NVIDIA BlueField-4 Powers New Scale-In Network Infrastructure for Agentic AI Factories

Scale-In evolves north-south networks into a coordinated infrastructure domain for the AI factory. Powered by NVIDIA BlueField-4 and NVIDIA DOCA and connected over NVIDIA Spectrum-X Ethernet, Scale-In accelerates the services that secure the full AI stack and move application, data, and storage traffic across the AI factory. Dedicated, host-independent processing keeps these infrastructure services off host CPUs, helping prevent security, data access, and operations from becoming bottlenecks as AI compute scales. The result is a more secure and efficient shared AI infrastructure with consistent access to AI services and data as demand grows.

This post explores the BlueField-4 architecture and explains how Scale-In delivers secure access, high-performance data movement, tenant isolation, simpler operations, and more predictable performance for agentic AI factories.

Scale-In accelerates north-south AI factory infrastructure[](#scale-in_accelerates_north-south_ai_factory_infrastructure)

AI factories have distinct infrastructure requirements at different scales:

• Scale-Up: NVIDIA NVLink unites GPUs as a coherent accelerator.

• Scale-Out: NVIDIA Spectrum-X Ethernet and NVIDIA Quantum InfiniBand connect servers across an AI factory.

• Scale-Across: NVIDIA Spectrum-XGS Ethernet connects distributed AI factories.

• Context Memory: NVIDIA CMX, built on the NVIDIA STX modular foundation for AI-native storage, provides shared KV-cache storage within the factory.

• Scale-In: NVIDIA BlueField-4, NVIDIA DOCA, and NVIDIA Spectrum-X Ethernet accelerate the access, security, data movement, and infrastructure operations surrounding AI compute.

North-south networks provide the access path into and out of a data center, connecting users, applications, data sources, storage systems, and services to individual systems. Traditional cloud data centers built these networks around software-defined infrastructure, composability, and elasticity, so resources and access could be provisioned and scaled as demand changed.

Agentic AI raises the demand on this infrastructure. Software-defined networking, composability, and elasticity remain essential, but they are no longer sufficient. AI factories bring together massive accelerated compute with growing numbers of users, agents, applications, enterprise data sources, and storage systems, all interacting continuously and at scale.

The infrastructure must be accelerated and co-designed as part of the AI factory. Security, multi-tenant networking, data and storage access, and infrastructure operations can’t rely solely on software running on general-purpose host CPUs or operate as independently managed layers. These functions must work together as a unified infrastructure domain, giving operators consistent control over access, data movement, security, provisioning, and observability across the AI factory.

Scale-In addresses these requirements by evolving north-south access into a unified, accelerated infrastructure domain across the AI factory. BlueField-4 provides host-independent acceleration and offload, DOCA provides a unified model for programming and operating infrastructure services, and Spectrum-X Ethernet provides high-performance Ethernet connectivity across the Scale-In access path, including external storage and data sources. Together, they give operators consistent control over access, data movement, security, provisioning, and observability across the infrastructure that supports accelerated compute and data storage.

Figure 1. Scale-Up, Scale-Out, and Scale-Across expand compute connectivity, and Scale-In connects, secures, provisions, and observes the infrastructure surrounding the compute domain