Seagate Technology Holdings STX

Reference table

1. What the business is, and how the disclosure actually works

Seagate describes itself in the 10-K as a supplier of mass capacity and legacy data storage solutions. The corporate structure matters more than usual here, so it is worth stating precisely from the filing: Seagate Technology Holdings plc is "a public limited company organized under the laws of Ireland", with principal executive offices at 121 Woodlands Avenue 5, Singapore — the 10-K notes that "The Company established Singapore as its principal executive offices in fiscal year 2024." The vendor profile field records the country as SG. The debt is issued out of two subsidiaries, Seagate HDD Cayman and Seagate Data Storage Technology Pte. Ltd., with cross-guarantees. Chief executive William David Mosley; chief financial officer Gianluca Romano; approximately 30,000 employees (vendor profile field). Founded 1978; listed on NASDAQ; ISIN IE00BKVD2N49, CUSIP G7997R103 — both Irish-issuer identifiers.

One reportable segment, and the vendor gives you nothing

This is the most important disclosure-structure fact in the file, and the vendor payload handles it badly.

The 10-K, Note 15, states: "The Company has determined that its Chief Operating Decision Maker ('CODM'), the Chief Executive Officer, evaluates performance of the Company and makes decisions regarding investments in the Company's technology platforms and manufacturing infrastructure based on the Company's consolidated results, including net income reported on the Consolidated Statements of Operations. As a result, the Company has concluded that its manufacture and distribution of storage solutions constitutes one operating segment."

The vendor's seg_prod block is EMPTY — an empty array. There is no product revenue disaggregation in the payload at all. That is not a defect so much as a consequence: with one reportable segment, the vendor has nothing to populate. But it means the entire mass-capacity versus legacy split — the single most important mix question about this company — is unavailable from the vendor and must come from the filings. The 10-K describes the markets, not the dollars:

• Mass Capacity Storage — "supports scalable, high capacity, cost efficient storage applications, including nearline cloud, nearline enterprise, video and image applications ('VIA') and network-attached storage ('NAS') and edge-to-cloud data storage infrastructures." This is enterprise nearline HDDs, enterprise SSDs, VIA and NAS drives, systems, and the Lyve edge-to-cloud platform.
• Legacy Markets — "include consumer, client and mission critical applications. We continue to sell to these markets but do not plan significant additional investment." Seagate is explicitly harvesting this.

The dollar split between them is not in this archive. The 10-Q says the company changed its end-market classification during FY2026 and "reflected these changes to our revenue and HDD exabytes shipped by end market retrospectively to the earliest period presented", adding that "The change had no impact on our previously reported consolidated net revenue." The table itself carries the numbers and the numbers are stripped from the extracted text. We state this as a gap rather than estimating around it.

The geography table is not what it looks like

seg_geo supplies four lines: United States, Singapore, Netherlands, Other Countries. They sum exactly — $9,097M for FY2025, $6,551M for FY2024, $7,384M for FY2023, matching reported revenue to the dollar in every year. The arithmetic is clean and we verified it line by line.

But the 10-K's own footnote destroys the natural reading: "Revenue is attributed to countries based on bill from locations." These are not customer geographies. They are the legal entities that issue the invoice. Seagate manufactures in China, Malaysia, Northern Ireland, Singapore, Thailand and the United States, and sells to hyperscale customers concentrated in the United States — and this table tells you nothing about either. Read naively it would suggest Seagate's demand base shifted 13 percentage points from Singapore to the United States between FY2024 and FY2025, which is an invoicing decision, not a demand shift. We present it, we do not interpret it, and we flag the label as misleading in Section 8. There is also no FY2026 row: the series stops at FY2025.

Customer concentration — lower than the industry reputation, and disclosed annually only

The 10-K states: "In fiscal year 2025, one customer accounted for approximately 10% of consolidated revenue. In fiscal year 2024, no customers accounted for more than 10% of consolidated revenue. In fiscal year 2023, one customer accounted for approximately 10% of consolidated revenue." Separately, under Concentrations: "The Company's customer base is concentrated with a small number of customers... One customer accounted for more than 10% of the Company's accounts receivable as of June 27, 2025 and June 28, 2024."

That is materially less concentrated than the semiconductor capital-equipment names in this batch, where single customers run to 19% of revenue. It is still concentrated in a small number of hyperscale buyers — the risk factors name "large hyperscale data center companies and CSPs" explicitly — but the disclosed 10% threshold is a fact and we report it as one. The caveat is timing: this disclosure is annual, the FY2026 10-K is not in this archive, and FY2026 revenue grew 34% on hyperscale nearline demand. The concentration figure we are quoting is thirteen months old and the direction of travel is almost certainly upward.

