NVIDIA Confidential Bare Metal

NVIDIA Confidential Computing option that lets frontier model builders run sensitive workloads on third-party bare-metal GPUs.

NVIDIA · Security and safety tooling · security

Published in Feed

Overview

NVIDIA Confidential Bare Metal, announced October 21, 2026 at GTC Berlin, is a deployment option for NVIDIA Confidential Computing that lets frontier model builders run sensitive workloads on dedicated third-party infrastructure while running their own host and platform software stack. NVIDIA added controls for data in transit, platform integrity measurement and hardware-level access restrictions. Support covers Grace Blackwell NVL72 and x86-based HGX Rubin NVL8 and DGX Rubin NVL8 systems. Google Cloud plans to extend it to A4X Max instances built on GB300 NVL72, and Cohere plans to integrate it into Cohere Model Vault.

This is a sourced directory entry. Features and capabilities are reported in the linked sources; this entry does not document independent Aymorphic testing.

Key features

  • Third deployment option beside Confidential Containers and Confidential Virtual Machines
  • Model builders keep control of host and platform software
  • Added data-in-transit protection, platform integrity measurement and hardware-level access restrictions
  • Supports Grace Blackwell NVL72, HGX Rubin NVL8 and DGX Rubin NVL8
  • Named cloud partners: CoreWeave, Firmus, GMI Cloud, Google Cloud, IREN, Lambda, Microsoft Azure, Nebius, Nscale, Vultr

Use cases

  • Training or serving proprietary frontier models on rented GPU capacity without exposing weights or data to the host
  • Protecting sensitive enterprise data used by agents while they run

Pricing

Unknown; the announcement gives no pricing or performance figures.

Pricing reflects the source record and may change. Confirm current terms with the provider.

Dates and sources

Released
21 Oct 2026
Record updated
27 Oct 2026
Record last checked
27 Oct 2026

Explore the Feed

Other entries filed under security and safety tooling; inclusion does not imply equivalent features.

  • DefenseClaw Edge Connector

    Cisco's open-source DefenseClaw gains an Edge Connector that allows or blocks AI agent tool calls directly on IoT and robot hardware.

  • Lakera Guard 2.0

    Lakera's real-time guardrail API for LLM apps, updated with multi-agent protection and a new benchmark for detecting indirect prompt injection in tool outputs.

  • Fleming-1

    Sierra's model that listens to a live phone call and scores whether the caller is an AI agent rather than a human.

Browse the Feed