Cisco and NVIDIA are partnering to bring hardware-accelerated firewall enforcement directly into AI servers using NVIDIA BlueField DPUs. By extending Cisco Hybrid Mesh Firewall to run on DPUs rather than host CPUs, the solution delivers 400G line-rate stateful segmentation without consuming GPU or CPU resources needed for AI workloads. Cisco Nexus One orchestrates network policy and collects Kubernetes workload metadata to generate application-aware segmentation rules enforced inline on the DPU. The architecture addresses multi-tenant AI environments where front-end network traffic requires strict inspection and lateral movement containment. Controlled availability is planned for Q3 CY26 with general availability in Q4 CY26.
Table of contents
The front-end network: The real security domainBringing security to the AI workload with DPUs“AI is transforming every industry, and the rapid rise of AI factories is driving a growing need for cybersecurity at scale across enterprise infrastructure. By embedding Cisco’s Hybrid Mesh Firewall policy into NVIDIA BlueField DPUs on AI servers, our joint customers achieve high-performance, multi-tenant, intent-driven enforcement and hardware-accelerated protection, seamlessly connected via Cisco Nexus One AI front-end fabrics.”Cisco Nexus One: Network policy orchestration and visibility for AI front-end fabricsBuilding the secure AI factory of the futureLearn more about how Cisco Hybrid Mesh Firewall can secure your distributed environmentSort: