Fortinet has announced a Secure AI Data Center solution, developed in collaboration with Arista Networks, and deployed at Monolithic Power Systems (MPS). Fortinet says this joint offering delivers a validated, modular, and scalable zero-trust architecture designed specifically for AI data centers, prioritizing integrated networking and security for high-performance environments.
The solution builds on Fortinet’s previously introduced Secure AI Data Center framework, now enhanced with multivendor integration. According to Fortinet, key technical features include a modular architecture for deployment flexibility, hyperscale-grade performance for AI training and inference, and zero-touch provisioning—enabling deployments up to 80 percent faster. The design also features support for future AI accelerators without requiring a complete infrastructure redesign.
At the core of this solution, Fortinet reports that HTTPS and TLS encryption is offloaded to its application-specific integrated circuit (ASIC), achieving up to 33-fold performance improvements with sub-single-microsecond latency. This allows server CPUs to dedicate resources to inference workloads and reduces network contention, bottlenecks, and tail latency, improving the processing rate of AI tokens-per-second.
The joint reference architecture combines Arista Networks’ high-performance, ultra-low-latency networking with cluster load balancing for large-scale data center requirements and AI or machine learning clusters. Fortinet’s ASIC-accelerated firewalls provide AI-aware threat protection, zero-trust segmentation, encrypted traffic inspection, and automated incident response. The integrated design is aimed at securing every layer of the AI stack, including compute, storage, data pipelines, and large language model (LLM) workloads, while aiming for high infrastructure uptime and operational efficiency.
Fortinet specifically targets data centers handling compute-intensive AI deployments and identifies critical challenges such as increasing infrastructure costs, vendor lock-in, and security threats like model tampering and data leakage. Fortinet claims the alliance addresses these issues through zero-trust protection across the stack, hardware-accelerated security, open modular architecture, automation for real-time detection and remediation, and unified management of security and networking.
According to Huy Ly, Head of Global Infrastructure/Security at Monolithic Power Systems, “Leveraging the scalable, affordable, and high-performing reference architecture and integrated solution gives us the visibility, performance, and protection we need to operate high-density GPU clusters with confidence.” Ly added, “With this solution integrated into our AI environment, we can safeguard sensitive models and data while maintaining hyperscale throughput with greater efficiency and cost performance.”
Source: Fortinet
