ArcOS running natively on NVIDIA BlueField-4 offloads and accelerates network and security functions for AI Clouds
Arrcus, the hyperscale networking software company and a leader in core, edge, and multi-cloud routing and switching infrastructure, today announced support for the newly announced NVIDIA BlueField-4 platform, unveiled at GTC Washington, D.C. Designed to meet the demands of gigascale AI factories, the combination of Arrcus ArcOS® and NVIDIA BlueField-4 DPU will deliver new levels of acceleration, efficiency, and security for AI training, inference, and cloud-scale workloads.
BlueField-4 introduces 6x the compute power of BlueField-3 and 800 Gb/s networking speeds, unlocking powerful architectural, performance, and scale benefits while maintaining backwards compatibility. The platform integrates the NVIDIA Grace CPU and NVIDIA ConnectX-9 networking, providing advanced infrastructure acceleration and cybersecurity.
ArcOS Advantage
Arrcus ArcOS is built to run natively on NVIDIA BlueField-4, enabling customers to offload resource-intensive network functionsincludingIPSec, NAT, Routing, EVPN/VXLAN, and ACLs. This offload capability increases throughput and efficiency, freeing up CPU cycles for AI workloads while ensuring secure multi-tenant connectivity and high-performance fabric orchestration.
Together, ArcOS and BlueField-4:
- Deliver accelerated performance across AI fabrics for training and inference workloads
- Extend AI fabric visibility, automation, and secure connectivity
- Open new monetization opportunities for NeoClouds, including Inference-as-a-Service
- Enable composable, cloud-ready infrastructure with NVIDIA DOCA microservices that deliver networking, data acceleration, and cybersecurity at scale
- Empower NeoCloud providers to unlock new monetization models, including Inference-as-a-Service, by delivering a highly efficient and optimized network fabric purpose-built for training and inference at gigascale
Together, Arrcus and NVIDIA are redefining the foundation of gigascale AI infrastructure. By pairing the flexibility of ArcOS with the acceleration of BlueField-4, customers gain a secure, high-performance fabric designed to scale effortlessly. This collaboration not only addresses today’s AI networking challenges but also lays the groundwork for future innovation in distributed AI, training, inference, and AI-as-a-Service.