin NVL8 systems. Its DSX OS software is set to support the confidential computing options for cloud providers.