NetApp.

Here is a slightly strange fact about the modern economy: some of the most valuable companies are ones you will never knowingly interact with. You do not buy their products at a store. You do not have an account. You have, however, almost certainly touched their machines - your bank balance, your medical records, a video you streamed - all of it sitting, at some point, on hardware sold by a company called NetApp.

NetApp sells storage. That sentence does a lot of quiet work. In 1992, three engineers - David Hitz, James Lau and Michael Malcolm - looked at the file server, an unremarkable box that held shared documents, and asked a question that sounds obvious in hindsight and heretical at the time: what if this were an appliance? Not a general-purpose computer someone had to babysit, but a dedicated, non-programmable box that did exactly one thing - serve files - and did it reliably. They called the company Network Appliance, which is the rare startup name that tells you precisely what the product is. The first units shipped in June 1993. The IPO followed on the Nasdaq in November 1995, ticker NTAP.

The appliance idea was good enough to build a Fortune 500 company on. But the more durable asset turned out to be the software running inside the box: an operating system called ONTAP. If you want to understand why NetApp is still relevant in 2026 and not a museum piece, ONTAP is the answer. It is the through-line. The hardware has been reinvented many times; the data-management brains have been continuously extended rather than replaced. That is an underrated kind of moat - the kind that compounds when you are not looking.

The interesting turn in the plot is what happened to the box. For a long time, storage was a physical thing you owned and racked in your own data center. Then the cloud arrived and threatened to make the whole category obsolete - why buy a NetApp array when Amazon will rent you storage by the gigabyte? A lesser company defends the castle. Under George Kurian, who took over as CEO in 2015, NetApp did something more interesting. It walked into the enemy camp and set up shop.

Today NetApp is the only data-storage company delivered as a first-party, native service inside all three major public clouds at once: Amazon FSx for NetApp ONTAP, Azure NetApp Files, and Google Cloud NetApp Volumes. Read that again. The three hyperscalers, who compete ferociously with each other and who could in principle build their own storage, instead resell NetApp's under their own logos. That only happens when the software is genuinely hard to replicate and customers keep asking for it by name.

The AI boom is usually narrated as a story about chips - who has the most GPUs, whose model is biggest. That narrative skips a step. Before a model can learn anything, its training data has to live somewhere, be governed, be moved to the compute, and not leak. That is a storage problem, and it is exactly the problem NetApp has been working on since before most AI startups existed.

The results are showing up in the deal count. Management said it closed roughly 200 AI infrastructure and data-lake modernization deals in a single recent quarter, up from about 125 the quarter before. That is not a rounding error; it is a category forming. NetApp's pitch to a company building AI is unromantic and effective: you already have decades of proprietary data, it is sitting on infrastructure like ours, and we can make it ready for your models without you shipping it somewhere risky.