Fortinet has announced a partnership with Nvidia that aims to strengthen security for artificial intelligence workloads running on GPUs. The collaboration, which brings together Fortinet's cybersecurity expertise and Nvidia's GPU technology, is expected to produce new safeguards for AI systems deployed in cloud environments.
Why the partnership matters
As companies race to adopt AI, the infrastructure that powers those models—often GPU clusters in the cloud—has become a target for attackers. Fortinet and Nvidia say their joint effort will focus on hardening those systems against threats like model theft, data poisoning, and adversarial attacks. The companies haven't disclosed specific products yet, but they described the initiative as a step toward setting industry-wide security standards for AI workloads.
What's in the deal
Under the agreement, Fortinet will integrate its security tools with Nvidia's AI Enterprise software and Morpheus cybersecurity framework. The goal is to give cloud operators real-time visibility into GPU-accelerated workloads and automatically block malicious activity without slowing down AI processing. Fortinet's CEO said in a statement that the partnership will help organizations run AI with confidence, while Nvidia's vice president of AI software highlighted the need for security designed specifically for AI infrastructure.
Cloud implications
The partnership arrives as major cloud providers—Amazon Web Services, Microsoft Azure, and Google Cloud—expand their GPU-as-a-service offerings. Customers renting Nvidia H100 or A100 instances often have limited insight into how their AI models are protected at the hardware level. Fortinet and Nvidia aim to close that gap by embedding security directly into the GPU orchestration layer. Early tests reportedly showed a 40% reduction in detection time for AI-specific attacks compared to traditional firewall-based approaches.
Neither company has announced a release date for the integrated solution. They said technical previews will begin in the second quarter of 2025, with general availability expected later in the year.
