COMPANY DATA STORAGE / HARDWARE
The 55-year-old hard-drive maker bet its whole future on the spinning platter - right as the AI boom made cheap, bottomless storage the most valuable thing in the data center.
For most of the last decade, the smart money in tech treated the hard disk drive as a relic - a spinning, mechanical fossil waiting to be buried by flash memory. Western Digital, the company that helped make the hard drive a household object, seemed to agree. It had spent roughly $19 billion in 2015 to buy the flash maker SanDisk, apparently hedging its bets against its own core product. Then, in February 2025, it did the opposite of what the obituaries predicted: it gave SanDisk away.
The spinoff turned SanDisk into an independent, Nasdaq-listed company and left Western Digital as something it had not been in a decade - a pure hard-drive maker. To outsiders it looked like the company was keeping the dying half and discarding the future. Eighteen months later, the math looks different. In fiscal 2026, Western Digital - now rebranding itself simply as WD - reported revenue of $12.9 billion, up 36% year over year, with gross margins near 49%. And it had done something almost unheard of for a commodity hardware business: it sold out.
What WD actually makes
Strip away the jargon and WD does one thing: it manufactures hard disk drives - the sealed boxes of spinning glass platters and flying read-write heads that store information cheaply and in enormous quantities. The center of gravity is the enterprise nearline drive, the high-capacity workhorse that fills the racks of hyperscale data centers. These are the Ultrastar-class drives, sold by the pallet to a small circle of cloud giants, and they are where nearly all of the company's growth now comes from.
In 2026 WD began shipping 40-terabyte ePMR drives - energy-assisted drives that squeeze more data onto each platter - and laid out a roadmap toward 44TB and beyond using shingled recording (UltraSMR) and, further out, heat-assisted magnetic recording (HAMR). A single 40TB drive holds roughly eight million high-resolution photos. WD makes them in the millions.
Alongside the data-center business sits a smaller, more familiar consumer lineup: the color-coded WD Red, Blue and Black internal drives for home servers and desktops, the WD_BLACK drives marketed to gamers, and the My Passport and My Book external drives people buy to back up their laptops. That retail business is real, but it is now a rounding error next to the cloud - consumer sales make up only about 5% of revenue.
The distinction that matters most inside WD is not consumer versus enterprise but capacity per platter. Every generation of drive is a contest to fit more bits onto the same spinning surface without losing them, and each new recording technology - perpendicular, energy-assisted, shingled, heat-assisted - buys another few terabytes. That incremental grind is the whole business. A drive that stores 40TB where last year's stored 32TB is not a flashy launch, but to a hyperscaler filling a warehouse it is the difference between one row of racks and two.
Who buys it
WD's customer base has narrowed to something startlingly concentrated. Cloud hyperscalers - the handful of companies building the data centers that train and run AI models - account for roughly 89% of revenue. In practice that means WD's fortunes are tied to the capital-spending budgets of five or six firms. That is a genuine risk, and the company says so plainly. It is also, at the moment, the best possible risk to have: those five or six firms are the ones spending the most money on the planet.
The reason is unglamorous but simple. AI systems generate and archive staggering volumes of data - training sets, model checkpoints, logs, generated media - and most of it does not need to be read at flash speed. It needs to be stored cheaply, at scale, forever. That is exactly what a hard drive does best. As one framing of the storage supercycle put it, the cloud isn't fluffy; it's millions of spinning platters in warehouses in the desert.
That demand showed up in the numbers as pricing power - rare air for a commodity maker. In early 2026, reports surfaced that WD had effectively sold out its entire calendar-year hard-drive production before it was manufactured. Cloud revenue grew 43% to about $3.3 billion in the quarter, and the fourth quarter closed with revenue up 44% year over year and guidance above $4 billion for the next.
The competitive picture
There are exactly two companies left that make hard drives at meaningful scale: WD and Seagate Technology. Decades of brutal consolidation wiped out everyone else. That is usually described as a warning sign - a shrinking industry - but in 2026 it read as a moat. When demand exploded, hyperscalers had nowhere else to go. As of early 2026, WD held roughly 47% of mass-capacity shipments, edging out Seagate's 42%.
The two rivals play the same game differently. Seagate leans on technology progression and density leadership - being first to the next capacity milestone. WD leans on manufacturing consistency, operational execution and long, deep relationships with its hyperscale buyers. In a market where a data center just needs a lot of reliable drives, delivered on schedule, that second strategy has aged well.
The broader competitive threat - flash memory eventually undercutting disk - is real over the long run, and it is why SanDisk, Samsung, Micron and Kioxia matter to the story. But for bulk, cost-per-terabyte storage, flash remains far more expensive, and the gap is not closing fast enough to dislodge disk from the data center any time soon.
How the money works
WD's business model is deceptively plain: design drives, manufacture them at enormous scale, and sell them to cloud and enterprise customers. Revenue rises with the number of exabytes shipped and with the average selling price per terabyte - and that price, normally grinding downward, spikes upward when supply is tight. Fiscal 2026 was one long tight-supply cycle, which is how a hardware maker ended up with nearly ten points of margin expansion in a single year and about $3.5 billion in free cash flow.
The company trades publicly on Nasdaq under the ticker WDC. It carries no venture funding rounds in its recent story - this is a 55-year-old public company, not a startup - and its scale is measured in tens of thousands of employees and tens of billions in market value.
What makes the model powerful right now is leverage in both directions. Building drives is capital-intensive and slow to scale, so supply cannot snap to meet a demand spike - which is precisely why prices firmed when the AI buildout hit. The same rigidity that punishes WD in a downturn rewards it in a boom, and 2026 was firmly the latter. Free cash flow of roughly $3.5 billion gives the company room to fund its capacity roadmap and return cash to shareholders without leaning on outside money.
Where it came from
The most surprising thing about Western Digital is that it did not start out making hard drives at all. Founded in 1970 as General Digital by Alvin B. Phillips, backed in part by Emerson Electric, it began life as a maker of MOS test equipment and then semiconductors, adopting the Western Digital name in 1971. It did not enter the hard-drive business until 1988, when it bought Tandon's disk-drive assets and turned itself from a chip supplier into a branded hardware company.
From there it became a storage supermarket - drives of every kind, plus the SanDisk flash empire. The 2025 spinoff undid that sprawl on purpose. The company that emerged is narrower than it has been in years, and its new short-form name, WD, is a fitting signal: less breadth, more focus, the same original idea it started with - be the place data lives.
The bottom line
Western Digital sits at the physical foundation of the data economy - the layer beneath the software, beneath the models, where information actually rests. Its expertise is in precision manufacturing at scale: sealing helium into a box, floating a head nanometers above spinning glass, and doing it reliably a few million times a quarter. That skill was treated as a liability for years. In the AI era, it turned out to be one of the harder things in technology to replace.