NetApp Novus Targets AI Factory Storage Bottlenecks with 100 TB/s Aggregate Throughput
NetApp (NASDAQ: NTAP) announced NetApp Novus, which the company calls the fastest storage on the planet, purpose-built for AI infrastructure providers operating large GPU environments. The next-generation architecture supports a zettabyte-scale file system and delivers up to 100 TB per second of aggregate throughput, designed to keep millions of GPUs continuously fed with data across AI factories.
- Disaggregated architecture separates metadata from the data path; performance, capacity, and concurrency scale independently under a single namespace over industry-standard NFS
- Combines Novus Data Director (on qualified Supermicro infrastructure) with ONTAP data services delivered through AFF A90 systems
- Built to exceed 100 TB/s throughput and support high utilization of hundreds of thousands of GPUs while preserving ONTAP enterprise data services
- Traditional architectures can see GPU utilization drop below 30 percent; Omdia auditing found near-linear scaling with effective capacity into the dozens of exabytes
- Orderable today, with a path to software-defined deployments under the same namespace and operating model
The AI factory storage bottleneck
Neoclouds and GPU-as-a-Service providers are building AI factories at a scale traditional storage architectures were never designed to support. When storage can’t feed enough data to GPUs, utilization can drop below 30 percent, undermining both results and return on investment. “AI factories struggle and GPU economics collapse when data can’t keep up,” said Syam Nair, Chief Product Officer at NetApp. “HPC-era storage alone cannot serve AI Factory scale — the industry needs a true architecture built for it from the ground up.”
Disaggregated by design
Novus separates metadata from the data path, removing scale limits so performance, capacity, and concurrency grow independently under a single namespace using industry-standard NFS. The initial release combines NetApp Novus Data Director — the metadata software running on qualified Supermicro infrastructure — with ONTAP data services delivered through AFF A90 systems, forming a disaggregated architecture built to exceed 100 TB/s while preserving the ONTAP enterprise data services customers already trust. Tony Palmer, Chief Analyst at Omdia, said audits of NetApp Novus demonstrated near-linear performance scaling as ONTAP clusters were added to a single global namespace, with modeling projecting sequential read throughput up to 100 TB/s and effective capacity reaching dozens of exabytes and beyond.
Three customer benefits
NetApp lists three adoption benefits. First, every GPU stays productive: separating metadata services from the data path reduces storage-related bottlenecks and keeps AI infrastructure fed. Second, scaling without fragmentation: billions of files and massive AI datasets grow under a single namespace, with metadata, bandwidth, and capacity scaling independently and no storage silos. Third, infrastructure efficiency: operators scale only the resources each workload needs, reducing power, space, and overhead while preserving the resiliency, security, and multi-tenant capabilities shared AI environments require. Mike Leone, Vice President & Principal Analyst at Moor Insights & Strategy, noted that storage unable to deliver data at cloud scale becomes a limiting factor for training performance and return on infrastructure investment. Jeremy Foster, SVP and GM of Cisco Compute, added that as AI pushes infrastructure to new levels of scale and speed, “the real differentiation isn’t the GPU — it’s the data and how customers use it.”
