In this episode of the AI Intelligence Lab series, NetApp’s Sandeep Singh, SVP & GM of Enterprise Storage, and NVIDIA’s Charlie Boyle, VP of DGX, unpack what an AI factory really means today. Two years ago, most people thought “AI factory” meant big training clusters for custom foundation models. Now, with so many strong models available off the shelf, the focus has shifted to a multifunction factory that flexes between training and inference, with the real output being tokens—units of intelligence that power recommendations, insights, and copilots. Charlie describes it like a factory with a big lever: one day you lean into training; but during peak retail season, for example, you pull the lever toward inference to crank out personalized recommendations at scale.
Data as the fuel for AI
A central theme is that data—not just GPU horsepower—is what really fuels the AI factory. Early models could only answer questions about the past, because everything they “knew” lived in pre-trained weights; now, useful AI systems must combine model intelligence with live enterprise and external data via tools and retrieval. That means enterprises need a real data strategy for AI: knowing where data lives, who can access it, how it’s segmented and protected, and how to make it AI-ready without copying it everywhere. NetApp’s long history in data discovery, governance, and intelligent storage becomes the foundation for letting customers “talk to their data” safely and effectively.
A deep, full-stack partnership
Sandeep and Charlie emphasize how long and deep the NetApp–NVIDIA partnership runs, from early work on storage for training to today’s fully validated “AI factory” blueprints. NVIDIA SuperPOD includes NetApp as a certified storage option, giving customers an end-to-end, tested design spanning compute, networking, storage, and software that’s optimized for performance and cost per token. On top of that, the joint AI Pod and AFX solutions deliver converged, full‑stack systems so customers don’t have to stitch components together themselves—they can start from a proven architecture and focus on outcomes, not infrastructure plumbing.
AFX: Disaggregated, AI-ready data infrastructure
NetApp’s AFX portfolio, announced at INSIGHT 2025, is positioned as the data backbone for these AI factories. It provides disaggregated data infrastructure that lets customers scale performance and capacity independently across training, post‑training, and inference workloads, instead of over‑provisioning one to get enough of the other. AFX combines high-performance storage with embedded compute nodes that run AI data services—metadata cataloging, classification and PII reduction, vector databases, and NVIDIA GPU‑accelerated pipelines—to turn raw data into AI‑ready data without massive data bloat. Because it’s built on ONTAP, customers also get enterprise‑grade resilience, ransomware protections, existing automation, and hybrid cloud connectivity out of the box.
Flexible architectures for evolving use cases
The conversation closes on flexibility: every AI factory looks a little different, and needs to evolve as use cases and models change. NetApp outlines a portfolio approach: StorageGRID and object for modern data lakes and long‑term AI data, E‑Series with parallel file systems for ultra‑high‑performance inner‑ring training, and AFX as the hot, intelligent tier in the middle. NVIDIA and NetApp work together to validate these combinations so customers can mix and match without getting stuck in “analysis paralysis” or forklift upgrades. The message to B2B tech buyers is clear: with a solid, scalable foundation from NetApp and NVIDIA, you can start building your AI factory now and grow it as your ambitions—and your data—expand.
Explore More of the AI Intelligence Lab
In Episode 5, learn more about how ISV partners like Domino Data Lab choose NetApp + NVIDIA infrastructure because it enables rather than constrains innovation.