Seagate Technology Holdings plc STX

Key data

The business

Seagate sells hard disk drives, and increasingly one kind: high-capacity "nearline" drives that sit in cloud data centers holding data nobody touches often but nobody deletes. It shipped 789 exabytes of drive capacity in fiscal 2026. The edge is cost per bit. Its Mozaic drives, built on heat-assisted recording, cram more terabytes onto each platter than rivals can, so the drive that costs about the same to build sells more storage. That is the whole moat: in a shrunken field of three makers, be the one whose bits are cheapest. The customers are the hyperscalers, and the concentration is real, three separate customers each account for more than 10% of receivables. The money is made once per drive, and the margin widens when capacity per drive climbs faster than the cost to add it. It is also violently cyclical: when cloud buyers over-order then digest, both revenue and margin swing hard, as fiscal 2023's loss showed.

The numbers

The recovery is not subtle. Revenue climbed every quarter of fiscal 2026 and net income nearly tripled across the year.

The June quarter was the strongest of the run, revenue up 17% on the March quarter and profit doubled. Revenue kept climbing and full-year gross margin held at 46%, so the June quarter was the turn, not the top, so far.

Zoom out and the cyclicality is the whole story.

Fiscal 2026 revenue just cleared the old 2022 peak, but earnings are nearly double that peak, because the balance sheet and cost base are leaner than they were then. How management ran cash through the trough is the tell that matters most for a cyclical.

They cut buybacks sharply in fiscal 2023 then stopped entirely for two years to protect the balance sheet, cut debt by $1.4 billion this year, and paid $634 million in dividends. Free cash flow was about $3.1 billion. Against a $190 billion market value that is a 1.6% cash yield, so almost none of the price is paid-for-today cash, it is paid-for-tomorrow growth. The market has set consensus fiscal 2028 earnings near $56, roughly four times this year's, and that is the bet in one number. What this memo believes and the tape does not settle: whether per-drive pricing holds as capacity per drive ramps. Watch nearline pricing and gross margin together next year. Hold both and the four-times ramp is real and 15x is cheap. Lose pricing as capacity climbs and this is a peak.

Management

CEO William David Mosley runs a company that behaved well in the downturn, buybacks slashed then zeroed when cash was tight, debt paid down first, dividend protected. The record on capital is the good news. The insider record is not: zero buys and 95 sales worth $43 million over the last year, and Mosley himself sold about $9.5 million across three transactions on September 1, days after the annual report landed and with the stock near a record. Plan status on those sales is not disclosed in the feed, so read them as neither reassuring nor damning, but they are what you expect near a high, not a bottom. Pay is tied to the same non-GAAP earnings the company has beaten four straight quarters, which flatters the scorecard. Nothing here says insiders think the stock is cheap.

How it fails or surprises you

Hyperscaler digestion. Three customers each above 10% of receivables means a small buyer group controls the order book. If cloud capacity buying pauses and they run down inventory, revenue can roll and gross margin drop under 45% inside two quarters, the same pattern that produced 2023's loss. The print that shows it first: sequential revenue and quarterly gross margin.

The AI storage supercycle (right tail). If AI workloads keep the data pile compounding and Mozaic yields keep climbing, the $56 consensus for fiscal 2028 could prove low, and the stock at 15x that number is cheap, not stretched. The market is paying a high trailing multiple but not fully pricing a multi-year ramp. Print: exabytes shipped rising while per-drive pricing holds.

The peak-on-peak risk. The fact this read explains least: you pay 61 times earnings on what may be peak earnings, with only a 1.6% cash yield underneath. Fiscal 2022 also looked like a new plateau right before revenue fell a third. The print that would prove the bull wrong: the first quarter-on-quarter revenue decline.

Closing thoughts

This is a name where one thing settles the argument. The stock at 61 times trailing but 15 times fiscal 2028 says the whole case rides on the earnings ramp being real, and next year's nearline pricing against next year's gross margin will tell you which world you are in before the income statement does. The tails are both fat and pointed opposite ways: this is a deep cyclical where the down move is fast and unforgiving, a $529 million loss two years ago, and where the up move, if AI storage demand compounds, makes today's multiple look modest. What is at risk if digestion hits is roughly the last two years of gains giving back; what is on offer if the supercycle holds is earnings the market is paying only 15 times for. Judgment, not a number: the left tail is the one to respect, because the customer base is small and the history is brutal.

The bet is still that cloud companies keep needing somewhere cheap to store an exploding pile of data, and that spinning hard drives stay the cheapest place to put it. It breaks if the hyperscalers pause to digest, or if flash falls far enough to steal the cold-storage job, and the one pair that tells you first is sequential revenue and gross margin read together. Hold above 45% margin with revenue still climbing and the ramp is intact. Slip below it as revenue stalls and June was the top, not the turn.