NetApp and Nvidia Redesign Storage Architecture for AI Factories

As artificial intelligence (AI) adoption accelerates, traditional storage architecture is being rewritten to meet the unique demands of AI factories. NetApp and Nvidia are collaborating on a new infrastructure designed to handle the combined heavy data movement and transactional metadata activity that comes with AI workloads.

NetApp's Novus architecture separates data and metadata functions, allowing them to scale independently based on different workload demands. This approach prevents smaller transactional operations from competing with large data transfers for resources, ensuring efficient use of GPUs and reducing storage delays.

The shift in storage architecture is also driven by the increasing complexity of AI infrastructure, which must be able to adapt to changing workloads without requiring significant redesign. Agentic workloads, where thousands of agents access data simultaneously, add an additional layer of concurrency requirements that traditional storage systems struggle to meet.