NVIDIA BlueField Platform

The NVIDIA BlueField platform is an advanced infrastructure computing solution designed to power AI factories, which are purpose-built data centers for AI workloads. It combines powerful computing, high-speed networking, and integrated software-defined hardware accelerators for networking, storage, and security to create a secure, accelerated infrastructure for any workload in any environment.

• Explore the NVIDIA BlueField platform overview for detailed infrastructure capabilities.

The BlueField-3 DPU is a 400 Gb/s infrastructure compute platform designed for software-defined networking, storage, and cybersecurity. The BlueField-4 DPU is an 800 Gb/s infrastructure platform purpose-built for gigascale AI factories, offering higher throughput and advanced capabilities to unlock new levels of performance and efficiency.

• Review the BlueField-3 DPU Datasheet for technical specifications.
• See the BlueField-4 DPU Datasheet to learn about gigascale AI factory performance.

The BlueField-4 STX is a storage processor purpose-built for AI storage platforms. It features the NVIDIA Vera CPU and ConnectX-9 networking to accelerate AI-native data movement, orchestrate data access, and provide in-silicon security directly in the I/O path.

• Download the BlueField-4 STX Storage Processor Datasheet for full product specifications.

An agentic AI factory is a new class of data center designed for agentic AI, where autonomous agents can reason, collaborate, and act at scale. BlueField supports these factories by providing secure, accelerated infrastructure that handles networking, storage, and security, allowing host CPUs to focus on AI model processing.

• Read the NVIDIA AI Factories Validated Design to learn more about infrastructure services.

BlueField enforces zero-trust security at the speed of AI by embedding security features directly into the hardware. This includes real-time threat detection, isolation, and runtime protection for distributed infrastructure, data platforms, workloads, and AI agents.

• Learn about NVIDIA DOCA capabilities for implementing secure, software-defined infrastructure.

NVIDIA DOCA is the software platform that delivers enterprise-ready libraries and microservices for NVIDIA BlueField and ConnectX. It is designed to accelerate, secure, and scale production AI workloads by providing developers with the tools to build infrastructure-aware applications.

• Review the NVIDIA DOCA software platform for library and microservice information.

The BlueField-4 STX improves storage performance by serving as a dedicated storage-side CPU and data-path engine. It uses the NVIDIA Vera CPU to deliver higher throughput for data pipelines, including compression and encryption, and integrates ConnectX-9 networking for accelerated data movement.

• Review the BlueField-4 STX storage benchmarks for performance details.

Yes, BlueField provides security and performance benefits at the edge. It acts as a dedicated security processor within operational technology (OT) systems to enforce zero-trust security and detect threats in real-time, helping protect mission-critical industrial processes.

• Read more about operational technology security with NVIDIA BlueField.

The NVIDIA BlueField-4 DPU delivers up to 800 Gb/s connectivity. This high-bandwidth capability is designed to support the demanding networking requirements of gigascale AI factories.

• Access technical details in the BlueField-4 DPU Datasheet.

BlueField enables high-performance, multi-tenant AI cloud services by offloading networking, storage, and security from host CPUs to the DPU. This improves isolation between tenants, enhances security, and increases overall GPU and CPU utilization for cloud providers.

• Learn about NVIDIA DSX AI infrastructure services.

The NVIDIA Vera CPU in the BlueField-4 STX is optimized for storage processing, delivering significant improvements in throughput for tasks like multi-stage compression, encryption, and data integrity checking, which are critical for AI-native storage systems.

• Explore the BlueField-4 STX storage benchmarks to understand the impact of the Vera CPU.

BlueField supports several industry-standard AI storage technologies, including Remote Direct Memory Access (RDMA), NVMe-over-Fabrics (NVMe-oF), and GPUDirect Storage, while providing support for block, file, and object storage interfaces.

• Learn about the NVIDIA AI-native storage solutions.

An AI Grid is a distributed computing architecture that scales AI across geographically dispersed locations. BlueField enables these grids by providing the secure, accelerated infrastructure needed to share resources across locations with lower latency and higher resilience for generative and physical AI workloads.

• Explore the NVIDIA AI Grid use case for more information.

BlueField acts as an infrastructure controller that integrates with Kubernetes to enable elastic, rapid provisioning of resources across diverse cloud environments, supporting multi-service architectures and service function chaining.

• See the BlueField platform documentation for integration details.

Yes, NVIDIA works with a robust ecosystem of partners, including providers of cloud, storage, cybersecurity, and infrastructure services.

• View the list of NVIDIA BlueField ecosystem partners.

The BlueField-4 STX serves as a storage-side CPU and data-path engine that orchestrates data access, manages metadata processing, and applies secure storage services directly within the I/O path, which accelerates data movement for AI agents.

• Review the BlueField-4 STX Storage Processor Datasheet.

BlueField offloads security tasks from the host to the DPU, allowing for real-time threat detection and enforcement of security policies (such as zero-trust file access) at the hardware level, protecting the system without impacting performance.

• Read about in-silicon security capabilities.

Elastic AI infrastructure refers to the ability of the BlueField platform to dynamically manage and provision infrastructure services, such as networking and security, allowing AI clusters to grow or shrink based on workload demands while maintaining predictable performance.

• Explore NVIDIA BlueField infrastructure services.

Yes, BlueField powers the infrastructure services that keep on-premises AI factories running by offloading networking, storage, and security, ensuring that enterprise workloads get predictable performance and secure isolation.

• Learn more about NVIDIA AI factory validated designs.