AI factories continue to grow at unprecedented scale, processing structured, unstructured and emerging AI-native data. With demand for trillion-token workloads exploding, a new class of infrastructure is required to keep pace.
At NVIDIA GTC Washington, D.C, NVIDIA revealed the NVIDIA BlueField-4 data processing unit, part of the full-stack BlueField platform that accelerates gigascale AI infrastructure, delivering massive computing performance, supporting 800Gb/s of throughput and enabling high-performance inference processing.
Powered by software-defined acceleration across AI data storage, networking and security, NVIDIA BlueField-4 transforms data centers into secure, intelligent AI infrastructure — designed to accelerate every workload, in every AI factory. It’s purpose-built as the end-to-end engine for a new class of AI storage platforms, bringing AI data storage acceleration to the foundation of AI data pipelines for efficient data processing and breakthrough performance at scale.
NVIDIA BlueField-4 combines an NVIDIA Grace CPU and NVIDIA ConnectX-9 networking to deliver 6x the compute power and support AI factories up to 4x larger than possible with NVIDIA BlueField-3, accelerating gigascale AI infrastructure.
BlueField-4 also features multi-tenant networking, rapid data access, AI runtime security and cloud elasticity with native support for NVIDIA DOCA microservices — containerized services that secure, scale and simplify AI deployment and operations.
The NVIDIA BlueField-4 platform supports a multiservice architecture with native service function chaining, enabling seamless integration and management of multiple network, security and storage services within a single, unified framework.
Security is anchored by the NVIDIA BlueField Advanced Secure Trusted Resource Architecture, enabling service providers to create secure, bare-metal compute instances with zero-trust tenant isolation and full software-defined infrastructure control.
The NVIDIA BlueField platform delivers consistent performance, efficiency and security across the entire NVIDIA AI infrastructure portfolio, from computing systems like NVIDIA RTX PRO Servers, NVIDIA HGX, NVIDIA DGX, NVIDIA GB200 and NVIDIA GB300 systems to data-center-scale designs including the NVIDIA Enterprise AI Factory validated design built on NVIDIA Enterprise Reference Architectures, the NVIDIA AI Data Platform reference design, NVIDIA DGX SuperPOD and the NVIDIA Cloud Partner reference architecture.
NVIDIA ConnectX-9 SuperNICs deliver ultralow-latency, 800Gb/s networking to maximize efficiency in gigascale AI infrastructure. Built for NVIDIA Spectrum-X Ethernet, ConnectX-9 SuperNICs accelerate data movement, optimize RoCE performance, and enable consistent and predictable networking for the most demanding AI workloads.
Growing Ecosystem Adopts NVIDIA BlueField, Powered by NVIDIA DOCA
AI infrastructure and cybersecurity pioneers will continue their track record of wide NVIDIA BlueField adoption and use BlueField-4 to build the next generation of accelerated, secure and software-defined data center platforms. Through DOCA, applications accelerated on today’s BlueField platforms will run seamlessly on BlueField-4, delivering substantial performance gains.
Server and storage leaders including Cisco, DDN, Dell Technologies, HPE, IBM, Lenovo, Supermicro, VAST Data and WEKA are building next-generation servers and AI storage platforms with BlueField — with planned adoption of BlueField-4 — to enable secure and efficient data movement at scale.
Cybersecurity leaders including Armis, Check Point, Cisco , F5, Forescout, Palo Alto Networks and Trend Micro are building new solutions with the NVIDIA BlueField platform and planning to integrate NVIDIA BlueField-4 to deliver zero-trust, AI runtime security and real-time threat protection across service provider and enterprise environments.
Cloud and AI providers such as Akamai, CoreWeave, Crusoe, Lambda, Oracle Cloud Infrastructure, Together.ai and xAI are building on NVIDIA DOCA microservices to accelerate multi-tenant networking, speed data movement and enhance security in AI factories and supercomputing clouds.
Cloud-native and infrastructure software platform providers Canonical, Mirantis, Nutanix, Rafay, Red Hat, Spectro Cloud and SUSE are looking to integrate BlueField-4 to automate networking, storage and security for AI-ready clouds and data centers.
Global systems integrators Accenture, Deloitte and World Wide Technology are working with NVIDIA to bring BlueField-powered AI infrastructure and cybersecurity services to enterprises, governments and manufacturers worldwide, as well as preparing to offer BlueField-4 when it arrives.
NVIDIA BlueField-4 is expected to launch in early availability as part of NVIDIA Vera Rubin platforms in 2026.
Watch the GTC Washington, D.C., keynote from NVIDIA founder and CEO Jensen Huang and explore sessions.