Recommended Free Tools
NetApp CEO George Kurian’s 2025 interview with CRN frames the company’s AI strategy as an effort to prepare and govern enterprise data where it already lives, rather than simply move it into a new AI system. The approach centers on AFX, DX data-processing engines and AI Data Engine (AIDE), with partners expected to help customers turn data-readiness work into production projects. Kurian also warned that a reported $100,000 H-1B application fee could influence where technology companies locate work. His comments describe NetApp’s strategy and his view of the policy; they do not independently verify product performance or establish the fee’s current legal scope.
What Kurian means by “bringing AI to your data”
In the CRN interview, Kurian argues that enterprise AI depends on making existing business data usable, governed and current. “Bringing AI to your data” is a strategy about data placement and management: process and prepare information near its authoritative location when possible, instead of repeatedly copying it into separate environments.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
NETAPP 74672-10 NTEAPP Network Appliance 3.5" Drive Caddy | $97.00 | Buy on Amazon |
That distinction matters because moving data can create duplicate stores that must be secured, synchronized and refreshed. A workflow that can identify changed information and process only what changed may reduce unnecessary movement and repeated work. It does not, by itself, create an AI model, guarantee better answers or remove the need for suitable compute, data engineering and model evaluation.
NetApp’s proposal is to extend its role from storage into data management, security, governance, hybrid-cloud operations and AI preparation. Storage remains the foundation of that pitch; the broader claim is that the company can manage more of the path between stored information and AI use.
#1 Best Overall
AFX, DX and AIDE: the roles NetApp describes
AFX and DX engines
Kurian described AFX as a composable or disaggregated infrastructure approach combining AFX storage platforms with DX engines for data processing and transformation. In that framing, storage-oriented access and data-preparation work are distinct parts of an architecture: one holds and serves data, while the other helps make it suitable for downstream AI workflows.
The interview, published around NetApp Insight 2025 in Las Vegas, describes the architecture at a strategic level. It does not establish AFX’s general-availability date, supported configurations or regions, pricing, performance benchmarks, deployment requirements, customer references, or which functions are production-ready rather than planned. Buyers should confirm those details directly for the specific configuration they are considering.
AI Data Engine
Kurian presented AI Data Engine, or AIDE, as a way to organize and prepare data for AI while applying governance and guardrails, protecting information used in workflows, and keeping AI-ready data current without unnecessary copies. The interview does not specify whether AIDE is a standalone product or a collection of platform capabilities, nor does it document supported NetApp systems, cloud environments, licensing, or implementation dependencies.
It also does not establish whether AIDE provides a full catalog, vectorization, retrieval-augmented generation, lineage and policy-enforcement stack, or how responsibilities are divided among NetApp, partners and third-party AI platforms. The stated purpose should not be mistaken for a complete product specification.
Formats, metadata and incremental processing
Kurian described NetApp’s intended scope as extending beyond file, block and object storage to include vector embeddings, tokenized data used by large language models, Apache Iceberg tables, Apache Parquet files, CSV, JSON and other structured, semi-structured and unstructured datasets. Those are his descriptions of the company’s data scope; they do not demonstrate equal native support for every format or make NetApp a replacement for every database, vector database, lakehouse or model-serving platform.
Metadata and change tracking are central to the logic. If a system can identify what has changed, a downstream pipeline may avoid repeatedly scanning or preparing an entire large dataset. Kurian cited SnapDiff as a scalable change-detection mechanism for this purpose. That is different from saying SnapDiff itself trains or serves models, or automatically controls every retraining workflow: the interview does not establish those functions.
The practical benefit depends on implementation. Buyers should test whether the system recognizes changes in their actual data sources, preserves access controls and lineage, and integrates with their chosen processing and model tools. A copy-efficient design can still require integrations, operational oversight and compute; it cannot remedy poor data quality or unclear ownership on its own.
NetApp’s competitive argument—and what remains unproven
Kurian contrasted NetApp’s approach with storage-centric “data cloud” positioning, including that associated with Pure Storage. NetApp’s argument is that it operates across storage and data management, can apply active metadata and tagging, supports a broader range of data forms, and can prepare information with fewer copies across hybrid environments.
This is a competitive position, not independently demonstrated market superiority. The interview does not provide comparative benchmarks, customer adoption figures, total-cost-of-ownership results, or measured reductions in copies, processing time or AI costs. Alternatives such as hyperscaler-native services, lakehouse platforms and specialist data or AI vendors may fit better for particular workloads. The meaningful comparison is whether a proposed architecture works with a customer’s existing data estate and produces measurable improvements without adding more operational complexity than it removes.
What Kurian said about H-1B rules
Kurian said NetApp uses H-1B visas for employees assigned to the United States over the long term and that international technical talent matters to product development. Discussing a reported $100,000 application fee, he argued that the added cost would make U.S. hiring harder and could lead global companies to place work where talent is available. He also made the issue personal, saying his family could not have afforded such a fee under the rules being discussed.