Competitors, and what the list tells you

The 10-K's principal-competitor list, verbatim in order: Kioxia Holdings Corporation; Micron Technology, Inc.; Samsung Electronics; Sandisk Corporation; SK hynix, Inc.; Toshiba Corporation; and Western Digital Corporation.

Only two of those seven — Toshiba and Western Digital — make hard disk drives. The other five make flash and memory. Seagate's own annual report is telling you that its principal competitive threat is substitution rather than rivalry. Hold that thought; it recurs in Sections 6c and 11, and it is the single best reason the exponential score is a 5.

2. The one thing to understand about these numbers: this is a price cycle, and the filings say so

FY2026 is the best year in Seagate's history by every margin measure, and roughly half of it is price.

Three facts follow from that table and they are the whole argument of this dive.

First, the amplitude is extraordinary in both directions. Revenue fell 43.8% peak to trough between FY2022 and FY2024. Operating margin ran from +16.8% to −4.6% to +33.6% inside four years — a 3,820 basis point swing. Josh Brown's single knowledge-base claim on this name says precisely this: "wildly swinging operating margins." He is describing the table above.

Second, the five-year trend line is flat. FY2021 revenue $10.681B; FY2026 revenue $12.195B. That is a 2.7% compound annual rate over five years, and FY2026's total is only 4.6% above FY2022's $11.661B. This company's revenue oscillates around a level; it does not compound. Any valuation that treats FY2026 as a new base rather than a cycle peak is making a claim that has never been true in this file.

Third — and this is the part that could genuinely be different — the FY2026 margin is not a volume effect. From the 10-Q filed 2026-04-29, verbatim:

> "Gross margin for the March 2026 quarter increased by 5 percentage points compared to the December 2025 quarter primarily driven by pricing actions undertaken by the Company."

> "Gross margin for the three and nine months ended April 3, 2026 increased by 12 percentage points and 9 percentage points from the three and nine months ended March 28, 2025, respectively, primarily driven by pricing actions undertaken by the Company, favorable volume and product mix, and $10 million of restructuring costs related to an inventory write down that did not recur."

> "Revenue in the March 2026 quarter increased by $287 million compared to the December 2025 quarter, primarily due to favorable pricing actions undertaken by the Company and an increase in exabytes shipped reflecting higher demand for nearline products."

And the volume data lets us split it. The 10-Q overviews give exabytes shipped directly:

From December to March, exabytes rose 4.7% and revenue rose 10.2%. Volume explains slightly less than half the sequential growth; price and mix explain the rest. This is a pricing cycle with a volume tailwind underneath it, not a volume cycle. That matters enormously, because volume is sticky and price is not.

A second, independent price lever confirms it. The 10-Q: "Sales incentive programs were approximately 10% of gross revenue for the March 2026 quarter, 11% for the December 2025 quarter and 14% for the March 2025 quarter." Seagate has withdrawn four percentage points of rebate from its channel in a year. That flows straight to net revenue and it is available only while supply is tight.

The June quarter, where we have less to go on. Revenue rose 16.6% sequentially and gross margin rose 5.8 points to 52.3%. The exabyte figure for that quarter is not in this archive — it would have been in the 8-K's Exhibit 99.1, which was not extracted (Section 8, finding 1). Given the March quarter's decomposition and the sequential margin step, it is highly likely that price again explains the majority. We say "likely" rather than asserting it, because we cannot verify it.

The point of this section. The FY2026 result is real, clean and impressive, and Sections 3 and 4 will show it produced real cash and a genuinely repaired balance sheet. But it is a price result in a commodity whose 10-K contains a section headed "Price Erosion", and the whole valuation question is whether the price holds.

3. Growth and margin trajectory

Quarterly, from the vendor income series, every revenue line cross-checked against earn_cal actuals:

Four consecutive beats, and they are getting larger: +6.5%, +8.8%, +11.9%, +16.8%, +12.0%. The June quarter beat revenue by 3.7% ($3.629B against $3.498B) and adjusted EPS by 12.0%. Sequential revenue growth accelerated all year — 7.6%, 7.5%, 10.2%, 16.6% — and gross margin rose in every single quarter for eight consecutive quarters, from 31.8% to 52.3%. That is a 2,050 basis point gross-margin expansion over two years. We have not seen a series like it elsewhere in this batch.

