NetApp in the age of AI: balancing sovereignty and cloud

Data sovereignty has become one of the biggest topics of discussion in enterprise technology. This is the case in particular in Europe, where geopolitical concerns and regulatory requirements are driving organizations to rethink their infrastructure strategies. NetApp claims it can address these challenges head-on with an approach that balances sovereignty requirements with the need for advanced AI capabilities and cloud partnerships.

For this latest episode of Techzine TV, we went to NetApp Insight Xtra in Eindhoven, the Netherlands. We sat down (in rather futuristic chairs from the past) with Jeff Baxter, VP of Product Marketing at NetApp, and discussed how the company is navigating the complex sovereignty landscape. With 18 years of experience at NetApp, Baxter can shed some light on both the company’s product evolution and the changing needs of enterprise customers.

The evolving definition of data sovereignty

The conversation around data sovereignty is progressing beyond simple data residency. While most major cloud providers have addressed where data is stored, the discussion is now moving toward control and accessibility. Organizations are increasingly concerned about what happens if a provider decides to switch off access to their infrastructure.

Baxter explains that NetApp has always focused on customer ownership of data. “NetApp has always been about our customers having ownership of their data, being able to protect their data, having the keys to their data,” he notes. This philosophy also means that NetApp systems can operate autonomously without requiring check-ins with central management systems.

However, the sovereignty discussion has expanded to include supply chain considerations too. Data center providers are now receiving RFPs that require them to list all components inside their racks down to the chip level, verifying whether each component is sovereign-compliant. This represents a blurring of the lines between sovereignty and secure supply chain management.

Sovereign cloud partnerships

NetApp’s deep partnerships with major cloud providers including AWS, Azure, and Google Cloud present both opportunities and questions in the sovereignty context. The company offers services like Azure NetApp Files, Google Cloud NetApp Volumes, and FSX for NetApp ONTAP on AWS.

When AWS launched its European sovereign cloud in Berlin, some customers questioned how a partnership between NetApp (considered a sovereign provider by default) and AWS (not considered sovereign by default) would work. Baxter addresses this “I personally believe that when they’re building out sovereign cloud regions like this, that they truly are sovereign. Now, obviously a lot of the engineering work, a lot of the innovation work may continue in the U.S., but in terms of data not leaving that region, that’s the fundamental.”

NetApp’s strategy is to work with all major cloud providers on their sovereign cloud initiatives while also building out private clouds and sovereign clouds with larger European service providers. The fact that NetApp partners with all major clouds rather than being exclusive to one actually provides comfort to some customers. At the very least it demonstrates that the company isn’t beholden to any single provider.

AI infrastructure and sovereignty

The rise of AI workloads has accelerated the sovereignty discussion, particularly in regions like the Middle East where sovereign AI clusters are a primary concern. Organizations want to perform mission-critical AI work on their data while maintaining full sovereignty over both the data and the compute infrastructure.

This has led to an explosion of neoclouds, specialized cloud providers offering GPU resources in specific regions. Baxter explains: “There’s two things behind the explosion of neoclouds. One is just getting access to GPUs and having these shared providers with access to a huge amount of GPUs. And the other is having neoclouds that can be sovereign in nature based upon where they’re located.”

Data at the petabyte and exabyte scale cannot easily be transported across borders. This means having sovereign GPU resources in the same location as sovereign data storage becomes essential. This creates regional AI ecosystems where both compute and storage can remain within sovereignty boundaries.

How does NetApp AFX address the issue?

At NetApp Insight in Las Vegas, the company announced AFX. You can see that as a fundamental rearchitecture of the NetApp stack. It is a new platform designed to enable exascale AI data infrastructure within customer data centers. This addresses the needs of organizations that want to build their AI infrastructure in their own sovereign environments.

What makes AFX interesting (in our opinion) is that it uses the same ONTAP operating system that NetApp customers have relied on for decades. “It’s not a fork, it’s not some copy that we put the same name on. It’s the same code,” Baxter emphasizes. This means that customers can leverage their existing ONTAP expertise, APIs, and controls while scaling to exascale dimensions.

AFX introduces a disaggregated architecture that separates storage nodes from compute nodes. This allows organizations to scale each component independently based on their specific needs. This could be adding more storage nodes for bandwidth or more compute nodes for data transformation tasks.

