NetApp introduces Novus, a storage architecture targeting zettabyte-scale capacity with 100 TB/second aggregate bandwidth.

NetApp this week rolled out Novus, a new storage architecture engineered to handle zettabytes of data and deliver 100 TB/second of aggregate bandwidth over NFS and parallel NFS—sufficient to sustain hundreds of thousands of GPUs productively.

"AI Factory math is unforgiving," writes Arindam Banerjee, NetApp's chief platform and technology officer, in a recent blog post. "A single GPU can demand as much as 2GB/s throughput to stay busy, and sustaining this across 50,000 GPUs requires 100TB/s cumulative bandwidth. Traditional storage arrays give you 40 to 80GB/s. So reaching 100TB/s with conventional arrays means deploying more than a hundred of them, each with its own namespace, failure domain, and management overhead."

This situation is untenable for AI factory operators, who face mounting time and expense simply maintaining storage infrastructure. In supercomputing, storage typically consumes around 20% of total system cost. With fragmented namespaces, siloed systems, and strained network connections, traditional architectures force AI factories to spend additional resources just to keep expensive GPUs and accelerators from sitting idle.

"AI teams should not have to become storage engineers to decide where data lives, which cluster to mount, how to rebalance workloads, or how to coordinate checkpoints across fragmented infrastructure," NetApp states in the Novus solution brief.

NetApp's engineers addressed this by rethinking a core element of their distributed storage architecture. Rather than simply adding controllers, they targeted a fundamental bottleneck preventing storage from scaling to zettabyte levels: metadata.

The key innovation is decoupling the storage metadata layer from the data layer. Instead of routing all metadata traffic through a fixed set of controllers, Novus allows each GPU client to read and write data directly to the underlying storage, eliminating traffic that can throttle large systems.

According to Banerjee, this decoupling allows both metadata and data to scale independently. "Metadata is a bottleneck even before capacity or bandwidth throttles GPU activity," he writes. "Every open, lookup, and layout request lands on the same controllers that are trying to serve data. A workspace import is millions of small file operations. A checkpoint is a sequential write at terabyte scale, issued by every node at once, arriving directly behind a burst of metadata traffic. When you scale the cluster, the metadata operations scale with it, until inevitably the file system slows or stalls."

The initial Novus implementation runs on NetApp's high-performance AFF A90 storage and leverages pNFS/NFS version 4.2, Linux clients, and the ONTAP storage operating system. NetApp created the Novus Data Director, a software component that offloads metadata traffic to an independently scalable, software-defined control layer running on standard x86 compute. "A client asks the Data Director where a file lives, receives a layout, and then talks directly to the storage," Banerjee explains. "Neither plane waits on the other, and data runs at line rate."

A unified namespace at zettabyte scale brings significant operational benefits. "Novus is architected to allow operators to add capacity and performance without forcing application changes, disruptive remounts, or tenant interruptions," Banerjee writes. "Instead of making the storage topology visible to every workload, Novus presents one consistent data environment built for AI factory scale."

NetApp is targeting AI factories—the massive data centers now under construction to house hundreds of thousands to millions of AI accelerators consuming hundreds of megawatts to gigawatts of power. These installations will inevitably face storage bottlenecks using conventional architectures.

"The architectures that got us to the start of the AI era won't be able to meet the demands we place on them as we continue to accelerate innovation," said Gary Grider, senior director for computing technologies at Los Alamos National Labs, in Banerjee's blog post. "As AI Factories continue to scale, we'll see hundreds of thousands of GPUs hitting a single namespace, creating a massive backlog of metadata operations that will slow or even stall the file system. Novus is a first-of-its-kind architecture that delivers the independent scaling of both the metadata tier and the data tier so AI Factories can grow without being constrained by their ability to access data."

NetApp announced Novus at NetApp INSIGHT 2026, its annual conference in Las Vegas. The company also unveiled new hybrid multi-cloud capabilities in NetApp Platform providing better control over hybrid storage fleets, autonomous operations, a new AI ChatOps interface, sovereignty-aware storage services, and integration with Nutanix's virtualized environments; new zero-copy data activation in NetApp AI Data Engine (AIDE) enabling in-place data access while maintaining governance, resiliency, and openness requirements; and expanded partnerships and new capabilities with SAP, Oracle, and Supermicro.