Fortinet Brings AI-Ready Data Center Security to NVIDIA DPUs

As enterprises race to operationalise generative AI and modernise data centres, the security conversation is shifting from “how fast can we inspect traffic?” to “how can we secure massive east–west flows without stealing compute from AI workloads?” Fortinet’s latest move—bringing its data centre security capabilities to NVIDIA BlueField DPUs—targets that exact pressure point: delivering high-performance security controls closer to where data moves, while keeping precious CPU and GPU cycles available for applications and AI training or inference.

At a high level, the announcement underscores a broader industry trend: security is becoming infrastructure-native. Instead of running all protections on general-purpose servers, organisations are increasingly offloading networking, storage, and security functions onto specialised accelerators such as DPUs (Data Processing Units). This approach reflects the reality of today’s data centres, where traffic volumes and microservice-to-microservice communications have grown dramatically, and where AI clusters introduce new performance and segmentation requirements.

Why DPUs matter for modern data centre security

DPUs are purpose-built processors designed to handle data movement and infrastructure tasks—often more efficiently than host CPUs. NVIDIA’s BlueField line, for example, is widely positioned as a way to isolate and accelerate networking and security functions in cloud and enterprise data centres. That matters because as AI deployments expand, CPUs are frequently tasked with orchestration and application logic, while GPUs are reserved for compute-heavy model work. Any security stack that competes for those resources can create bottlenecks or raise costs.

By integrating security services with DPUs, Fortinet is aligning with a key operational goal for AI-era infrastructure: lower latency and higher throughput security, with reduced host overhead. This is particularly relevant for large-scale environments where east–west traffic dominates and where segmentation and inspection must keep pace with rapid workload scaling.

What Fortinet is bringing to NVIDIA BlueField

The core idea is to extend Fortinet’s data centre security tooling so that protections can be deployed in conjunction with NVIDIA BlueField DPUs. In practice, this can help organisations enforce security policies closer to workloads and across high-volume traffic paths—supporting the kind of granular control modern architectures require.

While the exact deployment model depends on the environment, the value proposition is consistent: hardware-assisted security that can improve performance while strengthening isolation. This approach is well-suited to:

AI and HPC clusters where throughput and predictable latency are critical
Microservices and container platforms that generate heavy east–west traffic
Multi-tenant or segmented environments that require strong isolation boundaries
Hybrid data centres seeking consistent security policies across mixed infrastructure

The industry context: AI growth, cost pressure, and security architecture change

AI infrastructure investment is rising, but so is scrutiny over cost efficiency. Data centre operators are looking for ways to reduce the “security tax” on compute resources, especially when GPUs and high-core-count CPUs are expensive and often constrained. Offloading infrastructure functions to DPUs can help address this economic reality by improving utilisation and reducing the need for overprovisioning.

At the same time, AI introduces new risk considerations—ranging from protecting sensitive training data to securing model-serving pipelines and preventing lateral movement inside flat networks. Traditional perimeter-centric controls are not enough on their own. The push toward DPU-based security reflects a larger shift toward distributed enforcement, where security is embedded throughout the fabric of the data centre.

What this means for security teams and data centre operators

For practitioners, Fortinet’s support for NVIDIA DPUs signals a practical direction: security controls that can keep up with modern traffic patterns without becoming a performance bottleneck. It also suggests a maturing ecosystem where security vendors and infrastructure providers collaborate to deliver integrated, acceleration-friendly architectures.

Organisations evaluating this kind of approach often focus on a few key outcomes:

Performance: maintaining high throughput inspection without saturating host CPUs
Isolation: stronger separation of infrastructure services from tenant workloads
Scalability: support for rapid growth in east–west traffic and AI cluster expansion
Operational consistency: enforcing policies across data centre fabrics in a repeatable way

Conclusion

Fortinet’s move to bring AI-ready data centre security capabilities to NVIDIA BlueField DPUs reflects where the market is heading: security that scales with AI, delivered through infrastructure acceleration rather than added host overhead. As enterprises build larger AI environments and denser data centre fabrics, solutions that combine strong policy control with high-performance offload will become increasingly attractive—not just for speed, but for economics and resilience. In that landscape, DPU-enabled security is less a niche optimisation and more a blueprint for the next generation of secure, AI-capable data centres.