NetApp Extends the NetApp Platform for Agentic AI and AI Factory Scale

Key Highlights

• NetApp introduced Novus, an AI Factory architecture built for zettabyte-scale file systems and architected for up to 100 TB/s of throughput. Novus is orderable today.
• AI Data Engine extends discovery, metadata enrichment, and governance to NFS, SMB, and S3 data on ONTAP, StorageGRID, and third-party storage.
• NetApp Console adds autonomous operations, an AI ChatOps interface, MCP access, and a customer-installed delivery model for regulated and disconnected environments.
• Keystone Sovereign will give qualified European customers regional controls over data, support, and escalation, starting with pilots in Germany and France.
• NetApp broadened the NetApp Platform through Oracle, Supermicro, Nutanix, and Commvault.

The News

NetApp used INSIGHT 2026 to extend the NetApp Platform for agentic AI, hybrid cloud, and large AI factories. The announcements include Novus, an expanded AI Data Engine, Console autonomous operations, Keystone Sovereign, a planned first-party OCI storage service, and expanded Supermicro, Nutanix, and Commvault integrations. They continue a multi-year strategy to make the NetApp Platform a common data and control layer for enterprise applications and AI. For more details and press releases, see the NetApp newsroom.

Analyst Take

INSIGHT 2026 connects several themes HyperFRAME has tracked during 2026: metadata as a control layer for enterprise AI, zero-copy data activation, infrastructure choice, and sovereign control. HyperFRAME argued in March that AIDE could become a persistent context layer beneath AI workflows, supporting retrieval, governance, and policy enforcement. That model worked best when NetApp controlled the underlying data, which limited its value in enterprises whose NAS, object storage, and cloud services come from multiple vendors.

Heterogeneous NFS, SMB, and S3 support removes that structural limit. A metadata layer confined to the NetApp portion of an estate falls short as a control point for AI. The broader catalog gives NetApp a path to maintain context, governance, and discovery across data it does not store, while preserving deeper integration where it owns the system. Planned open integrations with query engines such as Starburst and Onehouse.AI would let NetApp compete for control of metadata and governance without owning every storage system or analytics engine.

Enterprise AI needs current data, and repeated ETL copying adds latency, capacity consumption, and governance boundaries. HyperFRAME's DataPelago analysis identified distributed processing close to the data as an important extension of NetApp's strategy because discovery alone does not activate data. AIDE can describe and govern the data; DataPelago can potentially process it where it resides. Humans can define policy, approve sensitive changes, and supervise exceptions, but agents act far faster than humans can approve individual actions. NetApp is responding with automated policy enforcement around data access and infrastructure management. MCP support and the ChatOps interface in Console give agents and administrators another route into those services within defined controls. The architectural issue is whether the same policies remain authoritative when actions originate from people, scripts, APIs, or agents.

HyperFRAME has asserted that sovereign infrastructure requires control over more than data location. Keystone Sovereign and the customer-installed Console extend the boundary to telemetry, management, support, and escalation. That turns sovereignty from a residency attribute into an operating model, with technical and human processes kept inside the same jurisdictional boundary. The next step is bringing AIDE, agentic interfaces, and automated policy enforcement inside that model.

Nutanix support, due this fall, expands hypervisor choice. JetStream supports VMware migration and protection, OCI NetApp Storage Service adds a first-party cloud destination, and Commvault connects third-party recovery management to ONTAP protection functions. These integrations keep NetApp attached to the data layer while customers change the infrastructure around it. The value lies in whether metadata, policy, resilience, and management remain consistent as those surrounding components change.

Large AI factories introduce requirements that conventional enterprise storage was not designed to meet. AFX gives NetApp a disaggregated scale-out architecture, but NetApp says its networking and coordination model becomes cumbersome when thousands of systems must share one namespace. Novus addresses the upper tier by separating metadata from the data path and letting clients reach data nodes in parallel. Namespace size, aggregate bandwidth, density, secure multitenancy, and provider control-plane integration directly affect GPU utilization and cost per token.

Novus puts NetApp against VAST in integrated AI-scale data infrastructure and against Hammerspace in standards-based pNFS and FlexFiles architectures, although the approaches differ materially. Hammerspace abstracts across heterogeneous storage, while NetApp scales ONTAP into the same class of parallel architecture. VAST brings an architecture designed around large-scale AI data services from inception. ONTAP's large enterprise installed base already provides NetApp with an incumbent advantage. AIDE, Console, cloud services, and now Novus extend that footprint into metadata, governance, control, and AI Factory infrastructure. The planned PEAK:AIO acquisition adds scale-out metadata and parallel NFS expertise.

Metadata is becoming an execution layer for enterprise AI, determining how data is discovered, governed, accessed, and moved across AI workflows and infrastructure. NetApp now has many of the components required to connect that metadata layer with distributed processing, resilience, mobility, autonomous management, and AI Factory infrastructure. The opportunity is to carry the same data, policy, and management model through on-prem, cloud, sovereign, and AI Factory environments.

What Was Announced

AI Factory Infrastructure

NetApp introduced Novus, an architecture for very large AI Factory environments. It separates metadata services from the data path so performance, capacity, and concurrency scale independently under one namespace using industry-standard NFS. NetApp said in the analyst briefing that Novus uses pNFS 4.2 and FlexFiles, letting standard Linux NFS clients reach storage nodes directly and in parallel without a proprietary client or separate parallel file system. NetApp describes Novus as built for zettabyte-scale file systems and architected for up to 100 TB/s of aggregate throughput. The company also cited approximately 2 GB/s per GPU in a 50,000-GPU environment during the briefing.

