Rating: Buy | 12-month Price Target: $720 | Current Price: ~$579 (post-earnings AH: ~$653)
Summary
- Seagate Technology ($STX) just reported a clean Q3 FY26 beat — revenue $3.11B (+44% YoY), non-GAAP EPS $4.10 vs. $3.48 consensus, gross margin a record 47%, operating margin 37.5%. The stock closed -2.8% in the regular session on high expectations but ripped +12.8% in after-hours as the order book detail landed.
- The Seagate AI HDD demand inflection 2026 is not cyclical noise. Nearline capacity is fully allocated through calendar 2026, build-to-order contracts now extend through the end of fiscal 2027, and customer discussions for 2028 are already active. CFO Romano on the call: "We have finalized our build-to-order for our fiscal '27."
- HAMR-based Mozaic 4+ began revenue shipments in late March. CEO Dave Mosley guided that Mozaic 4 will represent a majority of HAMR exabyte shipments exiting calendar 2026. The 40TB ramp is the lever that takes gross margin from 47% toward the 50% the bulls have been modeling.
- Bank of America and Cantor Fitzgerald independently moved to $700 price targets in the week before the print, citing structural HDD supply tightness and the HAMR transition. BofA's bull-case path runs to $45 EPS and 55% operating margin by calendar 2028.
- The bear case is real — flash substitution, hyperscaler in-housing, and the cycle — but each leg of it weakens when you read what hyperscalers actually ordered. Execution risk in AI-infra names is real (see POET's 47% crash for the cautionary case), but Seagate's order-book and contract structure put it on the opposite side of that curve. We rate STX Buy with a base-case 12-month target of $720. Live consensus and updated model targets on the STX forecast page.
Seagate's AI HDD demand inflection is structural, not cyclical (the thesis)
The tidy version everyone repeats — "AI generates data, HDDs are cheap, Seagate wins" — is correct but useless. It does not distinguish a one-quarter buying spree from a three-year capacity reset. The Seagate AI HDD demand inflection 2026 is the latter, and the evidence sits in the order book.
Three structural shifts have happened in tandem. First, the nearline HDD customer model has migrated from transactional to long-term agreements with build-to-order commitments 12–18 months out — hyperscalers (Google, Microsoft, Meta, Nvidia, Amazon) now sign multi-year supply assurance contracts because stocking out on storage mid-AI-training-cycle is operationally unacceptable. Second, the supply side has refused to expand unit capacity; Seagate and Western Digital are growing exabytes through areal density (higher TB per drive), not new factory floor. Third, the cost-per-TB gap between high-capacity HDDs and SSDs sits at ~16×, well above the 4× floor flash bulls model, meaning AI's warm and cold data has nowhere else to go.
The pricing environment proves the inflection. Enterprise HDD prices rose 40–50% in late 2025 and early 2026; high-capacity models are up over 120% YoY. CFO Gianluca Romano framed it on the Q3 call: "Revenue per terabyte has been remarkably stable in a period of rapid capacity growth." In a normal cycle, growing capacity drives ASPs lower. Here, supply discipline holds price flat while volume scales — that is what re-rates a hardware company into a structural growth story.
The thesis is simple: AI has permanently re-priced mass-capacity storage, the duopoly has chosen discipline over share grabs, and Seagate's HAMR lead means margin upside compounds. Q3 confirmed the inflection. The 2027 order book confirms the runway.
Q3 2026 earnings: beat the print, but the real story is the order book
The Q3 FY26 results released after the close on April 28, 2026 were a clean operational beat:
| Metric (Non-GAAP) |
Q3 FY26 |
Consensus |
YoY |
| Revenue |
$3.11B |
$2.98B |
+44% |
| Diluted EPS |
$4.10 |
$3.48 |
+115% |
| Gross margin |
47.0% |
~44% |
+9.5 pts |
| Operating margin |
37.5% |
~33% |
+14.4 pts |
| Free cash flow |
$953M |
— |
record |
Both gross and operating margin are all-time company highs. Q4 guidance is more telling: revenue $3.45B (±$100M), operating margin in the lower-40% range, EPS $5.00 (±$0.20). At midpoint, that implies sequential EPS growth of 22% — on top of the 32% sequential gain Q3 just delivered.
