💾

STX

💡
Thesis. Half of a two-supplier duopoly in nearline hard drives, growing 48% with a 52.7% gross margin — and the vast majority of nearline capacity is already committed into calendar 2028 under long-term agreements, with price per exabyte guided to +20%.
Reviewed 19 Sep 2026, after fiscal Q4 2026 results on 28 Jul. Next review after fiscal Q1 2027 earnings on 21 Oct 2026.

What it does

Seagate makes hard disk drives, specifically nearline HDDs — the high-capacity mechanical drives hyperscalers use for bulk datacentre storage. It is effectively a duopoly with
💾
WDC. Mechanical, not semiconductor: different physics from flash, same customers.
Two terms drive the numbers. Exabytes shipped is the real volume metric — a drive count means nothing when capacities differ. HAMR (heat-assisted magnetic recording, sold as Mozaic) is the technology transition that keeps pushing capacity per drive up, and with it keeps HDD cost-per-bit below flash for bulk storage.
The vast majority of nearline exabytes are already allocated into calendar 2028, and customers are asking to plan into 2029. This is no longer a spot market.

Bull case

  • Fiscal Q4 2026 revenue $3.6B, +48% YoY and +17% sequentially. Non-GAAP EPS $5.71, +121% YoY
  • Non-GAAP gross margin 52.7% (+570bps sequentially); operating margin 44.6% (+710bps) — a thirteenth consecutive quarter of margin gains
  • Free cash flow $1.1B, the strongest quarter in over a decade. FY2026 free cash flow $3.1B at a 31% margin
  • 195 exabytes shipped to datacentre customers, +43% YoY. Datacentre revenue $2.9B; Edge/IoT $697M
  • Pricing is accelerating, not just volume. Price per exabyte grew ~10% YoY in Q4 and is guided to ~20% in Q1 FY2027. CEO Dave Mosley: "The gap between supply and demand is now a little bit bigger" than expected, allowing premium pricing on incremental capacity outside long-term contracts
  • Mozaic (HAMR) drives are now 40% of nearline exabyte shipments. Mozaic 4 ramping with major cloud providers; Mozaic 5 on track for late calendar 2027 qualification
  • Q1 FY2027 guided to $4.1B ±$100M (+56% YoY), $7.30 ±$0.20 non-GAAP EPS, ~50% operating margin, with sequential revenue and margin expansion through the year
  • Balance sheet repairing fast: $300M of debt retired in Q4, $1.2B more planned in Q1, net leverage 0.4x, capex only 4–6% of revenue
  • New demand drivers named: KV cache for agentic AI, and physical AI in robotics and autonomous vehicles

Bear case

  • Long-term SSD substitution.
    🧠
    SNDK    's BiCS10 and QLC roadmaps push flash cost-per-bit down every generation. The HDD cost advantage is real today and erodes over time
  • A duopoly's pricing depends on both players holding capacity discipline. If either adds aggressively, the pricing evaporates — and the current margins are exactly the incentive to add
  • Elevated multiple. It has re-rated further than
    🧠
    MU     on trailing earnings
  • Trades as part of the memory complex and moves with it on days when nothing HDD-specific happened
  • Peak-cycle risk. A 52.7% gross margin and 44.6% operating margin in a mechanical storage business are historically extraordinary, and multiples applied to peak margins are the classic way to lose money in this sector
  • Committing capacity into 2028 locks in volume but also locks out repricing if scarcity worsens further

Major customers

  • Hyperscalers — Microsoft, Amazon,
    🏛️
    GOOG     and
    🏛️
    META     buying nearline drives for bulk storage — around 90% of segment demand
  • Cloud and colocation providers
  • Enterprise storage OEMs — Dell, HPE, NetApp
  • Video surveillance and consumer — a shrinking legacy share (Edge/IoT, $697M, 19% of revenue)
🔗
Read-through. Seagate and
💾
WDC are a pair — same duopoly, same customers, same exabyte cycle — so they should be sized as one position, not two. Both benefit from hyperscaler capex the way
🧠
SNDK does, but on the mechanical tier: if flash economics improve faster than expected the HDD names lose share to Sandisk, and if AI data volumes outrun flash supply they win. The most useful cross-read is the pricing disclosure. Seagate saying the supply-demand gap is widening, with price per exabyte going from +10% to +20%, is the same story
🏛️
NVDA told from the buyer's side when it guided gross margin down on memory costs — two independent confirmations that storage and memory scarcity is intensifying rather than easing.

What would change the view

  1. Exabyte shipment growth and price per exabyte — currently +43% and heading toward +20% respectively. Both, together
  2. HAMR/Mozaic ramp beyond 40% of nearline exabytes, and Mozaic 5 qualification in late 2027
  3. Long-term agreement coverage extending into 2029, as customers have requested
  4. Gross margin holding above 50%
  5. Any capacity discipline break by either duopolist — the fastest way this thesis ends
  6. Flash cost-per-bit progress at Sandisk and the other NAND makers

Update log

19 Sep 2026 — Converted to the standard template and the page's own "limited detail captured — pull the release" flag resolved. Fiscal Q4 2026 now carries the full picture: 52.7% non-GAAP gross margin, 44.6% operating margin, 195 exabytes to datacentre customers (+43%), Mozaic at 40% of nearline exabytes, $1.1B of quarterly free cash flow, Q1 FY2027 guided to $4.1B and $7.30 EPS, and long-term agreements covering the vast majority of nearline exabytes into calendar 2028. The Backlog property, which had said "not captured in this refresh," is rewritten.
28 Jul 2026 — Fiscal Q4 and FY2026 reported. Q4 revenue $3.6B (+48% YoY, +17% QoQ); non-GAAP EPS $5.71 (+121%); non-GAAP gross margin 52.7%; operating margin 44.6%; free cash flow $1.1B. FY2026 free cash flow $3.1B (31% margin). Datacentre revenue $2.9B on 195 exabytes. Price per exabyte +10% YoY, guided ~+20% for Q1. Q1 FY2027 guided $4.1B ±$100M and $7.30 ±$0.20. $300M of debt retired, $1.2B more planned; net leverage 0.4x.
Research and education only — not investment advice.