Seagate (STX): what it does, how it makes money, and why AI still runs on hard drives

The standard $STX story is that hard drives are a dying technology, slowly eaten by flash. Then the stock rose about 322% in a year to 2026-09-11. Both can be true at once, and the reason is the most useful thing to understand about the company.

The more accurate frame: flash wins the fast, hot tier of storage, but most of the data an AI program keeps is not hot. It is the training sets, checkpoints, logs and archives that have to be kept cheaply at enormous scale, and for that the hard drive is still the cheapest place to put a byte. Seagate is one of two companies left that make those drives at scale. This piece walks through what Seagate sells, how it makes money, where it sits in the AI storage stack, and why the cycle has not gone away. Figures are as of 2026-09-11 unless noted.

Why it matters now

A 322% one-year move turns a value stock into a momentum stock, and the multiple changed with it: STX trades near 60x trailing earnings. The stock has come off about 11% over the last three months. The evergreen question is what holds up if the cycle turns: the structure of the industry, or only the price.

What does Seagate do?

Seagate designs and builds hard disk drives, mostly the high-capacity drives that go into data centers. Consumer and PC drives still exist, but the business that sets the valuation is mass capacity: very large drives sold to cloud providers and hyperscalers by the exabyte.

The structural read from QA's desk: every AI training program needs petabytes, and for the largest programs exabytes, of warm and cold storage behind the GPU layer. NVMe flash handles the hot tier. Hard drives handle almost everything else, because on cost per terabyte they still win by a wide margin.

The technology that matters is HAMR (heat-assisted magnetic recording), branded Mozaic 3+. It uses a tiny laser to heat each spot on the platter so data can be packed more densely. Seagate ships 3 TB per platter today, with 4 to 5 TB on the roadmap and eventually 10 TB. More terabytes per platter means more capacity per drive without a proportional rise in cost, roughly doubling capacity per drive every 24 to 36 months.

How Seagate makes money

Seagate sells drives, priced per unit and increasingly per terabyte of capacity, with most of the revenue now coming from data-center customers.

The part that changed the economics is industry structure. Hard drives have consolidated into a duopoly: Seagate and Western Digital. With only two large suppliers, supply is rationalised, capex is patient, and hyperscalers compete for capacity slots, so pricing is structurally less commodity-like than it was a decade ago. That shift, plus the HAMR ramp, is what the market has been paying for.

Where it sits in the AI storage stack

Seagate sits in QA's DRAM / HBM Memory bubble, the memory-and-storage cluster that moves with AI capex, and maps onto the Compute Capacity theme.

The useful way to place Seagate is by tier. At the top is high-bandwidth memory next to the GPU, the tightest bottleneck in AI (see HBM: the tightest bottleneck in AI), made by names like Micron ($MU). Below it is flash storage, where Sandisk ($SNDK) plays. At the bottom, holding the most bytes at the lowest cost, is the hard drive. Seagate's nearest rival there is Western Digital; see the WDC Q4 2026 earnings recap for its side of the same cycle.

Its peers make the duopoly visible. STX's tightest correlation in the QA universe (252 trading days, market beta stripped out) is Western Digital ($WDC, 0.88), far above anything else; among the others are Sandisk (0.68) and Micron (0.64). Owning STX and WDC together is much less diversified than two tickers suggest. The full peer table is on /stocks/stx.

The numbers

The Street mean target is an estimate, an average of analyst opinions, not a promise or a QA view. On a stock that re-rated this far this fast, it describes where the consensus sits, not where the stock goes.

The bull case

• Structural demand: AI programs store exabytes behind the GPU layer, and hard drives remain the cheapest place to put most of it.
• A duopoly with Western Digital, which makes pricing less commodity-like than in past cycles.
• HAMR: 3 TB per platter shipping, 4 to 5 TB on the roadmap, so capacity per drive keeps rising without a proportional rise in cost.

The bear case

• The cycle has not gone away. Storage demand is real but rides the same hyperscaler capex as $NVDA and Micron; a pause in AI infrastructure spend hits the whole hard-drive complex.
• Flash encroachment. As the cost of flash per terabyte falls, the upper edge of what is cheaper on hard drives shrinks. The duopoly buys time, not immunity.
• The multiple. Near 60x trailing earnings, the stock has re-rated from a value name to a cycle-momentum name, with reversion risk if HAMR shipments come in below cadence.

How to access

Seagate trades on the Nasdaq as STX, a clean US listing. To trade it from a US-retail account alongside the rest of the AI-infrastructure names, see /stack/ibkr. The live ETF holdings breakdown is on /stocks/stx.

Bubble-correlation shifts and rule-based alerts on $STX are part of /pro. For how memory and storage cycles still turn inside a supercycle, see memory cyclicality: the supercycle that still has a cycle.

What to watch

• HAMR shipment cadence and the 4 to 5 TB per platter roadmap.
• Hyperscaler capex guides, the demand line for mass-capacity drives.
• Pricing discipline across the duopoly, the key difference from past cycles.
• The desk has no curated level on this name yet; for the chart, use the live levels on /stocks/stx.

Live data on this ticker: /stocks/stx. Price, ETF holdings, bubble correlation, curated levels, bot positions.

Bubble context: /bubbles/memory. The cluster this name belongs to and how it's moving.

QuantAbundance is educational research. Nothing here is investment advice. See /disclosures.