Nvidia Adds Scale-In to AI Factory Architecture, Pushing DPU Security for Agentic AI

Shainer said the move is intended to connect more of the AI factory's resources while extending security controls across the infrastructure. In Nvidia's positioning, the DPU moves beyond a device that secures access to a server and becomes a separate place to monitor and control activity across the factory. Nvidia is expanding the role of BlueField, its DPU, alongside DOCA software and the recently introduced OpenShell agent runtime.

The company argues that an AI system that simply answers a question presents one set of infrastructure requirements. A system in which agents make repeated model calls, retrieve information, access applications and then act presents a more pressing challenge. Fast communication between GPUs remains fundamental. Operators must also control what agents can access, what actions they can take and how their activity affects other workloads.

Nvidia's OpenShell 0.1.0 is an open-source runtime for defining and enforcing which systems and data an agent can access. It combines sandboxed execution, controlled service access, credential management and formal policy analysis, according to the company. Teams can grant agents the capabilities a task requires while OpenShell enforces those permissions outside the workload. OpenShell supports Codex, Claude Code, Pi, Hermes, and future frameworks across enterprise applications, frontier research and physical AI.

Nvidia describes scale-in as complementary to its existing scale-up, scale-out and scale-across architectures rather than a replacement for the networks that connect GPUs. Scale-up uses NVLink to combine GPUs into a larger unit of compute. Scale-out uses InfiniBand or Spectrum-X Ethernet to connect those units into a larger GPU cluster. Scale-across connects multiple AI factories so their compute resources can contribute to larger jobs. Scale-in brings users, models, agents, storage and compute resources into the AI factory with security controls that extend across its infrastructure.

In Shainer's account of earlier AI systems, the back-end network supported communication among compute resources, while the front-end network provided access for clients. Scale-in expands the role of that access infrastructure so it reaches further into the AI factory. He cited concepts originally presented on theCUBE in 2012 by Arista Networks CEO Jayshree Ullal: north-south traffic describes access into and out of the environment, and east-west traffic describes communication within it. Shainer said securing access at the entrance is insufficient for the agentic environment Nvidia sees coming. Controls must also extend across activity inside the factory, including its compute, memory and storage resources.

BlueField-4 is central to the expansion. Shainer described it as connected to the other infrastructures and to ConnectX network interfaces, with responsibility for security across a much broader set of traffic.

The potential customer benefit is stronger control without putting all the associated burden on the systems running applications, such as core GPUs and central processing units like Vera. For Nvidia, the strategic benefit is a larger role in defining how an AI factory operates, according to the report. The company's broader bet is that agentic AI raises the value of infrastructure that can enforce boundaries independently of the agents themselves.

Editor's Summary Nvidia is adding a scale-in category to its AI factory architecture, using BlueField DPUs, DOCA software and the OpenShell runtime to enforce security controls for agentic AI. The approach aims to give operators control over what agents can access and do without shifting that burden to core GPUs and CPUs. The move extends Nvidia's infrastructure role beyond connecting compute to monitoring and controlling activity across the AI factory.