DX50 compute nodes for offloading

A key component of the AFX platform is the DX50 data compute node, which represents NetApp’s first foray into including compute resources in their offerings. However, Baxter is clear that this doesn’t represent a strategic shift into the compute business.

The DX50 nodes provide resources specifically for AI data services like vectorized embeddings and data transformation. By offloading these tasks to dedicated compute nodes rather than embedding GPUs in every storage controller, NetApp can optimize both cost and performance. Customers might need 18 storage nodes for bandwidth but only three DX50 nodes for transformation work.

NetApp has already stated that the DX50 will be made available as software-only, allowing customers to use their preferred compute partners. “Our goal is not to get into the compute business,” Baxter confirms. The DX50 exists primarily for ease of deployment and to provide a complete solution for customers who want everything on a single purchase order.

AI Data Engine eliminates the 13-vendor pipeline

One of NetApp’s most recent value propositions is the AI Data Engine. This can potentially replace the complex multi-vendor pipelines that many organizations have built for AI data preparation. Baxter cites one customer that had 13 different vendors involved in their data preparation pipeline, with data moving 7 times between different systems.

This multi-vendor approach creates several problems. First off, it’s inefficient, as it requires constant data movement and transformation. It also creates sovereignty challenges. That is, organizations must verify that all 13 vendors in the chain are compliant with sovereignty requirements.

The AI Data Engine embeds data services directly within the NetApp platform, allowing organizations to perform vectorized embeddings and data preparation without moving data. By leveraging NetApp’s ability to see metadata in real time and access all data within the cluster, the platform can dramatically reduce the complexity of AI data pipelines.

The AI Data Engine also addresses storage bloat. When the same data is referenced multiple times in different vector embeddings, traditional approaches create multiple copies. NetApp’s technology allows data to be stored once while being embedded in multiple different contexts, similar to how snapshot technology can represent the same block logically 200 times while storing it physically only once.

NAND flash shortage and storage optimization

The enterprise storage industry is facing a significant challenge as global manufacturing capacity shifts toward GPUs and AI accelerators. This leaves less production capacity for NAND flash, coming at exactly the wrong time, as AI workloads are generating unprecedented data growth.

Baxter acknowledges this challenge: “We’re at the confluence of unfortunately two conflicting trends. We’re at the confluence of just massively increasing amounts of data that are required and needed for AI, but at the same time the entire global supply chain shifting and prioritizing towards GPUs and AI accelerated compute.”

According to Baxter, NetApp does not want to promise a magic, silver bullet solution but rather to help customers optimize their existing infrastructure. This includes better data tiering, using remaining hard drive capacity for older data, and leveraging the AI Data Engine’s ability to reduce storage bloat through efficient vectorization.

AI operating systems and platform plays

The conversation concludes with a discussion of the emerging trend of companies claiming to offer an “OS for AI.” Baxter expresses some skepticism about this terminology, noting that he’s not entirely clear what an AI operating system means in practice.

From his perspective, what many companies are really offering is not an operating system in the traditional sense but rather a platform that embeds and aggregates various AI data services. Multiple vendors, including startups, are pursuing this approach.

NetApp’s competitive advantage in this space comes from its 30+ year track record and the maturity of ONTAP as a data operating system. While startups may offer innovative data services, they lack the decades of hardening and real-world deployment that NetApp provides. “There’s also a reality to how many years or even decades it takes to truly harden and build out a data operating system,” Baxter notes.

NetApp’s value proposition is that customers can take everything they’ve built with ONTAP (all their confidence, controls, and expertise) and add these new AI data services on top. If competitors can offer significantly better data services, that’s a valid choice, but NetApp believes its combination of proven reliability and new capabilities provides an advantage.

Conclusion

NetApp’s approach to sovereignty demonstrates a pragmatic balance between meeting strict sovereignty requirements and enabling advanced AI capabilities. By partnering with major cloud providers on their sovereign regions, offering autonomous on-premises infrastructure, and embedding AI data services directly into a proven platform, NetApp is addressing the full spectrum of customer needs.

The company’s 30+ year track record is being touted as a strength rather than as a weakness. It provides stability and confidence that many newer entrants cannot match, while innovations like AFX, the DX50, and the AI Data Engine demonstrate that NetApp continues to evolve with changing customer requirements.

Also read: NetApp goes all out for AI with AFX and AI Data Engine