Now the honest test of that ramp — is it clean operating earnings? The brief for this dive flagged the earn_cal sequence (2.59 → 2.61 → 3.11 → 4.10 → 5.71) as violent enough to warrant checking for a tax benefit or a one-off gain. We tested it four ways and it survives.

Test 1 — the tax line. The effective tax rate went the WRONG way for a flattered result:

FY2026's earnings were achieved DESPITE the effective tax rate rising from 2.9% to 13.7% — a 1,080 basis point headwind worth roughly $400M of net income. If the FY2025 rate had persisted, FY2026 EPS would have been materially higher. The ramp is not a tax artefact; it is a tax-suppressed operating result.

Test 2 — the valuation allowance. The 10-Q discloses discrete tax benefits of "approximately $25 million of net discrete benefit" in the March quarter and "total discrete tax benefit of approximately $65 million" for the nine months, "primarily related to the release of certain valuation allowances in connection with the enactment of the One Big Beautiful Bill Act in July 2025, as well as net excess tax benefits related to share-based compensation expense." $65M against nine-month pre-tax income of roughly $2.2B is under 3%. It is real, it is disclosed, and it does not explain the ramp.

Test 3 — one-off charges, which run the other way. The March quarter absorbed a $105M legal settlement charge and $7M of restructuring, both above the operating line, plus $69M of net loss from debt transactions below it. For the nine months, net loss from debt transactions was $141M, of which $120M was a non-cash induced-conversion expense on the exchangeable-note retirements. These depress GAAP earnings, they do not inflate them. Cleaned of the legal settlement alone, March-quarter operating income would have been roughly $1.103B rather than $998M, and the operating margin 35.4% rather than 32.1%.

Test 4 — the GAAP-to-adjusted gap, quarter by quarter. Q3 FY2026: adjusted $4.10 against GAAP $3.27, a gap of $0.83 (≈$190M pre-tax, consistent with $105M legal + $69M debt loss + $7M restructuring). Q4 FY2026: adjusted $5.71 against GAAP $5.60 — a gap of eleven cents. The June quarter had almost no adjustments at all. The single largest number in the ramp is the cleanest one in the file.

Conclusion of the class-3 test: the earnings ramp is clean operating earnings driven by price and volume. The honest trailing multiple is 60.8x GAAP diluted EPS of $13.90 and 54.4x the adjusted total of $15.53 (the sum of the four earn_cal actuals: $2.61 + $3.11 + $4.10 + $5.71). We use those, and we do not discount either for one-offs, because the one-offs are working against the company.

Research and development was $755M in FY2026, 6.2% of revenue, against $724M (8.0%) in FY2025, $654M (10.0%) in FY2024 and $941M (8.1%) in FY2022. Absolute R&D is up only 4.3% year on year while revenue rose 34.1% — the entire operating-margin expansion at the opex line is operating leverage on a nearly flat cost base, not investment restraint that would show as a cut. But note the direction: R&D intensity is the lowest in six years, and this is the company defending an areal-density lead against flash. The 10-Q attributes product-development increases to "an $8 million increase in outside services costs and a $3 million increase in facilities costs" — small numbers. Marketing and administrative expense was $577M, up 2.9%.

Operating expense discipline is the second half of the margin story and it is worth quantifying. Between FY2025 and FY2026 revenue rose $3.098B and total operating expenses rose $154M. Ninety-five percent of every incremental revenue dollar above cost of sales fell to operating income.

4. Balance sheet and the enterprise-value rebuild

As at 2026-07-03 (fiscal year end), from the vendor payload. The most recent FILING-verified balance sheet in this archive is 2026-04-03, from the 10-Q filed 2026-04-29 — see Section 8, finding 1.

The deleveraging is the single most underappreciated fact in this file. Total debt fell $1.430B in one fiscal year, a 28.6% reduction, and net debt fell $2.243B, a 54.7% reduction. Shareholders' equity crossed from negative to positive for the first time in the six years of history available here. A year ago Seagate had a shareholders' deficit of $453M; today it has $2.167B of equity. The 10-Q filed in April still uses the caption "Shareholders' Deficit" when describing where the capped-call cost sits — the balance sheet was still negative at 3 April 2026 and turned in the final quarter.

Correction check 1 — does netDebtTTM omit short-term investments?

