Storage & Data Management Watch: NetApp Novus and the Race to Build Storage Infrastructure for AI Factories
Storage & Data Management Watch · November 11, 2026
NetApp Novus and the Race to Build Storage Infrastructure for AI Factories, Plus Dropbox, VAST Data and Snowflake
|
THE CODEW · Partner Intelligence
Ship better with data and AI skills
Hands-on Python, SQL, and AI learning from DataCamp, for technical teams and founders who want to build with data.
> python · sql · data analysis · ai
|
Run DataCamp → |
AI Infrastructure Is Also a Race to Store and Deliver Data
AI infrastructure is not just a compute race. It is also a competition to store, organize, protect, and deliver the data that keeps AI systems running. On November 10, NetApp announced that Nebul, a European neocloud, is the first to adopt its Novus architecture, which is now generally available and designed for up to 100 TB/s of aggregate throughput. This edition examines NetApp's shift from enterprise storage toward AI factories and sovereign AI clouds, then tracks three related developments: Dropbox's planned Tigris Data acquisition, VAST Data's reported $10 billion in commitments, and Snowflake's autonomous data engineering.
Editorial Note: The Nebul announcement is from November 10; Novus itself was introduced on September 29. The Dropbox, VAST, and Snowflake items are from November 3 to 5 and are included as follow-up monitoring. This edition pairs with Hardware Watch, which covers processors, servers, and accelerators.
Lead Story
NetApp Novus: Nebul Becomes the First Neocloud to Adopt NetApp's AI Factory Storage Architecture
Announced November 10, 2026. Novus was introduced at NetApp INSIGHT on September 29, 2026 and is now generally available.
What happened: On November 10, NetApp (NASDAQ: NTAP) announced that Nebul, a European AI-native cloud provider, is the first neocloud to adopt NetApp Novus, the company's high-performance storage architecture for AI factories. NetApp said Novus is now generally available and is designed to scale to up to 100 TB/s of aggregate throughput. That figure is a company-stated architectural target, not evidence that Nebul is operating at that performance level; the announcement gave no deployment size or benchmark results for Nebul. Novus itself was introduced on September 29 at NetApp INSIGHT 2026 in Las Vegas, so the November 10 news is the first named neocloud customer and the move to general availability.
NetApp's transformation: NetApp built its business on ONTAP data management software and enterprise storage, and now describes itself as an intelligent data infrastructure company. Its pitch is a single data management model that spans branch systems, its disaggregated AFX platform, first-party storage services on AWS, Azure and Google Cloud, and now GPU clusters at neocloud scale. Around Novus it also announced Keystone Sovereign for qualified customers in the European Economic Area, and it completed its acquisition of PEAK:AIO, a UK developer of disaggregated pNFS metadata technology, on November 2. NetApp has not explicitly said PEAK:AIO technology is part of the first Novus release, which ships with NetApp's own Novus Data Director, though analysts see the two as aimed at the same architectural constraint.
The Novus architecture: Novus separates file-system metadata from the data path. In the first release, Novus Data Director metadata software runs on qualified Supermicro x86 servers, while NetApp AFF A90 systems running ONTAP serve as the data layer. GPU hosts see a single namespace over standard parallel NFS, mount once, and data nodes can be added without remounting, which lets performance, capacity and metadata concurrency scale independently. NetApp also positions Novus as retaining ONTAP's multi-tenancy, security and governance. As a validation step, NetApp had analyst firm Omdia audit its testing: Omdia reported near-linear performance gains as ONTAP clusters were added, but its 100 TB/s figure is a modeled projection rather than a result from a full-scale deployment. NetApp's more expansive claim, of a zettabyte-scale file system, remains a design goal.
Why Nebul: Nebul, founded in 2022 and headquartered in Leiden, the Netherlands, builds a sovereign AI cloud that emphasizes data residency, security, compliance and end-to-end multi-tenancy for regulated customers. NetApp's case study says Nebul serves areas such as healthcare, and NetApp's executives pointed to government, healthcare, and biotechnology as regulated sectors. Techzine reported that Nebul chose AFF A90 storage in 2024, so Novus extends an existing relationship rather than starting a new one. For a neocloud, the draw is keeping tenants' data separate while scaling GPU utilization; for NetApp, it is the ability to extend existing customers' ONTAP data estates into a sovereign cloud under the same access and jurisdiction policies, as the company describes it.
The storage bottleneck in AI: NetApp argues that when storage cannot feed data fast enough, GPU utilization in poorly fed AI factories can fall below 30%, so storage performance becomes an input to cost per token. This is a company claim, but the underlying engineering problem is widely recognized: large training runs write checkpoints that hit many storage nodes at once, and metadata operations such as opens and lookups can limit how far a shared file system scales. Newer agentic and inference workloads add many small, concurrent reads against indexes and vector stores, plus key-value cache that spills from GPU memory into shared storage, according to NAND Research's analysis.
