← Back
NVIDIA Blog

NVIDIA Confidential Computing Expands, Enabling Frontier Model Builders to Securely Scale Across Global Infrastructure

•4 min read•
#deployment#compute#nvidia
NVIDIA Confidential Computing Expands, Enabling Frontier Model Builders to Securely Scale Across Global Infrastructure
Level:Advanced
For:AI Engineers
✦TL;DR

NVIDIA announced the Confidential Bare Metal deployment option for its Confidential Computing stack, allowing frontier model builders to run sensitive workloads on dedicated third‑party bare‑metal infrastructure while retaining full control of models and data. The expanded security architecture now includes three deployment modes—Confidential Bare Metal, Confidential Containers, and Confidential Virtual Machines—and integrates with NVIDIA Grace Blackwell, Rubin systems, and the DSX OS reference architecture. This enables global scaling of training and inference with hardware‑rooted isolation, as demonstrated by Cohere’s Model Vault integration and Google Cloud’s A4X Max instances powered by NVIDIA GB300 NVL72 systems. Engineers can now match deployment choices to trust boundaries, achieving end‑to‑end cryptographic protection across memory and interconnects.

⚡ Key Takeaways

  • Confidential Bare Metal adds a bare‑metal deployment option to NVIDIA Confidential Computing.
  • The platform now supports three deployment options: Confidential Bare Metal, Confidential Containers, Confidential Virtual Machines.
  • Integration with NVIDIA Grace Blackwell, Rubin systems, and DSX OS provides a reference architecture for AI factories.
  • Cohere will embed Confidential Bare Metal in its Model Vault to protect models and data in GPU memory and interconnects.
  • Google Cloud’s A4X Max instances (NVIDIA GB300 NVL72) will offer non‑virtualized, cryptographically protected compute for training and inference.
💡 Why It Matters

By offering hardware‑rooted isolation on bare‑metal servers, engineers can securely scale high‑value AI workloads across sovereign and global infrastructures without sacrificing performance, simplifying compliance for sensitive model training and serving.

✅ Practical Steps

  1. Deploy workloads on Confidential Bare Metal instances to retain full control of the software stack and enforce hardware‑level access restrictions.
  2. Integrate NVIDIA Confidential Computing into your model serving pipeline via the reference architecture with DSX OS for cloud providers building AI factories.
  3. Leverage Cohere Model Vault or Google Cloud A4X Max instances (NVIDIA GB300 NVL72) to extend hardware‑rooted security directly to the GPU for inference.

Want the full story? Read the original article.

Read on NVIDIA Blog ↗

More like this

Introducing workload recovery and expanded infrastructure checks on Amazon SageMaker HyperPod

AWS ML Blog•#deployment

Introducing Unity Gateway Insights

Databricks Blog•#deployment

Introducing Gemini 4 Neon

Google DeepMind•#llm

DaisyRose-27B: Pushing the Frontier of Post-Training Efficiency

Hugging Face Blog•#llm

EXPLORE AI NEWS

Daily hand-picked stories on LLMs, RAG, agents and production AI — curated for engineers who ship.

BROWSE NEWS

GET THE WEEKLY DIGEST

Join engineers getting the Monday signal-over-noise AI breakdown. No spam, unsubscribe anytime.

LEARN AI ENGINEERING

Curated courses, research papers, repos and tutorials built for engineers leveling up in AI.

START LEARNING