No. We tested it and the vendor is CORRECT. This is the defect that broke META ($45.7B), AAPL (sign inverted) and BRK-B ($321.4B of T-Bills), and on this name it does not apply, for a documented reason: Seagate holds no short-term investments. The balance sheet reports shortTermInvestments: 0 in all three years, and the 10-Q corroborates from two directions:

> "Our primary sources of liquidity as of April 3, 2026, consist of: (1) approximately $1.1 billion in cash and cash equivalents, (2) cash we expect to generate from operations and (3) $1.3 billion available for borrowing under our senior unsecured revolving credit facility."

> "Our cash and cash equivalents are maintained in investments with remaining maturities of 90 days or less at the time of purchase."

Three sources of liquidity, and marketable securities is not one of them. netDebtTTM of $1.861B = total debt $3.565B − cash $1.704B, and there is nothing else to subtract. We report this as a verified-clean check because a clean check is a finding.

Correction 2 — the share count used for market cap is one exchange stale

This is the real correction, and it is a share-count issue, not a debt issue.

• Vendor market cap $189,551,985,150 ÷ price $845.35 = 224,229,000 shares.
• 10-Q cover page: "As of April 27, 2026, 224,228,992 of the registrant's ordinary shares, par value $0.00001 per share, were issued and outstanding."
• Agreement to within eight shares — 0.000004%. The vendor is using the 10-Q cover count. Clean.

But on 2026-05-26/27, after that cover date, Seagate issued 2,023,124 new ordinary shares. The 8-K filed 2026-05-28 records it verbatim: the company and Seagate HDD Cayman entered privately negotiated exchange agreements "to exchange... $185.908 million principal amount of Notes for consideration consisting of an aggregate of $185.908 million in cash and 2,023,124 ordinary shares of the Company."

The vendor's market capitalisation is therefore understated by 2,023,124 × $845.35 = $1.710B. Corrected shares 226,252,116; corrected market cap $191.262B.

The rebuilt enterprise value

> Corrected market cap $191.262B (226,252,116 shares × $845.35)

> + total debt $3.565B

> − cash and equivalents $1.704B

> − short-term investments $0 (there are none)

> = corrected enterprise value $193.123B

The vendor reports $191.413B. The understatement is $1.710B, or 0.89% — precisely the omitted share issuance.

Correction 3 — the vendor's EBITDA is struck from net income, not operating income. The inc_a ebitda field of $4.250B equals net income $3.184B + interest expense $284M + tax $506M + D&A $276M. We compute operating EBITDA as operating income $4.094B + D&A $276M = $4.370B (35.8% margin), a $120M difference reflecting non-operating items. Both are defensible; we use the operating construction and say so, and note it moves EV/EBITDA from 45.0x to 44.2x — immaterial to any conclusion.

The debt stack, rebuilt from the filings

The 10-Q's debt table is a numeric table and the numbers are stripped from the extracted text. We rebuilt the structure from the prose, which is intact, and it is more informative than the table would have been.

• Obligor Exchange, settled 30 June 2025. All outstanding notes of eight series issued by Seagate HDD Cayman were exchanged for new notes issued by Seagate Data Storage Technology Pte. Ltd. (SDST), guaranteed by Seagate Technology Unlimited Company, Seagate HDD and Seagate Technology Holdings plc. "No gain or loss was recorded as the Exchange Offers were accounted for as a debt modification." Terms — rate, payment dates, redemption prices, maturity — identical. This is a guarantor and obligor reshuffle, not a refinancing, and it is why year-on-year debt-line comparisons in any screen may look odd.
• 3.50% Exchangeable Senior Notes due 2028 ("the 2028 Notes"). Issued 13 September 2023 in a private placement, $1.500B aggregate principal including a $200M over-allotment; maturity 1 June 2028; effective interest rate 3.94%.
• Revolving Credit Facility. $1.300B senior unsecured, entered 30 January 2025 with the Bank of Nova Scotia as administrative agent, term through 30 January 2030. "On April 3, 2026, no borrowings were outstanding under the Credit Agreement." Up to $150M available for letters of credit, $50M for swing-line loans.
• Financial covenant. "a total net leverage ratio of less than or equal to 6.75 to 1.00, commencing with the fiscal quarter ended June 27, 2025 and declining over time so that the maximum permitted net leverage ratio for each fiscal quarter ending after July 2, 2027 is 4.25 to 1.00." At 3 April 2026 the company states it "was in compliance with all of the covenants." Our computed net leverage of 0.43x sits an enormous distance inside a 6.75x limit — but note that the covenant tightens to 4.25x from July 2027, which is the point at which a cycle turn would matter.
• Repurchases. "During the March 2026 quarter, $40 million principal amount of the New June 2029 Notes were repurchased for cash at a discount to their principal amount."

