Cohere, Google Cloud adopt NVIDIA Confidential Bare Metal for secure AI workloads

NVIDIA has announced the expansion of its Confidential Computing portfolio with the addition of Confidential Bare Metal. This new deployment option allows organisations building frontier models to execute sensitive workloads on dedicated third-party infrastructure. The move addresses the need for strong privacy protections while operating across global cloud, neocloud and enterprise environments.

Root cause

NVIDIA states that this architecture enables safe agents across the stack. The company builds tools to enable safe agents across the stack, with the NVIDIA OpenShell Reference Architecture forming part of this portfolio. Confidential Computing serves as the infrastructure layer that protects the models and data those agents depend on while running.

Attack path

This release is not a security incident but an architectural enhancement. It introduces platform security controls designed to protect data in transit and measure platform integrity. The system enforces hardware-level access restrictions, allowing builders to operate bespoke software stacks, including host and platform infrastructure software, while maintaining strict control over their intellectual property and sensitive datasets.

Affected versions

NVIDIA Confidential Computing now offers three distinct deployment options: Confidential Bare Metal for bare-metal infrastructure, Confidential Containers for Kubernetes environments, and Confidential Virtual Machines for virtualized environments. Support extends to NVIDIA Grace Blackwell NVL72, x86-based NVIDIA HGX Rubin NVL8 and NVIDIA DGX Rubin NVL8 systems.

Mitigation

• Verify that your chosen cloud provider supports the specific hardware configurations required for Confidential Bare Metal.
• Ensure your software stack is compatible with the non-virtualized environment and hardware-rooted security features.
• Review your data handling policies to align with the new cryptographic protection capabilities offered by the platform.

What it means for organisations

Teams usually need to assess whether their current trust boundaries align with bare-metal deployment models. Security leads should consult with infrastructure managers to understand the operational implications of managing bespoke software stacks on third-party hardware. Procurement teams must verify that suppliers can guarantee the integrity of the physical infrastructure. Internal governance groups should review data residency requirements, as this option expands geographic deployment possibilities.

What to do and how to stay safe: NVIDIA

• Audit your current AI workload deployment models to identify those requiring enhanced data privacy.
• Engage with your cloud providers to confirm their support for NVIDIA Confidential Computing features.
• Evaluate the compatibility of your existing software stacks with non-virtualized, hardware-isolated environments.
• Update your risk assessments to include the benefits and constraints of bare-metal confidential computing.

General security guidance from the TechSec Observer newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

Which companies are already using Confidential Bare Metal?

Cohere plans to integrate it into Cohere Model Vault, and Google Cloud is extending support to A4X Max instances.

What hardware systems support this new feature?

It supports NVIDIA Grace Blackwell NVL72, x86-based NVIDIA HGX Rubin NVL8 and NVIDIA DGX Rubin NVL8 systems.

How does this differ from previous options?

It provides a bare-metal infrastructure option, distinct from the existing Confidential Containers and Confidential Virtual Machines.