So why did the stock close -2.8% in the regular session? Two reasons. First, the run-in: STX had returned over 100% YTD 2026 on top of tripling in 2025, and BofA/Cantor Fitzgerald had each moved targets to $700 the week before. The bar was not "beat consensus" — it was "beat the whisper, which already priced beat-and-raise." Second, options had flipped sharply bearish into the print, with 30-day put/call ratios surging to 2.15 from 0.72.
The after-hours move tells the better story. Once investors had time to read the order book detail — 2027 visibility, Mozaic 4 ramp, 75% of leading global cloud customers qualified on the 4+ TB/disk product — STX rallied 12.8% to ~$653. The regular-session pullback was positioning; the after-hours move was the order book.
STX nearline sold out 2026 2027 — and 2027 orders already booked
This section requires no interpretation. The nearline market has moved from "tight" to "rationed."
The headline facts from the Q3 FY26 call and industry reporting:
- Nearline capacity allocated through calendar 2026 — every drive Seagate produces in CY26 has a customer name on it. No spot inventory.
- Build-to-order finalized through end of fiscal 2027 (June 2027). CFO Romano: "The vast majority of our nearline capacity is allocated during the next 4 quarters... we have finalized our build-to-order for our fiscal '27."
- Discussions with hyperscalers for CY28 demand are already active. Several CSPs have placed intent letters for 2028 capacity.
- Western Digital is in the same position — both major HDD producers are sold out on 2026 production.
- Average nearline drive capacity rose 22% YoY to ~23TB. Cloud customers buy significantly higher-capacity SKUs than the average — where Mozaic 4+ 40TB drives land.
BofA and Cantor Fitzgerald moved to $700 PTs ahead of the print specifically on order-book visibility. BofA's bull-case math gets to $45 EPS for CY28 with 55% operating margin — aggressive, but Seagate just guided lower-40% operating margin for Q4 FY26, with three more quarters of Mozaic 4+ mix shift to come.
The point is not that Seagate had a strong quarter. The point is that revenue is locked in for the next ~14 months at known volumes and improving ASPs, with visibility on a 14-month-plus tail from a customer base that physically cannot replace HDDs with anything else. Morgan Stanley, which named STX a Top Pick in early April, called the visibility "unprecedented." A thesis-driven view of STX must refuse the "cyclical hardware" frame — the cycle has been replaced by a contracted backlog.
Seagate HAMR Mozaic 4+ 40TB ramp: the technology that lifts gross margin toward 50%
If the order book is the volume story, HAMR Mozaic is the margin story. They compound: more drives at higher per-drive capacity at better per-TB cost takes gross margin from 47% to 50%+ over a multi-quarter ramp.
Mozaic is Seagate's branding for its Heat-Assisted Magnetic Recording (HAMR) drives. Mozaic 3+ shipped in volume in 2024 (3TB per platter, 30TB+ drives). Mozaic 4+ — qualified at two leading hyperscalers, with capacities up to 44TB — began revenue shipments in late March 2026. CEO Mosley on the Q3 call: "We began revenue shipments from Mozaic 4 in late March, and based on current ramp plans, we expect Mozaic 4 to represent a majority of our HAMR exabyte shipments exiting calendar 2026."
Ramp economics matter. Each HAMR generational step delivers ~25% higher capacity per platter at lower marginal cost — more terabytes from the same factory and head/media supply chain. That is the gross-margin expansion. Seagate's goal is HAMR at 70% of nearline exabytes by June 2027. The customer adoption breadth is underappreciated: 75% of leading global cloud customers are qualified on the 4+ TB-per-disk product. Hyperscale qualification is a 12–18 month process — once qualified, customers don't switch lightly.