The exchangeable notes: the most important single item on this balance sheet

This is where the convertible-instrument defect class bites, and it bites in an unusual way — the vendor's numbers are fine, but the instrument is about to change the share count and no vendor field tells you.

The retirement sequence, all filing-verified:

Then, on 11 June 2026, Seagate HDD called ALL remaining notes for redemption. From the 8-K filed 2026-06-12, verbatim:

> "On June 11, 2026, Seagate HDD Cayman, a subsidiary of Seagate (the 'Issuer' or 'Seagate HDD'), issued a notice (the 'Redemption Notice') to holders of the Issuer's 3.50% Exchangeable Senior Notes due 2028 (the 'Notes') calling for redemption (the 'Redemption') of all outstanding Notes."

> "On September 8, 2026 (the 'Redemption Date'), all then-outstanding Notes that are called for Redemption and have not been submitted for exchange will be redeemed for cash at a price... equal to the principal amount of such Notes plus accrued and unpaid interest."

> "The exchange rate for Notes exchanged after today and through the Exchange Deadline is currently equal to 12.1363 ordinary shares of Seagate, par value $0.00001 per share... per $1,000 principal amount of the notes... The principal amount of any Notes submitted for exchange shall be paid in cash, and Seagate HDD has elected to settle the remainder of its exchange obligations in excess of the principal amount... by delivering Ordinary Shares."

The arithmetic, at today's price. The exchange rate of 12.1363 shares per $1,000 principal is an exchange price of approximately $82.40 per share (10-Q, at 3 April 2026). The stock is $845.35 — 10.3x the exchange price. Every holder will exchange rather than take principal back.

On the $185M–$214M of principal still outstanding: total exchange value is $1.898B–$2.196B, of which $185M–$214M is settled in cash and the excess — $1.713B–$1.982B — is settled in shares. At $845.35 that is 2.03M to 2.35M new ordinary shares, roughly 0.9% to 1.0% of the current count, plus $185M–$214M of cash out, all landing by 2026-09-08. This is not a large dilution. It is, however, a dated, certain, unavoidable event that appears in no vendor field, and it is 35 days away.

And the hedge is worthless. The 10-Q: "the Company and Seagate HDD entered into privately negotiated capped call transactions... The current cap price of the capped call transactions, which was adjusted on March 25, 2026... is $107.750 per share. The cost of the capped call transactions was $95 million." Seagate paid $95M to hedge dilution up to $107.75. The stock closed at $845.35 — 7.85x the cap. The capped calls offset dilution only between the $82.40 exchange price and the $107.75 cap; every dollar of the move from $107.75 to $845.35 is unhedged economic dilution. The company was not wrong to buy the hedge in 2023; it is simply that nothing in the structure contemplated a seven-fold move.

Interest cost is falling fast as a result. Interest expense was $321M in FY2025 and $284M in FY2026, and the quarterly series runs $80M, $72M, $68M, $64M. Interest coverage is 14.4x and rising.

Two things on this balance sheet that deserve naming

Supplier finance has grown eighteen-fold and sits inside accounts payable. The 10-Q: "Outstanding amounts related to suppliers participating in the SCF program, recorded within Accounts payable... were $368 million and $20 million as of April 3, 2026 and June 27, 2025, respectively." The company has not pledged assets or provided guarantees, and the payments run through operating cash flow. But $368M of a $1.748B payables balance is now third-party-financed, and it inflates days payable outstanding — reported at 96.1 days against days inventory outstanding of 86.4 and days sales outstanding of 45.9, for a cash conversion cycle of just 36.2 days. That is an excellent working-capital profile, and roughly 20 days of the payables leg is a financing arrangement rather than genuine trade credit.

Book-equity metrics are meaningless and are rejected as lenses. priceToBookRatioTTM and priceToFairValueTTM are both 85.4x; returnOnEquityTTM is 348.2%; debtToEquityRatioTTM is 1.645x; financialLeverageRatioTTM is 4.60x. All are arithmetically correct and all are artefacts of equity having been negative until three months ago. The vendor's own composite rating scores STX 1 out of 5 on price-to-book, 1 out of 5 on debt-to-equity and 1 out of 5 on price-to-earnings, dragging the overall grade to B− / 3 — a rating artefact, not a judgement. returnOnInvestedCapitalTTM of 50.3% and returnOnAssetsTTM of 31.9% are the honest return measures and are what we use.

