NVIDIA expands confidential computing for frontier model builders
NVIDIA said on Wednesday it is expanding its confidential computing stack with a new bare-metal deployment option aimed at frontier model builders that need to run sensitive AI workloads on third-party infrastructure while retaining control of models and data.
Evidence and context
In a post published Oct. 21, 2026, NVIDIA said the new option, called Confidential Bare Metal, is part of a broader set of tools designed to let organizations deploy agents at scale while protecting the models and data those agents rely on. The company said the feature adds platform security controls for data in transit, platform-integrity measurement and hardware-level access restrictions, allowing users of bare-metal instances to own and operate their bespoke software stack on NVIDIA accelerated computing infrastructure across third-party cloud, neocloud and enterprise environments.
NVIDIA said its confidential computing portfolio now includes three deployment modes: Confidential Bare Metal for bare-metal infrastructure, Confidential Containers for Kubernetes environments and Confidential Virtual Machines for virtualized environments. The company said the expansion also reaches NVIDIA Grace Blackwell and Rubin systems and will integrate with NVIDIA DSX OS as a reference architecture for cloud providers building AI factories.
The post named several providers and systems that will support or extend the capability. Cohere said it plans to integrate Confidential Bare Metal into Cohere Model Vault, using NVIDIA Confidential Computing trusted execution environments to extend hardware-rooted security directly to the GPU, protecting models and data in memory and across interconnects during execution, according to a quoted statement from Phil Blunsom, Cohere's chief technology officer. Google Cloud said it is extending support for Confidential Bare Metal to A4X Max instances powered by NVIDIA GB300 NVL72 systems, describing the setup as hardware-isolated, non-virtualized compute with end-to-end cryptographic protection, according to a quoted statement from Andrés Lagar-Cavilla, distinguished engineer and AI infrastructure security lead at Google Cloud.
NVIDIA also said support now extends to NVIDIA Grace Blackwell NVL72, x86-based NVIDIA HGX Rubin NVL8 and NVIDIA DGX Rubin NVL8 systems. The company listed CoreWeave, Firmus, GMI Cloud, Google Cloud, IREN, Lambda, Microsoft Azure Infrastructure, Nebius, Nscale and Vultr as supporting NVIDIA Confidential Computing deployments, and said systems partners including Cisco, Dell Technologies, HPE, Lenovo, Supermicro and Bull will support the capabilities with NVIDIA-Certified Systems.
The company described confidential computing as using hardware-rooted isolation and cryptographic attestation to help protect models and data during execution. The source material also notes that deployment support depends on the GPU SKU, the confidential-VM environment and compatible drivers, so the capability should not be assumed across all NVIDIA hardware. NVIDIA's Sept. 22, 2026 technical guidance further said performance should be benchmarked for the intended workload rather than assumed to be unchanged.