In its annual customer and partner conference, NetApp kicked it off with some well-deserved bragging about its success. Their success in the financial industry, federal space, and media/entertainment is built on the trusted relationships between their customers, NetApp's customer teams, customer support, and their business partners. Since last year's conference, NetApp has released the AFX 1K and, based on initial field test results, the AFX 2K. So, what's new this year?

AIDE

There were two major announcements during the initial keynote. First, an updated scope for the previously announced AI Data Engine (AIDE). Last year, they announced the AIDE product to discover data, unify data access for data pipelines, and keep things synchronized across various environments around an enterprise. While we know AIDE in that form has been tested in the wild, NetApp recently acquired DataPelago, which may have led them to delay AIDE's general availability release to allow integration of its capabilities. At Insight this year, they announced the renewed AIDE, this time with the ability to work across a customer's entire data estate, not just that which sits on NetApp. This seems a wise move since having tooling unique to just data on NetApp limits applicability and decreases broader adoption in favor of other, more open options. We are eager to see NetApp's foray into what seems to be an increasingly contentious arena.  The specifics of AIDE's internals are limited, but it's claimed to be a full-featured way to analyze and vectorize your file data estate.

NetApp AI data engine software logical stack
NetApp AI Data Engine logical diagram (image source: NetApp)

Novus

Next was something that surprised many. Without burying the lede, its name is Novus and, assuming NetApp's claims are real, it's fast. Really fast. There's a certain cachet when you have the fastest storage on the planet, as claimed by NetApp. They are proud of the accomplishment and it's a mind-boggling amount of data. Claiming 100TB/s of aggregate performance, Novus can feed 50K GPUs at 2GB/s each and, theoretically, keep going. 

The way they accomplish this is twofold:

  1. Use parallel NFS (pNFS) and split out metadata. At scale, metadata is the bottleneck, so separating it from the data path brings performance gains. Clients can run all the getattr requests they want, and the in-memory metadata servers will respond with the data nodes none the wiser. Likely an outcome of the Peak:AIO acquisition, whose Lattice NFSv4 metadata server allows independent scaling of the individual components of NFSv4 metadata: state, core, and metadata catalog. Depending on the type of metadata workload, those three components can expand as needed. Out of the gate, NetApp has approved specific Supermicro servers as the hardware platform for running Lattice; additional hardware platforms will be qualified in the future.
  2. A scalable number of tweaked AFF A90 arrays. A few things here. First, they'll have double the NVRAM of normal A90s to handle AI's checkpoint write surges. Next, they'll upgrade to a 400G frontend; it all seems pretty straightforward. Finally, to break free from traditional tightly-coupled clustered systems, ONTAP or otherwise, there's no backend network; all data nodes (A90+) are independent contributors to the cluster. This eliminates the backend network as a potential bottleneck so the raw performance of the A90s can be realized.
NetApp Novus storage system logical diagram
Novus Logical Design (image source: NetApp)

With current positioning guidance, Novus starts to make sense in 8,000+ GPU environments that can drive a mass of A90 arrays with their 400G uplinks. Its applicability to most customers will be…limited, especially considering the lack of data services enterprises expect. It suggests the old question, "What's faster, a supercar or an SUV?" It all depends on the terrain, and in that metaphor Novus is a sports car, eschewing enterprise data services like ARP and SnapMirror. One thing it gets right is that it can be used by stock Linux clients; no specialized client is needed. In my discussions with customers over the years, this makes adoption more palatable, since there's nothing unique to qualify, compile, or keep up to date.

Starburst

Hinted at by the re-launch of AIDE and the acquisition of DataPelago, the data platform market is an area of competition amongst the storage OEMs at present. Bringing a curated set of analytics tools to bear on your data is a hot topic and considered a basic need for growing enterprise AI services. Though it was not an announcement at Insight itself, NetApp's growing partnership with Starburst is worth discussing. Starburst is a data engine that has connectors into data sources like Kafka, MySQL, or MongoDB, to name a few. It performs federated queries across its sources and pipes results into an open table like Iceberg, and can mask data to prevent PII leakage. These tables form 'materialized views' that can be reused repeatedly by upstream apps as needed to save repeated query cycles. Starburst is complementary to AIDE in that it excels at querying structured sources while AIDE is good with unstructured data; with your data unified by each platform's respective MCP capabilities.

ONTAP in Oracle Cloud

NetApp and Oracle Cloud Infrastructure targeted use cases
NetApp + OCI targeted use cases (image source: NetApp)

In addition to their first-party offerings in AWS (FSxN), Azure (ANF), and GCP (GCNV), NetApp added Oracle Cloud Infrastructure (OCI) to the mix, along with the third-party Cloud Volumes ONTAP (CVO). It is a fully managed, OCI-native storage service that runs ONTAP data management software inside OCI. While it can be used for anything, it's a good fit for databases, enterprise applications, VMs, and AI data pipelines, to name a few. 

Existing ONTAP workloads should move with minimal application redesign or storage management changes. When it comes to integrated AI services, both companies plan to collaborate and bring Oracle's AI efforts into ONTAP so AI can run where the data lives and eliminate data movement; general availability is planned within the next 12 months. This gives NetApp a first-party storage service across all four hyperscalers (AWS, Azure, GCP, and OCI) and provides ONTAP-based customers with a low-risk onramp into OCI.

Why WWT and NetApp work better together

NetApp has laid out an ambitious set of announcements. Turning them into working infrastructure, on time, in the customer's data center or cloud, is where World Wide Technology (WWT) comes in.

A partnership proven in the market

On October 1, 2026, at Insight itself, NetApp named WWT its FY'26 North America Partner of the Year and AI Partner of the Year. The AI award recognizes AI business performance, customer impact, and leadership in scaling AI solutions globally. WWT's Bob Olwig, EVP Global Alliances, said the relationship lets WWT "help customers build a foundation, not a patchwork." WWT lists NetApp among its strategic partners, alongside Cisco, Dell Technologies, HPE, Intel, Microsoft, AWS, and Google Cloud.

That breadth matters because the Insight announcements touch many of those partners. Novus sits in a GPU-heavy environment. The cloud announcements involve Azure, Google Cloud and OCI. Customers do not buy these pieces separately; they need them to work as one design. WWT's long-standing relationships with each vendor let us design across them, rather than advocate for a single layer.

Conclusion

OEM conferences are an opportunity to celebrate the past year and show off what's coming, and Insight 2026 doesn't disappoint. Claiming the title of the fastest performance system is something to be proud of. Diving into the architecture of Novus shows a serious engineering effort that keeps NetApp's core strengths in the mix. It remains to be seen whether NetApp will make it easier to get data into it and whether the technology in Novus will move down-market into something that makes sense for customers other than neoclouds. Perhaps the most interesting part of it is the metadata clusters, presumably derived from Peak:AIO's Lattice, in which each of the three components of NFSv4 metadata can be scaled. Separately, the broadening of AIDE's focus combined with the Starburst partnership is a welcome change. The walled garden approach of AIDE's initial incarnation limited its applicability and would have hindered adoption; NetApp got this one right. 

All in all, we see NetApp putting its bets in the right places. They focus on their forte of freeing data so it can be put to use, one of the founding principles of NetApp and ONTAP themselves. 

Technologies