5. Cash flow and capital returns

Operating cash flow more than tripled, from $1.083B to $3.674B. incomeQualityTTM of 1.154 — operating cash flow above net income — is a healthy reading and reflects $276M of depreciation and $213M of stock compensation against only $167M of working-capital absorption, despite receivables rising $575M on a 34% revenue increase.

The FY2026 capex and free-cash-flow figures require a correction and it is a large one.

The vendor's cf_a for FY2026 reports investmentsInPropertyPlantAndEquipment: 0, capitalExpenditure: 0 and therefore freeCashFlow: $3,674M — free cash flow equal to operating cash flow. That is false, and three independent sources say so:

1. The 10-Q filed 2026-04-29, verbatim: "Net cash used in investing activities for the nine months ended April 3, 2026 was $369 million, primarily attributable to payments of $382 million for the purchase of property, equipment and leasehold improvements." And for the March quarter alone: "paid $161 million for the purchase of property, equipment and leasehold improvements."

2. The vendor's own km_ttm contradicts its own cf_a. capexToRevenueTTM of 0.028536 × $12.195B = $348M. capexPerShareTTM of $1.5890 × 219M basic shares = $348M. capexToDepreciationTTM of 1.2609 × $276M = $348M. evToFreeCashFlowTTM of 57.55 against EV of $191.413B implies free cash flow of $3.326B, not $3.674B. Four km_ttm fields agree with each other and disagree with cf_a.

3. The cf_a line otherInvestingActivities: −$554M in FY2026 is where the capex has been misfiled; the equivalent line was +$26M in FY2025 and +$40M in FY2024.

The filing wins. Nine-month capex is $382M, filing-verified, which already exceeds the vendor's implied full-year $348M. We estimate FY2026 capex at approximately $543M — $382M for nine months plus a fourth quarter at the March quarter's $161M rate — and we label it derived. Free cash flow is therefore approximately $3.131B, not $3.674M, and the free-cash-flow yield is 1.64%, not 1.94%. The precise fourth-quarter figure is not in this archive.

Capital expenditure is rising sharply and that is the number to watch. $316M (FY2023) → $254M → $265M → ~$543M (FY2026), roughly 2.0x depreciation of $276M. In an oligopoly whose pricing power rests on nobody adding capacity, the incumbent adding capacity is the beginning of the end of the pricing regime. We are not saying that is what this is — some of it is HAMR tooling and some is maintenance catch-up after three lean years — but it is the single cheapest early-warning indicator available and no vendor field carries it.

Depreciation itself has a discontinuity worth flagging. The series reads $397M (FY2021), $451M, $513M (FY2023), $264M (FY2024), $251M, $276M. A halving between FY2023 and FY2024 is not a normal depreciation pattern; the FY2023 figure almost certainly contains impairment or accelerated depreciation associated with the downturn. We do not use FY2023 D&A for any comparison and flag the series as non-comparable across that boundary.

Capital returns are modest and dividend-led.

• Dividend. The 8-K filed 2026-07-28 records: "the Board of Directors of the Company declared a quarterly cash dividend of $0.74 per share, which will be payable on October 7, 2026 to shareholders of record as of the close of business on September 24, 2026." Forward run-rate $2.96 annualised, a 0.35% yield. The vendor's dividendPerShareTTM of $2.94 and lastDividend of $2.94 are the trailing-twelve-month total, not the last declared amount — the same labelling trap seen elsewhere in this batch, and here it happens to be almost identical to the forward rate because the dividend barely grows. FY2026 dividends paid were $634M; FY2025 $600M; FY2024 $585M. Growth of about 3% a year on a 0.35% yield is not an income story and should not be sold as one. The payout ratio is 19.9%.
• Buyback. $176M in FY2026 — 0.09% of market capitalisation. FY2024 and FY2025 were zero. The 10-Q shows $59M of ordinary-share repurchases plus $111M of tax withholding on vesting equity across nine months. Seagate has spent four years deleveraging rather than repurchasing, which was the right call and is why the balance sheet is repaired — but it means there is no buyback support under this stock. Note also the 10-Q's phrasing, an Irish-plc quirk: "All repurchases are effected as redemptions in accordance with our Constitution."

Total shareholder yield is approximately 0.44% (0.35% dividend plus 0.09% buyback). Free cash flow yield is 1.64% on our corrected free cash flow. Neither number provides any support at this price; the entire return case is earnings growth and the multiple.

