Phase Space AI

Valuation

Seagate Technology [STX]

STX — Valuation

Spot $764.20 (2026-07-29). Diluted shares 232m. Net debt $1,861m. EV $179,155m.

---

1. COMPANY STATE — declared first

STATE B — cyclical / commodity-sensitive, with an argued upward reset of the mid-cycle level

Test satisfied: revenue and margin are driven by an exogenous cycle (hyperscaler storage capex and nearline supply/demand balance). The evidence is unambiguous: FQ1'24 gross margin 10.2% and operating margin −8.9%; FQ4'26 gross margin 52.3% and operating margin 43.0%. A 42-point gross-margin swing in eleven quarters is the definition of State B. This is the same state as MU, SNDK, LRCX and the semicap complex.

State evidence: 81 quarters of AV normalized statements plus 13 transcripts plus the FY2026 8-K filed today. Ten-year gross-margin range 10.2%–52.3%; ten-year operating-margin range −16.9% to +43.0%.

Consequence (mandated by valuation.md State B): terminal margin must be mid-cycle, normalised on BOTH revenue and margin. A trailing-peak margin with a cycle-average growth forecast is internally inconsistent and always flatters.

The cycle must be DATED, not merely normalised. It is dated here explicitly:

• Seagate is at or very near a cycle peak in margin — FQ4'26 operating margin 43.0% is an all-time record and FQ1'27 guidance implies higher again.
• Seagate is early in a volume cycle — nearline capacity allocated through CY2027, LTAs at the peer extending into CY2029, lead times ~12 months, exabyte demand CAGR guided >20%/mid-20s%.
• Own-multiple position: trailing P/E is 50.0x, the 94.4th percentile of 2,408 daily observations of Seagate's own ten-year history (median 9.8x, p75 15.5x). Every observation in the top decile was created inside the last twelve months — measured to twelve months ago, the maximum ex-trough realised-forward P/E was 15.5x. This is precisely the pattern valuation.md flags on the semicap names.

So: peak margin, early volume cycle, top-percentile multiple. The forecast begins at a margin peak. That is stated, not buried.

Why State B does not mean reflexive mean reversion here

The State-B instruction is mid-cycle, and the honest reading of the evidence (Research §8) is that the mid-cycle level has reset upward, because the mechanism that historically converted margin recovery into price collapse — capacity investment — has been explicitly withdrawn ("we are not making any investments in adding unit capacity", WDC CEO; Seagate capex 4.7% of revenue) and replaced by contracted volume at defined prices. Mean-reverting to the 2016–2024 average operating margin of roughly 10–14% would be applying a bound calibrated on a supply regime that no longer exists — the exact error the brief names.

Equally, extrapolating 44.6% is extrapolating a peak.

---

2. TERMINAL MARGIN — derived, bridged, and defended

Terminal (FY2031) operating margin: 35.0%

Basis: the company's own demonstrated margins across the full cycle AND its current trajectory.

35.0% sits ABOVE the trailing full-year GAAP actual (33.6%) and 1.5pp below the trailing full-year non-GAAP actual (36.5%). It is ~9.6pp below the exit-rate non-GAAP margin.

This satisfies the brief's test directly: the terminal margin is not below the company's own trailing actual. The 9.6pp discount to the exit rate is the mid-cycle normalisation State B requires — applied to the reset level, not to the pre-2025 level. Had I normalised to the 2016–2024 mean (~12%), the model would have carried a terminal margin 21.6pp below the trailing actual, which is the AVGO/MSFT/XZO defect class the brief names (21 of 84 covered names).

Opex bridge — the real test

m_EBIT,T  =  m_gross,T − R&D − S&M/G&A − other
    35.0% =    44.0%   −  4.5%  −   4.0%    −  0.5%

• m_gross,T = 44.0%. FY2026 actual 45.6%; FQ4 exit 52.3%. 44.0% is a 1.6pp discount to the FY2026 full-year actual and an 8.3pp discount to the exit rate. It is ~10pp above the pre-2025 decade mean (~34%) — justified by (a) permanent mix shift to 40TB+ HAMR drives (70% of nearline exabytes on HAMR by FY2027-end, roadmap to 10TB/disk), and (b) revenue/TB no longer declining.
• Opex 9.0% of revenue. FY2026 non-GAAP opex was $1,170m = 9.6% of revenue, and it is near-fixed: it grew 3.7% y/y (from $1,128m) while revenue grew 34%. FQ4 non-GAAP opex was $293m = 8.1% of revenue. At FY2031 revenue of ~$28bn and opex compounding at 6%/yr from $1,170m → $1,565m = 5.6% of revenue. 9.0% is deliberately conservative — roughly 1.6x the mechanical bridge — to absorb a step-up in R&D for post-HAMR roadmap and any reversal of the FY2026 opex leverage.
• Hard constraint check: m_EBIT,T (35.0%) ≤ m_gross,T (44.0%) ✓, same basis, same period.
• Sufficiency check: 44.0% gross leaves 9.0pp of revenue for all R&D, selling and administrative cost. FY2026 actual was 9.6pp at a lower revenue base. Economically feasible with headroom. ✓

terminal_margin_source: company's own FY2026 full-year actual (33.6% GAAP / 36.5% non-GAAP) as the lower anchor, FQ4'26 exit rate (44.6%) as the upper anchor, set between them at 35.0% with a mid-cycle discount to the exit rate. No industry percentile, no peer median, no cap applied.

---

3. EXIT MULTIPLE — derived from the identity, not asserted

EV_T / EBIT_T = (1−t)(1−g/ROIC) / (WACC−g)

(1 − 0.16) × (1 − 0.04/0.40) / (0.10 − 0.04) = 0.84 × 0.90 / 0.06 = 12.6x

Warranted terminal EV/EBIT = 12.6x.

Sanity check against the traded set: STX's own current EV/EBIT is 40.3x on FY2026 non-GAAP EBIT and 24.5x on the FQ1'27 annualised run-rate. The comparator set (WDC 34.9x forward P/E, MU 4.7x, SNDK 5.6x) spans 7.4x with no growth ordering and therefore cannot discriminate a multiple — declared UNIDENTIFIED per valuation.md rule 6. The identity is the only defensible source, and the implied compression from today's 24.5x forward EV/EBIT to a 12.6x exit is −48.6%. That compression is stated as a number because the framework requires it, and it is the dominant driver of the result below.

---

4. IMPLIED-PATH TEST (the Valuation Criteria)

Solved parameter: revenue CAGR. Held fixed: terminal margin 35.0%, exit EV/EBIT 12.6x, WACC 10.0%, horizon 5 years, FCF margin 26.0%, revenue base $12,195m (FY2026 actual), net debt $1,861m, diluted shares 232m.

Terminal value is ~85% of EV under this construction, so the reverse DCF is mandatory and primary.

The FCF-inclusive figure is the one to use — Seagate converted 25.5% of FY2026 revenue and 30.8% of FQ4 revenue to free cash flow, so ignoring five years of interim FCF (as the shipped script does) overstates the required CAGR by ~5.1pp. This is a finding about the tool, not about Seagate.

The margin: demonstrated − required

Read this honestly. The price requires Seagate to sustain, for five consecutive years, a revenue growth rate it has achieved exactly once — in the single best year of the best cycle in its history — and which management itself guides to 20%. On management's own number the gap is −14.8pp. The implied-path test is FAIL.

Sensitivity over the exit multiple (never over probabilities)

Equivalently, holding growth and solving for the multiple:

At management's own 20% growth target, today's price requires a 24.0x terminal EV/EBIT — 1.9x the warranted multiple. That is the whole valuation statement in one line: the price is not asking Seagate to grow, it is asking the market to still be paying a growth multiple in 2031.

---

5. TWELVE-MONTH TARGET

The multiple anchor is UNIDENTIFIED — and the reason is a regime change, not missing data

valuation.md requires the 12-month multiple to be anchored on the name's own trading history with the percentile stated, and requires UNIDENTIFIED where the history spans a regime change. Both conditions bind:

• Own trailing P/E today: 50.0x = 94.4th percentile of 2,408 daily observations. Median 9.8x.
• Own realised-forward P/E (price at report date ÷ the next four quarters of actual EPS), which removes the trough-earnings artifact: 15.5x (Jul-2024), 13.5x (Oct-2024), 10.9x (Jan-2025), 7.9x (Apr-2025), 12.2x (Jul-2025). Median of the clean observations ≈ 12x.
• Today's forward P/E on guided FQ1'27 annualised EPS: 26.2x.

Seagate is trading at roughly 2.2x its own historical forward multiple. Two regime changes make the history non-comparable: the shift from spot to build-to-order/allocated pricing, and the HAMR cost curve. The historical anchor is declared UNIDENTIFIED. It is reported, not used.

Target built forward, with the multiple assumption stated as the assumption it is

Earnings path (non-GAAP diluted EPS; FQ1'27 is guided, the rest is a modest sequential progression consistent with mid-20s% exabyte growth and high-single-digit $/TB):

In twelve months (July 2027) the market prices FY2028.

Base multiple justification, stated explicitly: 24.0x is (a) an ~8% de-rating from today's 26.2x forward multiple, (b) the multiple that today's price already requires as a terminal exit multiple at management's 20% growth target (24.0x — §4), so it is internally consistent rather than an independent parameter, and (c) far above any percentile of Seagate's own history — it is off the top of the distribution and is declared as a regime-change assumption, not a historical anchor.

12-month target: $860 (+12.5%).

Cross-check on cycle position: the stock's 252-day peak close was $1,093.26 on 2026-06-22 and it is −30.1% from that high as of today, with MU −12.8% today alone and SNDK −9.5%. The base target is 21.3% below the June high. A +12.5% target off a 30% drawdown is a modest statement, and it is the honest one: the 5-year implied path fails by 14.8pp against management's own growth target, so the 12-month upside is entirely a multiple-persistence bet, not a valuation gap.

Sanity band against the external professional target — REQUIRED, and the divergence explained

EARNINGS_ESTIMATES returned empty, but AV OVERVIEW carries AnalystTargetPrice:

Per valuation.md, this is a check on the output, never a calibration target. The divergence is explained, not tuned away:

1. The Street target is stale relative to the drawdown. STX peaked at $1,093.26 on 2026-06-22 and is −30.1% from there. A $1,016 target is 7% below the June high — it is consistent with a target set before the last five weeks, not with a view on today's price.
2. The house number embeds a multiple de-rating that the Street does not. $1,016 on FY2028E EPS of $35.80 is 28.4x — i.e. the Street is holding today's 26.2x forward multiple and letting earnings carry the target. The house uses 24.0x, an 8% de-rating, for the reason given in §4: at 20% growth today's price already requires a 24.0x terminal multiple, so 24x is the internally consistent figure and 28.4x is not.
3. Direction of the check: the house is below Street by 15.4% but still above spot by 12.5%. Item B16 warned that 16 of 16 house targets below spot expresses a market view rather than a valuation. This target is above spot, so that failure mode does not apply here.

The gap is a multiple disagreement of 4.4 turns, fully attributable and stated.

---

6. Criteria summary

No position verdict is issued. The book decides.