Artificial intelligence is no longer experimental—it is now the primary engine of enterprise growth and market leadership. As organizations race to operationalize AI at scale, many are moving away from general-purpose computing toward AI factories, purpose-built infrastructures that manage the full AI lifecycle. With this shift comes a critical requirement: enterprise-grade security that does not compromise performance.
To address this challenge, Palo Alto Networks announced that Prisma® AIRS™, accelerated on the NVIDIABlueField DPU, is now officially part of the NVIDIA Enterprise AI Factory validated design—marking a major milestone in secure, scalable AI infrastructure.
A New Security Architecture for AI Factories
The inclusion of Palo Alto Networks Prisma AIRS within the NVIDIA Enterprise AI Factory enables enterprises to embed zero-trust security directly into AI infrastructure. Instead of layering security on top of AI workloads, Prisma AIRS integrates protection at the infrastructure level, ensuring continuous defense without affecting AI performance or efficiency.
By deploying Prisma AIRS Network Intercept directly on NVIDIA BlueField and extending protection into the cloud, organizations gain a unified zero trust governance fabric for the AI factory. This allows enterprises to accelerate innovation while maintaining strict control over sensitive data, models, and workloads.
Performance Without Compromise Using NVIDIA BlueField
One of the biggest barriers to secure AI adoption is performance degradation. This architecture removes that concern by offloading security processing to an isolated domain on the NVIDIA BlueField DPU. Hardware acceleration via NVIDIA DOCA enables security policies to be enforced at line speed, while real-time workload intelligence captured through DOCA Argus feeds directly into Cortex XSIAM® for AI-driven threat detection and response using Cortex XSOAR®.
As a result, enterprises can dedicate 100% of host compute resources to AI workloads, eliminating resource contention and maximizing token throughput, infrastructure efficiency, and return on investment.
Defense-in-Depth for Enterprise AI Operations
Beyond distributed runtime protection, this validated design also emphasizes the importance of a centralized Hyperscale Security Firewall (HSF) cluster at AI factory ingress and egress points. This layered approach supports advanced threat detection, lateral movement prevention, and elastic scalability—delivering the high availability required for mission-critical AI operations.
By combining distributed DPU-based security with centralized hyperscale firewalls, organizations achieve a robust defense-in-depth strategy that aligns with modern AI risk profiles.
Future-Proofing the Enterprise AI Factory
The NVIDIA Enterprise AI Factory validated design is built for immediate deployment and long-term growth. Organizations can transition from general-purpose clusters to secure AI factories without costly infrastructure overhauls. More importantly, the architecture is forward-compatible—ready for next-generation NVIDIA BlueField-4, which will deliver up to 6x compute power and double the bandwidth.
This ensures AI factories remain resilient and scalable as AI demands move toward gigascale workloads.
Securing Innovation with Confidence
With Palo Alto Networks Prisma AIRS embedded into the NVIDIA Enterprise AI Factory, enterprises gain end-to-end protection across the AI stack. Proprietary data stays secure, models are protected from adversarial threats such as prompt injection, and AI innovation can move forward without hesitation.
This collaboration delivers a future-proof blueprint for enterprises seeking to lead in the AI economy—securely, efficiently, and at scale.
Industry leaders can learn more about this architecture at NVIDIA GTC 2026, happening March 16–19 in San Jose, California.