6. Valuation — priced in or room?

At $845.35 (corrected market capitalisation $191.262B, 226,252,116 shares, corrected enterprise value $193.123B):

Read the implied-net-margin row before anything else. It is the whole valuation argument.

We derive it as epsAvg × 229M diluted shares ÷ revenueAvg, and we validated the method against the FY2026 estimate row before using it: FY2026 epsAvg of $14.910 × 229M ÷ $12.0575B = 28.3%, against the actual non-GAAP net margin of 29.2% ($15.53 × 229M ÷ $12.195B). The method reproduces the delivered figure to within a point, so the forward readings are trustworthy as a description of what the estimates embed.

And what they embed is a net margin of 44.2% in FY2027 and 52.5% in FY2028 for a company that manufactures physical hardware in six countries. The best quarter in Seagate's history — the June 2026 quarter, the one that just beat by 12% — delivered a non-GAAP net margin of 36.0%. FY2027 consensus asks for 820 basis points more than the best quarter ever, sustained across a full year. FY2028 asks for 1,650 basis points more.

Estimate coverage collapses beyond FY2028 and we use nothing past it. FY2027 rests on 11 EPS analysts and 15 revenue analysts; FY2028 on 10 and 14; FY2029 on TWO EPS analysts and FY2030 on ONE. The FY2029 and FY2030 rows imply net margins of 59.2% and 71.9%, which are not estimates of anything — they are extrapolation artefacts. Both years are excluded from every conclusion in this dive.

est.ebitdaAvg and est.ebitAvg are FABRICATED and are not used. This is worth a full sentence because the signature here is cleaner than any we have documented. ebitdaAvg is exactly 19.7062% of revenueAvg and ebitAvg is exactly 15.7328% of revenueAvg in every single year from FY2024 through FY2030 — seven consecutive years, to four decimal places. These are not forecasts; they are a fixed ratio applied to the revenue line. The data contract documents a related signature (exactly 50.000% and 33.000%, five files); this is the same fabrication at different constants. A second, independent proof: FY2027 ebitdaAvg of $3.616B is below the FY2026 actual EBITDA of $4.370B while revenue is forecast 50.5% higher, and it is below the same row's netIncomeAvg of $8.342B, which is arithmetically impossible. All forward valuation in this dive runs on epsAvg and revenueAvg only.

Basis note, per the house rule. est.epsAvg is an adjusted (non-GAAP) measure. We verified: the FY2026 row's $14.910 sits between GAAP diluted EPS of $13.90 and the reported adjusted total of $15.53, and the FY2025 row's $7.962 sits 17.6% above GAAP's $6.77. Every forward multiple in this dive is struck on an adjusted basis. The trailing multiple is shown on both (60.8x GAAP, 54.4x adjusted).

Peer context. The vendor peer set is close to useless — Autodesk, Datadog, Electronic Arts, Fortinet, Infosys, Monolithic Power, Motorola Solutions, NXP and Super Micro are not comparables for a disk-drive manufacturer. The one genuine peer in the set is Western Digital at $548.62 and a $189.099B market capitalisation — within 0.2% of Seagate's own $189.552B. The two direct competitors in the three-player HDD oligopoly now carry effectively identical equity values. No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The 10-K's own competitor list (Kioxia, Micron, Samsung, Sandisk, SK hynix, Toshiba, Western Digital) is the more informative document, and its composition is discussed in 6c.

6a. What today's price assumes (the inversion)

At $845.35 — 23.9x FY2027 adjusted consensus and 14.8x FY2028 — the price embeds the following falsifiable claims, each with a number and a date:

• Revenue rises 50.5% in the twelve months to July 2027, to $18.348B, and a further 35.5% to $24.865B by July 2028. (Consensus-derived; 15 and 14 analysts.) The FY2027 figure is 57.3% above Seagate's previous all-time revenue peak of $11.661B, set in FY2022. The nearest check is the 2026-10-28 print: consensus revenue of $4.090B for the September quarter would be +55.6% year on year and +12.7% sequentially on a June quarter that itself grew 16.6% sequentially.
• Non-GAAP net margin expands from 29.2% (FY2026 actual) to 44.2% in FY2027 and 52.5% in FY2028. (Derived by us from consensus EPS and revenue, method validated against the FY2026 actual to within one point.) This is by a wide margin the most fragile assumption in the price. The best quarter ever recorded by this company delivered 36.0%. There is no precedent in six years of history — or in the FY2022 peak, when net margin was 14.1% — for anything approaching these figures.
• Gross margin holds at or above roughly 52%. (Our number, derived from the June-quarter print.) Gross margin has risen for eight consecutive quarters from 31.8% to 52.3%. The 10-Q attributes the move to pricing actions, and the 10-K contains a section headed "Price Erosion" describing the historical default behaviour of this market.
• The pricing regime survives the next capacity decision. (Our claim.) Seagate's own capital expenditure roughly doubled to approximately $543M in FY2026, about 2.0x depreciation. Three suppliers, no new capacity, is the entire mechanism, and one of the three is now spending.
• The market keeps paying roughly 15x or better on the forward year at every date between now and then. (Our number.) At 12x FY2028E the stock is $684; at 20x it is $1,140. For a business with a 3,820 basis point operating-margin swing inside four years, the multiple is not a stable parameter, and this is the widest source of variance in the outcome.