Competitive implications: Novus is NetApp's entry into the AI factory storage tier, already contested by VAST Data, WEKA, DDN, Dell Technologies and Pure Storage. NAND Research notes that NetApp has so far played a limited role in the largest GPU build-outs, where VAST, WEKA and DDN have won most neocloud storage deals. NetApp's case rests on assets those specialists lack: ONTAP's data protection and governance, native services across the three largest hyperscale clouds, and an enterprise installed base whose AI workloads may increasingly run in neocloud tenancies. Whether traditional storage providers can capture a meaningful share of AI infrastructure spending will depend on software and data services as much as raw bandwidth, and on recurring revenue from them.
Why it matters: This is a stronger story than a product launch because it names a real infrastructure customer and shows how an established storage vendor is repositioning for AI factories and sovereign AI clouds. The open questions are the ones the announcement does not answer: how large Nebul's deployment is, what throughput it achieves in production, and how Novus compares with competing systems in independent benchmarks.
Evidence status: Company-announced adoption and general availability (NetApp and Nebul, November 10). Performance figures are design targets or modeled by Omdia; no independent production benchmarks or Nebul deployment scale have been published. NetApp's forward-looking statements carry its standard disclaimers. Connect to: Company Intelligence · Hardware Watch · Emerging Technology Intelligence
Secondary Monitoring
Dropbox's AI Infrastructure Ambition: From File Storage to Object Storage
Announced November 5, 2026. The acquisition is planned, not completed.
What happened: Dropbox announced a definitive agreement to acquire Tigris Data, a globally distributed, S3-compatible object storage provider founded by former Uber engineers, and introduced DBX Agent Cloud, a storage and data layer for companies and developers building AI applications and agents. Reported deal value is about $350 million in cash, and Dropbox expects closing in the fourth quarter of 2026, subject to customary conditions. DBX Agent Cloud begins with standalone S3-compatible object storage that customers can buy separately from Dropbox's file-sync plans.
Strategy: Dropbox plans to add filesystem and content-processing capabilities, which would turn documents, images, audio and video into inputs AI models and agents can use, though no timeline was given. Tigris says existing customers keep the same product, API, pricing and zero egress fees during the transaction. Dropbox has not published pricing, egress terms or a general availability date for DBX Agent Cloud. The same announcement included AI actions across customers' existing Dropbox files, which Dropbox says have rolled out to more than 70,000 teams in the US, UK and Australia.
Why it matters: The question is whether Dropbox can credibly move from file synchronization into developer infrastructure and AI data storage, a market with entrenched cloud object-storage providers. Tigris brings a developer product and a multi-cloud, no-egress-fee position; Dropbox brings scale operating storage.
Evidence status: Company-announced plan; transaction value is as reported, and the acquisition has not closed. Product availability and pricing for DBX Agent Cloud are unannounced. Connect to: Tech M&A Watch · Company Intelligence
VAST Data: Why AI Infrastructure Is Driving Massive Storage Commitments
Announced November 5, 2026. Figures are company-reported.
What happened: VAST Data said that in the first three quarters of its current fiscal year, 14 of its more than 1,000 customers made aggregate commitments of more than $10 billion for VAST software and supporting platform hardware. Each of the 14 committed more than $100 million, and four committed more than $1 billion each. VAST also said it structured agreements with supply-chain partners to align more than 200 exabytes of flash capacity through 2029, and that its systems support nine of the top ten frontier AI labs.
Scrutiny: These figures merit caution. Commitments are not recognized revenue, VAST did not name the customers or split software from hardware, and it did not say how much has converted to revenue. Trade coverage noted that four customers likely account for most of the total, that hardware pass-through is included, and that VAST's bookings had passed roughly $4 billion by April, implying sharp acceleration this year. Terms and manufacturers of the flash supply agreements were not disclosed.
Why it matters: VAST is the benchmark for AI-native storage demand, so its numbers frame what NetApp is up against, and the flash supply arrangements show that NAND availability is becoming a competitive factor.
Evidence status: Company-reported commitments and supply arrangements; customers and revenue conversion undisclosed. Connect to: Company Intelligence · Technology Intelligence · Emerging Technology Intelligence
Snowflake's Agentic Data Platform: Automating the Enterprise Data Pipeline
Announced November 3, 2026 at Snowflake Expedition 2026.
