Seagate sells high-capacity “nearline” drives and systems. Data Center was 80% of FY2026 revenue; the remaining 20% is Edge IoT (surveillance, NAS, consumer). The industry is a three-firm oligopoly of Seagate, Western Digital and Toshiba, and its only real substitute is NAND flash. Three things drive earnings: nearline exabyte volume, price per exabyte set under supply discipline, and cost per terabyte falling as HAMR raises capacity per disk. Cycle exposure is high. Revenue fell 37% in FY2023 and gross margin fell to 18%; FY2026 set records on revenue, exabytes and margin.
Cloud operators need somewhere cheap to keep data that must stay online but is rarely read: video, backups, logs, and data that AI systems keep and reuse. For that tier, cost per terabyte, including energy, decides the purchase; speed does not. Hard drives hold 87% of exabytes in large data-center deployments (IDC, cited in the FY2025 10-K).
Revenue follows exabytes, not drives. In FY2016 Seagate shipped 169 million drives, mostly for PCs, totalling 233 EB at about $61 per drive. In FY2026 it shipped 789 EB. Management now holds drive units flat while the heads and disks inside each drive rise; they grew 15–20% last year. Heads are made from Seagate’s own wafers and media is sputtered in-house. Drives are assembled in Asia, Northern Ireland and the US.
Since July 2023 nearline drives are built to order. Seagate secures orders before starting production, the contracts fix configuration and price for all of calendar 2027, and most nearline exabytes are allocated into calendar 2028. That matters because a wafer start takes three to four quarters to become a finished drive.
OEMs take 81% of revenue, and one customer took 14% in FY2026. The PC business was reclassified as “legacy” and starved of investment. The chip-design unit was sold to Broadcom in 2024 for $600M. Moving from a commodity product sold per unit to a capacity product sold per terabyte to buyers who plan years ahead explains most of the margin change.
Consolidation left three drive makers:
Components come from a thin, shared chain:
Seagate and Western Digital make their own heads and media, while Toshiba buys them in. The profit pool now sits with the drive makers. Seagate is number two by exabytes: in Q1 2025 Coughlin counted Western Digital at about 180 EB, Seagate about 144 EB and Toshiba about 41 EB.
The barriers that bind are recording know-how, the thin supply chain and hyperscaler qualification. Capital is not one of them: capex runs at only 4–6% of revenue.
Seagate’s specific edge is HAMR, where a laser heats a nanoscale spot so that smaller grains can be written:
That is a lead of roughly 12–18 months (inferred). It came about six years after Seagate first promised it, for 2019.
Flash is the substitute to watch. IDC put its per-terabyte premium at 5–10x in 2024. NAND contract prices then rose 33–38% in Q1 2026 and 70–75% in Q2 2026, and one hybrid-storage vendor (an interested party) now puts the gap at about 18–23x. Everpure has still won a second hyperscaler, and QLC flash uses about 30% less power. High NAND prices slow substitution; they do not stop it.
The winners in this industry have the lowest cost per terabyte, a steady density cadence, and the discipline not to add drive units. Seagate qualifies on technology timing and, since 2023, on discipline. It does not yet demonstrably qualify on cost.
FY2026 revenue rose 34%, and almost all of it was volume. Exabytes rose 33% and nearline exabytes 40%, while blended revenue per exabyte rose only about 1%. That flat figure hides two opposite forces. Like-for-like price per exabyte rose about 10% year on year in the June quarter, but the shift to 30–44 TB drives lowers revenue per terabyte. Acquisitions were immaterial, since Intevac cost $119M, so reported growth is organic apart from about 2 points from the 53rd week.
Gross margin rose from 35% to 46% because cost of revenue grew 13% while revenue grew 34%. Materials scale with capacity. Assembly, test and overhead scale with units, and those are flat. The CFO says incremental gross margin is “well above” 60%, and operating expense is only 8% of revenue.
