At NetApp Insight 2026, enterprise storage providers highlighted a dual challenge facing artificial intelligence deployments: basic data preparation issues that continue to stall average enterprise pilots, and physical storage throughput limitations that threaten large-scale GPU clusters. According to ITPro's coverage of the event, traditional storage systems are failing to meet the continuous data supply demands of high-performance AI infrastructure.
Addressing Enterprise Data Quality
Data quality remains a primary hurdle for organizations attempting to move AI initiatives into production. Analysis from IDC indicates that 52% of surveyed companies identify data quality as the most critical factor in determining AI deployment success. During the conference, NetApp Chief Executive Officer George Kurian underscored that AI implementation is fundamentally a business transformation process requiring structural readiness alongside software tools.
Solving Massive GPU Throughput Constraints
Beyond data organization, extreme hardware scale has created physical operational bottlenecks. AI processing facilities are expanding from traditional clusters to planned installations housing 50,000 to 100,000 GPUs. When storage infrastructure fails to sustain the required throughput for these setups, processing units sit idle. NetApp Chief Platform and Technology Officer Arindam Banerjee noted that a stalled 100,000-GPU cluster can cost tens of millions of dollars daily in lost operational efficiency.
Novus Architecture and Market Focus
To tackle high-end throughput bottlenecks, NetApp announced Novus, a dedicated storage architecture engineered for large-scale GPU clusters and neocloud operators. Built to operate at zettabyte scale, Novus is designed to deliver throughput of up to 100Tbps. While the architecture targets upper-tier infrastructure providers and massive compute environments, standard enterprise deployments will still need to resolve foundational data management and preparation challenges.
AI-generated editorial illustration; not a documentary photograph.
Source: I can’t believe we’re still talking about AI data readiness but I can’t fault NetApp for persisting
SIGNAL TITANS
Leave a Reply