The margin path quantitatively:
| Period |
GM (Non-GAAP) |
Driver |
| Q2 FY26 |
42.2% |
Pricing power, mix shift starts |
| Q3 FY26 |
47.0% |
Mozaic 4 begins shipping |
| Q4 FY26 guide |
~48–49% |
Implied from "lower-40% OM" guide |
| FY27 model |
49–51% |
Mozaic 4+ majority of mix |
| Bull CY28 |
55%+ OM |
BofA scenario, full HAMR mix |
Western Digital is going to a similar place via a different road — WDC has guided to 47–48% GM, matching Seagate. This complicates the "STX has a HAMR moat" story but reinforces the broader thesis: this is an industry-wide margin reset, not single-vendor share take. A duopoly where both prioritize price and capacity discipline is a more durable margin structure than one where Seagate stands alone.
For broader AI infrastructure context, see our Lam Research (LRCX) coverage on the wafer-fab equipment side, and the MU vs. SNDK memory analysis. Storage, memory, and capex equipment run on different gears of the same engine.
The bear case: flash substitution, hyperscaler in-housing, cycle exposure
A thesis that does not engage the bear case is a sales pitch. Three counterarguments deserve real answers.
1. Flash substitution. Argument: NAND falls on a Wright's Law curve until QLC/PLC SSDs replace HDDs in nearline. Counter: the cost-per-TB gap is ~16× today and consensus has it remaining >4× through 2030. Even on the most aggressive flash trajectory, the ~90% of AI-generated data that lives in warm and cold tiers economically must sit on HDDs. Flash takes the hot tier and cache. Verdict: real on a 5+ year horizon, manageable for the 2-year window.
2. Hyperscaler in-housing. Argument: Google's TurboQuant compression (disclosed late March 2026) reduces storage footprint per workload, and hyperscalers could design custom stacks consuming fewer HDDs. Counter: TurboQuant compresses logical data, but AI training datasets, checkpoints, and outputs grow faster than compression saves. Bernstein called TurboQuant "zero" impact on HDD demand; the post-disclosure sell-off lasted four sessions before the order book reasserted itself. Hyperscalers can compress. They cannot decline to store. Verdict: a recurring narrative risk that doesn't survive the order book.
3. Cycle exposure. Argument: HDDs are still hardware, ~$700B 2026 hyperscaler capex is unprecedented, and any pullback crashes the stack. Counter: this is the realest bear point, and it's why the LTA structure matters. If hyperscaler capex moderates in 2027, Seagate's 2027 production is already booked. The cycle risk is real for 2028 — bake it into the price target, not the rating.
A fourth flag: insider selling. CEO Mosley sold meaningfully through 2025 and early 2026 with no open-market purchases. Long-tenured executives diversifying after a 4–6× run is common, but it argues against pressing the position to maximum exposure.
Three-scenario valuation: Bull / Base / Bear price targets
We anchor on FY27 (ending June 2027) consensus EPS of ~$20.61 with sensitivity to execution and macro:
| Scenario |
FY27 EPS |
Fwd P/E |
12-mo PT |
Probability |
| Bull |
$24.00 |
~32× |
$760+ |
25% |
| Base |
$20.50 |
~26× |
$720 |
50% |
| Bear |
$16.50 |
~22× |
$480 |
25% |
Bull ($760+): Mozaic 4+ ramps faster than guided, gross margin clears 50% by late FY27, operating margin upper-40s. Hyperscaler capex sustains above $700B and grows in 2027. Multiple expands as STX re-rates as a pure AI infrastructure name (BofA's $45 CY28 EPS scenario).
Base ($720): Mozaic 4+ ramps on schedule. GM holds 47–49%, OM low-to-mid 40s. 2027 order book delivers as booked, no acceleration into 2028. Multiple modestly compresses as AI narrative matures, offset by clean execution. Our central case and the anchor for the Buy rating.