The same fixed base cuts the other way. In FY2023 idle factories cost $75–79M a quarter in underutilisation charges, over 400 basis points of margin.
In FY2016 Seagate spent $587M of capex to ship 233 EB. In FY2026 it spent $569M (4.7% of revenue) to ship 789 EB.
| Metric | FY2016 | FY2020 | FY2023 trough | FY2026 |
|---|---|---|---|---|
| Revenue ($M) | 11,160 | 10,509 | 7,384 | 12,195 |
| Gross margin | 23.4% | 27.0% | 18.3% | 45.6% |
| Operating margin | 4.0% | 12.4% | -4.6% | 33.6% |
| Exabytes shipped | 233 | 442 | 441 | 789 |
| Free cash flow ($M) | 1,093 | 1,129 | 626 | 3,105 |
| Total debt ($M) | 4,091 | 4,175 | 5,451 | 3,565 |
FY2026 cash uses, ranked:
Management used the upturn first to repair a balance sheet stretched by the downturn. The dividend was never cut, and buybacks, paused since late 2022, are only now restarting with $4.8B of authorisation. Diluted shares fell from 302M (FY2016) to 207M (FY2023), then rose to 229M as the notes converted.
About 90% of exabytes go to data centers, so a few buyers’ capex budgets set demand. In Q1 2023 industry nearline unit shipments fell 54%, and Seagate lost $529M in FY2023. Price did not hold. Disclosed price per terabyte went $31 (FY2018), $22, $17, $15 and then $14 (FY2025), and Western Digital’s average drive price fell 17% in mid-2023. Price per terabyte falls every year, so profit depends on cost falling faster. In a demand air-pocket, price falls faster than cost can follow.
Where it sits now: at peak. Revenue of $12.2B is above the FY2022 peak of $11.7B. Exabytes of 789 are above the prior peak of 631. Margins are far above the prior decade high of about 30%. Lead times exceed 52 weeks, and capacity is allocated into 2028.
Build-to-order prevents a repeat of the 2023 inventory glut. It does not stop a buyer cutting volume or price at renewal, and Western Digital’s contracts use “predetermined or formula-based” prices. A storage-digestion pause, a NAND collapse, or Western Digital’s 2027 HAMR supply would hit renewal pricing first. With incremental margins above 60%, a modest price fall would remove a large share of gross profit (inferred).
| Indicator | Why | Where published |
|---|---|---|
| Nearline / data center exabytes and revenue | Volume engine | Seagate quarterly release and call |
| Price per exabyte; revenue per EB | Discipline at renewal | Seagate and Western Digital calls |
| HAMR share of nearline EB; Mozaic 4+/5+ | Cost per TB, capacity | Seagate calls |
| Western Digital non-GAAP GM; HAMR volume date | Does the lead convert? | Western Digital releases and 10-Q |
| NAND contract prices | Substitution pressure | TrendForce |
| Hyperscaler capex guidance | Demand budget | Amazon, Alphabet, Microsoft, Meta 10-Qs |
| Capex % of revenue; contract horizon | Supply discipline | Seagate 10-K/10-Q, calls |
The HAMR lead may never become a cost lead. If Western Digital ramps HAMR cleanly in 2027, Seagate’s one structural edge narrows before it has shown up in margins.
Reliability at fleet scale. Warranties issued rose from $68M to $114M in FY2026, and HAMR drives, with lasers in the head, are new to production fleets. A fleet-level problem would raise costs and hand share to Western Digital at requalification. The two risks compound.
Buyer power. One customer took 14% of Seagate’s revenue, and Western Digital’s top three customers took 44% of its revenue. Buyers’ leverage is muted while supply is short and returns at the first renewal after supply catches up.
Flash comeback. If NAND prices normalise just as power becomes the data-center constraint, QLC flash could move up into warm data and cap exabyte growth below the mid-20% target.
Legal and tax drag.
Not answerable from the sources:
The deciding event for any thesis has not happened yet: a hyperscale digestion under the build-to-order model.