Novus is orderable today, with lighthouse deployments expected roughly one to two months after INSIGHT. The first implementation combines NetApp Novus Data Director on qualified Supermicro infrastructure with ONTAP data services on specialized AFF A90 systems. NetApp plans software-defined ONTAP on selected x86 servers, and existing NetApp hardware is not intended for Novus-scale deployments. In the briefing, NetApp expected lighthouse customers to begin roughly one to two months after INSIGHT, followed by wider AI Factory deployment. The planned PEAK:AIO acquisition, announced September 25 and subject to regulatory approvals, adds metadata and parallel NFS technology aligned with the Novus design.

NetApp and Supermicro are developing jointly validated AI infrastructure spanning AI factories, enterprise deployments, neoclouds, and sovereign AI. NetApp AIPod with Supermicro targets enterprise training, inference, and agentic workloads from dozens to hundreds of GPUs, complementing Novus at the larger AI Factory tier.

Governed Data Activation

AI Data Engine now extends beyond NetApp-resident data. AIDE can discover, index, extract, and enrich metadata across NFS, SMB, and S3 repositories on ONTAP, StorageGRID, and third-party storage, creating a common repository. Data on the NetApp Platform gets faster metadata updates because AIDE interacts directly with the underlying systems, while external stores participate in the same catalog. AI applications can discover and access governed data where it resides without copying it into a separate AI repository. The expansion reflects customer feedback that a catalog covering part of the data estate fell short, even where NetApp held most of the installed storage.

NetApp also previewed upcoming AIDE capabilities: expanded AI-powered data understanding, deeper governance intelligence, open analytics integrations with engines such as Starburst and Onehouse.AI, and new agentic services. NetApp positions these as steps toward semantic search, knowledge graphs, and decision context.

The DataPelago acquisition adds distributed processing close to enterprise data. Its integration with AIDE and the wider NetApp Platform remains a key development milestone.

Hybrid Cloud, Autonomy, and Sovereignty

NetApp Console becomes the common management front end for the NetApp Platform. Autonomous Operations delivers AI-assisted, outcome-based automation with human oversight, including in air-gapped environments. The AI ChatOps interface offers conversational, intent-based provisioning, reporting, predictive analytics, and automated remediation across ONTAP systems. It runs through an open LLM gateway, so customers can use their own LLM while actions stay within defined storage classes, policies, and governance boundaries. REST APIs, Terraform, and Ansible remain available for traditional automation, while MCP adds an agent-facing interface. NetApp is also bringing Active IQ Unified Manager functions into Console. A second Console delivery model, installed in the customer's environment, brings fleet management, observability, and AI to dark sites without requiring telemetry to leave the customer perimeter.

NetApp and Oracle announced OCI NetApp Storage Service, a planned, fully managed OCI-native service that brings ONTAP data management into Oracle Cloud Infrastructure. It is designed for databases, enterprise applications, virtualized environments, EDA/HPC, regulated workloads, and AI data pipelines while preserving familiar NetApp operations. Customers will manage it through the OCI Console and SDKs, ONTAP APIs, and existing workflows. General availability is planned within 12 months.

Keystone Sovereign adds a sovereignty-focused option to NetApp's as-a-service model for qualified European customers. Initial pilots start in Germany and France, with additional countries planned. The planned controls cover European-based data residency for relevant customer data, logs, backups, and telemetry where applicable; European-based support and escalation; European-controlled access and management processes; documentation of telemetry and data flows; and contractual structures that reinforce sovereignty expectations.

Infrastructure Choice and Data Resilience

NetApp's Nutanix integration, scheduled for general availability this fall, lets customers use NetApp storage and ONTAP data services with Nutanix Cloud Infrastructure while retaining VMware, KVM-based, and container options. That gives customers a migration path without a corresponding storage migration.

The expanded Commvault integration connects ONTAP autonomous ransomware protection signals on primary storage to Commvault recovery workflows. It adds near-real-time threat detection, deeper scans for evasive threats, cleanroom validation in an isolated environment, and clean recovery from the Commvault console. Joint customers can also invoke NetApp Snapshot, SnapMirror, SnapLock, and SnapRestore functions from the Commvault console.

NetApp reiterated the role of JetStream Software in VMware protection and migration, which remains as set out in its August acquisition announcement.

Looking Ahead

Maintaining current, trustworthy metadata gets harder when NetApp does not control the underlying storage. AIDE's broader reach raises its value and the requirements for identity, access, lineage, and policy consistency across heterogeneous systems. Metadata is becoming an execution layer that governs retrieval, transformation, agent access, and infrastructure behavior. Integrating DataPelago into AIDE, Console, and NetApp Platform workflows would extend that control from discovery and governance into distributed processing.

Novus faces a different proof point, but it connects to the same platform thesis. Lighthouse deployments need to demonstrate sustained throughput, metadata performance, density, failure recovery, secure multitenancy, and GPU utilization at target scale. PEAK:AIO adds relevant engineering expertise and Supermicro provides a defined infrastructure starting point, while mixed workloads and sustained production use will provide the harder validation. Cost-per-token claims gain meaning once customers can connect storage behavior directly to GPU utilization and economics.

Oracle, Supermicro, Nutanix, and Commvault extend the NetApp Platform into cloud, server, hypervisor, recovery, and AI infrastructure choices while customers keep the data layer underneath them. The value comes from keeping metadata, policy, resilience, and management consistent as those surrounding components change. Keystone Sovereign and local Console apply the same principle to jurisdictional boundaries, where data, telemetry, support, escalation, and administration must remain inside the selected operating model. Extending AIDE, autonomous policy enforcement, and agent interfaces into those environments would make that architecture more complete. NetApp's opportunity is to make the NetApp Platform the common data and control layer across enterprise and AI workloads, with the same metadata, policy, resilience, and management model preserved as infrastructure choices change.