NetApp Novus
Ambition is your only limit
Scale beyond boundaries with the fastest storage* on the planet
NetApp Novus is a new architecture for gigawatt-scale AI factories, powered by NetApp ONTAP. Designed for leading-edge throughput and low latency, and capable of supporting hundreds of thousands of GPUs, NetApp Novus decouples metadata from data to deliver over 100TB/s in a single namespace.
At AI-factory scale, storage bottlenecks leave GPUs waiting. NetApp Novus federates ONTAP-based data resources beneath a common namespace, separating metadata from data movement to scale throughput without fragmentation, remounts, or infrastructure sprawl.
GPUs at ~2GB/second
Read throughput
Namespace for all your GPUs
“NetApp Novus directly addresses the storage architecture limit by combining disaggregated metadata management with high-performance data services, giving neocloud and GPU-as-a-service providers a more scalable foundation for AI factory operations.”
Mike Leone, VP & Principal Analyst, Moor Insights & Strategy
FAQs
How does NetApp Novus scale without fragmenting my data?
A decoupled metadata layer tracks where everything lives. When a client reads or writes data, they get the data map from the metadata server and then directly access the data layer; the data layer, powered by proven NetApp ONTAP® software on NetApp AFF A90 systems, moves the bytes. The architecture scales performance with each additional storage system. Add a cluster, add its throughput, no remounts and no fragmentation. Every GPU still sees one unified namespace.
How does NetApp Novus improve my return on GPU investment?
NetApp Novus keeps utilization high by removing the storage bottleneck that leaves GPUs waiting, so more of your compute runs at full capacity and more of it converts to revenue. Higher utilization on the same hardware means a better return without buying another rack.
Do my GPU teams need a proprietary client to use NetApp Novus?
No. NetApp Novus works with standard, in-kernel NFSv4.2 / pNFS Flex Files clients, nconnect, and GPUDirect Storage. There's nothing custom to install across a fleet of thousands of GPU servers, which keeps operations simple and onboarding fast.
Ready to get started?
.png?width=5760&format=avif&disable=upscale)