Bear ($480): Mozaic 4+ hits a yield issue or qualification setback; macro slowdown defers 2027 orders into 2028; or WDC closes the HAMR gap faster than expected. Margin stalls at 45%, growth decelerates to mid-teens. Multiple compresses to ~22× on the cyclical reframe.
Our 12-month PT is $720. From the $579 regular-session close (or ~$653 after-hours), that implies +24% to +10% base-case upside. Bear case ($480) is ~17% downside; bull ($760+) is ~31% upside. Risk-reward is positively skewed. Buy.
Frequently asked questions
Q1: Is Seagate (STX) a buy after Q3 2026 earnings?
Yes, with caveats. The Q3 FY26 print beat across every line (revenue $3.11B vs. $2.98B; EPS $4.10 vs. $3.48), and the order book — sold out through CY26, build-to-order finalized through end of FY27 — supports a structural thesis rather than a cyclical trade. The regular-session pullback was pre-earnings positioning; after-hours, the stock rallied 12.8% on the LTA detail. Base-case 12-month PT: $720.
Q2: Why does AI need HDDs instead of flash storage?
Cost-per-TB favors HDDs by ~16×, with the gap projected to remain >4× through 2030. AI workloads generate enormous training data, checkpoints, and generated content — ~90% of which is warm or cold data that does not require flash latency. Flash dominates the hot tier and GPU-adjacent cache; HDDs dominate everything else. This is a tiering story, not a transition story.
Q3: What is Seagate's HAMR Mozaic 4+ and why does it matter?
HAMR (Heat-Assisted Magnetic Recording) increases areal density on HDD platters. Mozaic 4+ delivers up to 44TB per drive (4+ TB/platter) and began revenue shipments in late March 2026. It matters because each generational step takes more terabytes off the same factory floor at lower marginal cost — that is the gross margin lever. CEO Mosley guided Mozaic 4 to majority of HAMR exabyte shipments exiting CY26. By June 2027, Seagate targets HAMR at 70% of nearline exabytes.
Q4: What is the bear case for STX?
Three concerns: (1) flash substitution long-term, (2) hyperscaler in-housing or compression breakthroughs reducing demand, (3) cycle exposure if hyperscaler capex pulls back. (1) is manageable in the 2-year window; (2) was tested by TurboQuant — order behavior did not change; (3) is the most legitimate but is mitigated by the LTA structure locking in CY26 and FY27. CEO Mosley's insider selling is a yellow flag worth monitoring.
Q5: How does Seagate compare to Western Digital (WDC)?
The two are a functional duopoly (WDC ~47% share, STX ~42%). WDC has guided to similar GMs (47–48%), compressing the STX premium gap. WDC is pursuing a dual-track ePMR/HAMR strategy and is ~2 years behind on HAMR scale. STX is the technology leader; WDC has shown faster margin recovery. For concentrated exposure, STX has the cleaner technology story; for diversified exposure, owning both makes sense.
Conclusion
The Seagate AI HDD demand inflection 2026 is the rare hardware thesis where the order book does most of the work. CY26 is sold out. FY27 is booked. Mozaic 4+ is shipping. Gross margin is 47% with a clear path to 50%. The bear case (flash, in-housing, cycle) is real but has answers. Valuation is full but not extreme on FY27 numbers, and the post-earnings price action — regular-session pullback, 12.8% after-hours rally — is the market repricing the order book.
Our rating is Buy with a base-case 12-month target of $720. Risk-reward is positively skewed; insider selling and cycle risk argue against pressing the position, but the structural setup is too clean to fade.
For broader AI infrastructure context, see our semiconductor sector overview, the parallel memory thesis in MU vs. SNDK, and capex equipment via LRCX. Storage is the third leg — and right now, the one with the longest visibility.
David Hartley is Senior Equity Analyst — Technology at Edgen Research, focused on data-center hardware, semiconductors, and AI infrastructure. This article reflects analysis as of April 28, 2026 and is for informational purposes only; it is not investment advice. Edgen and the author do not hold positions in STX as of publication.
