Artificial intelligence has moved from an experimental tool to the main engine of market leadership, pushing enterprises to rethink how they build, scale, and protect their computing environments.
As companies shift away from general-purpose clusters toward purpose-built AI factories, one challenge has become impossible to ignore: how to secure these high-performance systems without slowing them down.
Palo Alto Networks has taken a major step in addressing that challenge by announcing that its Prisma AIRS platform, accelerated on NVIDIA BlueField data processing units, is now part of the NVIDIA Enterprise AI Factory validated design.
The integration embeds zero trust security directly into the AI infrastructure itself, protecting data, models, and workloads while preserving the performance AI factories are built for.
By deploying Prisma AIRS Network Intercept directly on the NVIDIA BlueField DPU and extending protections across hybrid and cloud environments, the solution establishes a zero trust governance fabric that spans the full AI lifecycle.
Security processing is offloaded to an isolated domain on the DPU, allowing host CPUs and GPUs to focus entirely on AI workloads while policies are enforced at line speed using NVIDIA DOCA acceleration.
The architecture also taps into real-time workload telemetry through DOCA Argus, feeding intelligence into Cortex XSIAM and enabling automated, AI-driven responses via Cortex XSOAR. The result is a security model that operates continuously and invisibly, aligned with the scale and speed of modern AI operations.
According to Rich Campagna, senior vice president of product management at Palo Alto Networks, the shift elevates AI security to the highest levels of enterprise decision-making. He said the AI factory has become the new engine of value creation, making its protection a board-level imperative, and described the validated design as a new security architecture built specifically for the AI era—one that embeds trust directly into the infrastructure so organizations can protect proprietary intelligence and deploy AI with confidence.
From NVIDIA’s perspective, security must evolve alongside AI itself. Kevin Deierling, senior vice president of networking at NVIDIA, noted that as AI transforms every industry, protection has to be distributed and embedded within the infrastructure to scale effectively.
Running Prisma AIRS on NVIDIA BlueField, he said, delivers robust runtime security for AI factories without compromising performance or efficiency.
The design goes beyond distributed protection at the workload level. Enterprises are encouraged to maintain centralized hyperscale security firewall clusters at AI factory ingress and egress points, creating a defense-in-depth posture.
Individual workloads can selectively route traffic through these clusters to inspect application-layer threats and prevent lateral movement, while elastic scaling ensures high availability and session resiliency for mission-critical AI operations.
This approach also reshapes the economics of AI infrastructure. By isolating security functions on the BlueField DPU, organizations can dedicate 100 percent of host compute resources to AI tasks, eliminating resource contention and maximizing token throughput.
The result is a lower total cost of ownership and higher capital efficiency, especially as AI models and workloads continue to grow in size and complexity.
Perhaps most importantly, the validated design offers a clear upgrade path. Built today on high-performance BlueField-3, the architecture is designed to transition seamlessly to next-generation BlueField-4, unlocking up to six times the compute power and doubling bandwidth as future demands arrive.
That forward compatibility allows enterprises to invest now while preparing for gigascale AI requirements ahead.
With Prisma AIRS integrated into the NVIDIA Enterprise AI Factory validated design, enterprises gain a comprehensive security blueprint that protects sensitive data end to end and defends models against emerging threats such as prompt injection and adversarial attacks. The collaboration signals a broader shift in how AI infrastructure is built and secured, giving leaders the confidence to innovate faster, scale smarter, and deploy AI boldly in an increasingly competitive landscape.