Those remarks establish Kurian’s position and prediction, not the final legal status or scope of the policy. The interview does not say whether the fee applied to all cases, what exemptions or waivers existed, when it took effect, whether it was temporary, or how subsequent agency, court or congressional action affected it. It also gives no NetApp visa-volume figure, financial exposure or evidence that the company changed hiring plans. Readers should not treat the interview as current immigration guidance.
For technology companies generally, a substantial upfront immigration cost could make hiring or relocation more expensive and could encourage distributed teams, overseas engineering centers, contractor use or competition for workers already authorized to work in the United States. These are possible business responses, not measured effects attributable to NetApp. Kurian’s concern is that limiting access to international talent may influence where specialized engineering work is performed.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Why partners matter to NetApp’s AI plan
Kurian described the channel as integral to NetApp’s go-to-market strategy, not an adjacent sales route. The company wants solution providers and integrators to help customers design data architectures and deliver consulting and services around AI readiness, governance, security, organization, transformation, data lakes and hybrid-cloud modernization.
The business logic is that many customers need more than storage equipment to move from an AI proof of concept to a production workload. Partners may assess data quality and access, integrate systems, prepare information, and help operate the resulting environment. That creates potential services work before large-scale AI infrastructure purchases arrive.
The interview does not identify a formal AI-readiness service or competency, partner compensation, margins, incentives, customer-relationship ownership or rules for how NetApp and partners divide professional-services work. It also does not establish whether the same opportunity applies equally to resellers, managed-service providers, global systems integrators and cloud marketplaces. Those details matter: partners need a viable recurring-services role, not just implementation responsibility, for the strategy to generate durable economics.
As later context—not part of the original interview—CRN reported in 2026 that NetApp appointed former Microsoft executive Alvaro Celis as chief partner and ecosystem officer. The appointment signals continued attention to the partner organization, but it does not disclose program economics or prove that partners are already capturing AI-services revenue.
Free tools Windows power users keep installed
One-click scans. No signup required.
Are enterprise customers and partners ready?
Kurian characterized enterprise AI adoption as early, with proofs of concept preceding the first wave of production deployments. Moving from demonstration to sustained use involves more than choosing infrastructure. Organizations must establish data quality, ownership, permissions, privacy, security, lineage and evaluation practices, while integrating structured and unstructured sources with the systems that run AI workloads.
Partners can help close skills and integration gaps, but their role does not remove the customer’s responsibility for governance or model risk. Data residency requirements may restrict where processing occurs; legacy applications may expose information in difficult-to-use forms; frequently changing data can make repeated indexing costly; and archival datasets may not benefit from near-real-time change detection. Organizations already committed to a cloud lakehouse or vector database may need integration rather than replacement. Smaller firms without substantial data-management needs may not benefit from an enterprise-scale platform.
What to test before adopting the approach
A pilot should measure the customer’s own workload rather than rely on broad claims about AI-ready data. Agree on a baseline and require the vendor and implementation partner to demonstrate the following:
- Data locality and copies: Identify which data remains in place, which must move, and how many persistent or temporary copies the workflow creates.
- Incremental work: Compare a full reprocessing run with processing after representative files or records change. Record time, compute and storage consumed.
- Format coverage: Test the actual formats and applications in use, including any proprietary or application-specific data—not just a short list of common formats.
- Governance and security: Verify that permissions, encryption, lineage, auditability and relevant protections persist through preparation and downstream use.
- Interoperability: Confirm connections to the customer’s on-premises systems, cloud environments, data platforms and AI tools. Account for network and cloud egress costs where data crosses boundaries.
- Operations and recovery: Test monitoring, failure handling, rollback and recovery, and determine who supports each layer.
- Commercial terms: Obtain written details on availability in the buyer’s region, supported configurations, licensing, renewals and any consumption charges.
- Evidence: Request workload-relevant production references and independent performance evidence; do not treat a proof of concept or product presentation as proof of production results.
These tests also reveal the trade-offs. Processing near data may reduce movement but add infrastructure or management layers. Active metadata can improve discovery only if tagging and ownership remain reliable. A tighter integration may simplify support while increasing dependence on one vendor’s stack, and partner services can speed deployment while adding project cost.
What the interview says—and what it does not
Kurian’s interview is useful for understanding NetApp’s intended direction: turn its storage and data-management footprint into an operating foundation for AI, with partners helping customers prepare data and deploy workloads. It is not independent proof that AFX or AIDE delivers lower costs, faster AI deployment or better model quality, nor does it settle the legal details of the H-1B fee discussed. Those outcomes depend on product availability, architecture, workload evidence and the applicable rules at the time of a decision.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