What happened: Snowflake introduced new autonomous data engineering capabilities in its CoCo coding agent. CoCo Cloud Agents let teams run agent sessions in a Snowflake-managed cloud runtime rather than on an engineer's laptop, and CoCo Automations let teams schedule recurring tasks, such as pipeline health checks and investigations of failed jobs, described in plain language. When a run finds a problem, CoCo identifies the likely cause and proposes a fix that engineers review and approve before it reaches production. Snowflake says Automations are generally available in Snowsight and the CLI, and in public preview in CoCo Desktop.
Why it matters: Snowflake is positioning agents as the layer that builds, runs, and repairs pipelines on governed data, with engineers keeping approval control. For enterprise data engineering, the implication is a shift of routine pipeline maintenance toward agents; Snowflake itself says data quality and governance remain the main obstacle to deploying autonomous agents safely. Its benchmark claims for CoCo come from internal testing and should be treated as company-reported. The competitive picture with Databricks and others is a question to watch rather than something this announcement settles.
Evidence status: Company-announced capabilities, with availability as stated by Snowflake; performance claims are internal and unverified. Connect to: Technology Intelligence · Enterprise AI Intelligence
The Day at a Glance
| Company | Main development | Significance |
|---|---|---|
| NetApp / Nebul | Novus adoption, general availability | AI factory storage and sovereign AI |
| Dropbox / Tigris | Planned acquisition, DBX Agent Cloud | AI object storage for developers |
| VAST Data | $10B+ commitments, 200 EB flash supply | AI storage demand economics |
| Snowflake | CoCo Cloud Agents and Automations | Autonomous data engineering |
Why It Matters
Storage is a bottleneck for AI economics. If data cannot reach GPUs fast enough, utilization and cost per token suffer, which is why metadata and file-system design are now strategic.
Established vendors are repositioning. NetApp is taking ONTAP governance into neocloud and sovereign AI markets where AI-native specialists lead.
Demand and supply are both under pressure. VAST's reported commitments and flash supply deals show how large AI storage buying has become and how much NAND availability matters.
The data layer is moving up the stack. Dropbox is entering object storage, and Snowflake is adding autonomous pipeline agents.
What to Watch
- Nebul deployment detail: the size of Nebul's Novus deployment and any production throughput results.
- Independent Novus benchmarks: third-party results against VAST, WEKA, DDN, Dell and Pure Storage systems.
- More neocloud customers: whether other GPU-as-a-service providers adopt Novus beyond Nebul.
- NetApp execution: PEAK:AIO integration and how metadata technology evolves in later Novus releases.
- VAST revenue conversion: how commitments translate into recognized revenue and how flash supply holds up.
- Dropbox closing and pricing: completion of the Tigris deal, DBX Agent Cloud pricing and availability.
- Snowflake adoption: customer uptake of CoCo Automations and how enterprises govern autonomous pipeline fixes.
NetApp's Nebul win is a credible first step, not yet proof. A named neocloud customer and general availability move Novus beyond slideware, and the sovereign-cloud angle plays to NetApp's strengths in governance and multi-tenancy. But 100 TB/s is a target, Nebul's scale is undisclosed, and the AI storage tier is already crowded.
The differentiator may be software, not speed. Specialists such as VAST have won the largest GPU build-outs on performance and scale. NetApp's bet is that enterprises moving AI to the cloud will value the same data protection, security, and governance they already run on ONTAP.
AI infrastructure is also a data race. VAST's reported commitments, Dropbox's object-storage move and Snowflake's pipeline agents all point to the same conclusion: storing, organizing, protecting and delivering data is now as strategic as compute.
Sources and Evidence Notes
NetApp Novus and Nebul (November 10; September 29) — NetApp and Business Wire press releases; NetApp Nebul customer case study; StorageReview; NAND Research; Blocks & Files; Business Today Malaysia; Techzine; ITPro. PEAK:AIO acquisition completion (November 2) — NetApp via Business Wire. Dropbox and Tigris (November 5) — Dropbox and Tigris announcements; StorageReview; TechCrunch; StorageNewsletter; itdaily. VAST Data (November 5) — VAST Data press release; Blocks & Files; IT Brief; Channel Insider; StorageNewsletter. Snowflake (November 3) — Snowflake press releases and blog; Techzine; IT Brief.
Evidence labeling: Novus's 100 TB/s figure is a design target, with Omdia modeling and near-linear scaling results from NetApp-commissioned testing; no independent production benchmarks have been published. VAST's commitments are company-reported and are not recognized revenue. The Dropbox acquisition is planned and has not closed. Snowflake's benchmark claims are internal. This briefing does not provide investment advice.
Reviewed by Erwin Castro
on
Wednesday, November 11, 2026
Rating:
No comments: