NetApp moved beyond enterprise data storage into higher-level data management with the preview of a new AI Data Engine product and AFX hardware in partnership with Nvidia.
AI Data Engine (AIDE), expected to become generally available in the first calendar quarter of 2026, is a fourfold set of software tools that collect, curate, sync, protect and prepare data for use by AI applications. It's different from the data pipeline systems in widespread use by data scientists, such as Apache Kafka, because it's focused on centrally collecting and curating an organization's data, rather than transforming and transmitting it from a source to a destination.
AIDE uses the NetApp metadata engine to generate a structured, centralized view of an organization's data, along with a set of APIs to query that data. It uses NetApp's SnapMirror and SnapDiff snapshot functions to keep data in sync and automatically detect changes in data sets. A feature called Data Guardrails automatically classifies and protects data.
Finally, a Data Curator component generates vector embeddings at the data storage layer before data is moved to vector databases to improve efficiency. Data Curator can act as a retrieval-augmented generation endpoint using Nvidia NIM microservices. In the future, AIDE will also run on Nvidia RTX PRO Servers.
NetApp and several other major enterprise storage vendors have already partnered with Nvidia as part of its AI Data Platform reference architecture, launched in March. But Gagan Gulati, senior vice president and general manager of AI, cloud and security at NetApp, said in an interview with Informa TechTarget that it's the start of bigger things to come for NetApp with Nvidia.
"This is all being done in collaboration with Nvidia," Gulati said of AIDE, which led the news announcements at NetApp's Insight 2025 conference this week. "Everybody's running with Nvidia, but to get the power of enterprise data, NetApp's going to be the first big data infrastructure company that they're partnering with."
NetApp brings unstructured data chops to AIDE
Simon Robinson
NetApp is not alone in offering a data engine product for AI data management -- Oden Technologies, Scale AI and Vast Data also offer such systems. But NetApp has focused on unstructured data and metadata for 30 years, which could give it a leg up in the AI infrastructure race, said Simon Robinson, an analyst at Omdia, a division of Informa TechTarget.
"There are certain aspects of managing a data pipeline in an AI context that NetApp has some unique perspectives on because of its proximity to the underlying data," Robinson said. "Classifying that data and applying policies around it takes overhead; it's very processor-intensive. Building dedicated processing resources around that is pragmatic, but there is also some really interesting innovation."
Nancy Gohring
For example, AIDE's Data Guardrails identifies sensitive data, for which users can set usage policies. The feature can also redact information or exclude sensitive data from use by AI apps.
"NetApp is arguing that masking sensitive data at the beginning of the data pipeline is a strong way to exclude that data in advance, rather than try to prevent access to that data later in the pipeline," said Nancy Gohring, an analyst at IDC. "In reality, it’s probably not an either/or, but an added layer of protection. It will be key for them to build the right partnerships and integrations with the data and AI governance tools that enterprises are beginning to adopt."
NetApp AFX decouples compute and capacity
AIDE is accompanied by a generally available NetApp AFX all-flash array certified as part of Nvidia's DGX SuperPod architecture, which comprises Nvidia DGX servers, InfiniBand and Ethernet networking components, management nodes and storage. AFX represents NetApp's foray into disaggregated storage, a trendy term for systems that decouple storage controllers from storage capacity, or offer an alternative to hyperconverged infrastructure, depending on the vendor.
For NetApp, “disaggregated” refers to infrastructure spread between on-premises data centers, remote sites and the public cloud. With a minimum configuration of four controllers, AFX systems aren't meant for the smallest edge computing locations but can blend in with customers' existing deployments with public cloud storage services that already run NetApp's ONTAP storageoperating system.
AI is inherently a hybrid cloud workflow.
Sandeep SinghSVP & GM, NetApp
"AI is inherently a hybrid cloud workflow," said Sandeep Singh, senior vice president and general manager of enterprise storage at NetApp, in an interview. "So having the ability to unify and enable a hybrid workflow across on-prem and public clouds or new clouds, sovereign clouds, becomes critical."
IDC's Gohring said NetApp's approach to disaggregated storage could also help enterprises efficiently use GPUs when training AI models.
"At times, you want to scale storage capacity for very large training data sets, but you don’t want to have to scale your very expensive GPUs at the same rate, since that compute isn’t actually being used," she said.
NetApp's disaggregation vs. Dell Automation Platform
Robinson said that with the launch of NetApp AFX, two major data storage vendors are using “disaggregation” in different ways, potentially creating confusion in the market.
Dell rolled out a disaggregated infrastructure strategy earlier this year with its Dell Automation Platform. But for Dell, the term has more to do with changing competitive dynamics than a technical distinction, he said, and offering a replacement for the VxRail hyperconverged infrastructure Dell sold when it owned VMware. Now that Broadcom owns VMware and has alienated some VMware customers with pricing and other changes, Dell has looked to capitalize with an alternative to VMware's vSAN and a partnership with partnership with Nutanix.
"Basically, every non-VMware company is like, 'Okay, how do we help customers who want to move off of VMware move off VMware?'" Robinson said. "But from NetApp's point of view, disaggregation is about separation, within the storage architecture, between the storage processing and back-end disks."
This has traditionally been done using scale-out storage systems, which NetApp also offers, but the new AFX system parallelizes its file system between separate controllers, rather than using one centralized set of controllers to maintain consistency between systems, according to NetApp's Singh.
"With AI, the scaling model can be very different [from other workloads]," Robinson said. "There are companies that are going to scale their capacities way, way out of whack with their performance, or more likely, they're going to be doing a hell of a lot of I/O on a relatively small amount of back-end storage for inference at scale."
Beth Pariseau, a senior news writer for Informa TechTarget, is an award-winning veteran of IT journalism covering DevOps. Have a tip? Email her or reach out @PariseauTT.
Simon Robinson
Nancy Gohring