6b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: EPS growth (+61.0%, from FY2027E $35.393 to FY2028E $56.990) + multiple drift (COMPRESSION, from 23.9x to roughly 15.5x on the respective forward year, −35%) + shareholder yield (+0.44%) ≈ +5% to +15%, with an enormous variance around it.

Our base case assumes very substantial multiple compression, and it is important to be clear that this is NOT the rolldown artefact the addendum warns about. Today the market pays 23.9x the next fiscal year's estimate. Our base assumes that in twelve months it pays roughly 15.5x the then-next fiscal year's estimate. That is a genuine de-rating of 35%, and we assume it for a specific, defensible reason: this is what cyclicals do at the top of a cycle. A commodity manufacturer earning peak margins trades at a low multiple of those peak earnings precisely because the market does not believe they persist. Seagate at 24x forward is currently priced like a secular grower. Our base prices it like the cyclical the knowledge base's one claim insists it is.

Nearly all of the expected return in our base is earnings growth, and the multiple is fighting it hard. If the multiple simply HELD at 23.9x on FY2028E, the price would be $1,362 (+61.1%) — which is roughly where our bull case lands, and roughly where the street's $1,130 sits between. If it compressed to 12x, the price would be $684 (−19.1%). The entire spread between a good and a bad outcome sits in a multiple we do not control and have no edge on, applied to an estimate whose implied margin has no historical precedent.

The bull case at $1,360 requires no re-rating at all — it applies 17x, well below today's 23.9x, to the FY2028 consensus high of $79.892. The bear case at $490 is the mirror image: 18x applied to the FY2027 consensus low of $27.296.

6c. Variant perception (where we differ, what would surprise)

• We differ from the street on the multiple, not the earnings, and by a wide margin. The street's consensus target is $1,130.15 (+33.7%), with a low of $875 that is 3.5% ABOVE the current price — meaning every published analyst target in this file sits above spot. That is the single strongest piece of evidence against our own view and we put it here rather than burying it. Our base of $885 is 21.7% below consensus. We are not disputing the FY2027 revenue line; we are disputing that a business with a demonstrated 3,820 basis point operating-margin swing should be capitalised at 24x peak-cycle forward earnings.
• We differ from the knowledge base's own classification of this company, and the filing supports us. The one claim naming Seagate groups it with MU, SNDK and WDC as "memory names". Seagate is not a memory company. It makes mechanical disk drives, and its 10-K lists Micron, Samsung, SK hynix, Kioxia and Sandisk as principal COMPETITORS — the memory makers are the substitute product, not the same trade. The practical consequence runs in Seagate's favour and almost nobody frames it this way: the NAND and DRAM scarcity that has repriced the memory complex raises the cost per terabyte of flash, which WIDENS Seagate's cost advantage in cold and nearline storage. The same physical shortage that is revenue for Micron is a competitive tailwind for Seagate. That is a genuine variant perception and it is filing-grounded. (It also cuts the other way in the market: if Seagate is not a memory name, it should not be traded as one — and today's fade suggests the market may be starting to agree.)
• We differ on what today's price action means. The consensus reading of 2026-08-04 is "storage participated in the semiconductor melt-up". What actually happened is that STX opened +5.6% and closed +1.72%, 4.48% below its own high, while MU closed +7.62%, SNDK +10.79% and MRVL +12.81%. Seagate gave back three-quarters of its gap on a day everything around it held. Watchable number: whether STX can close above the 50-day average of $892.66 while the complex is rising. If it cannot, the relative-strength break is real.
• Positive surprise that would force a re-rate upward: September-quarter revenue materially above $4.090B with gross margin at or above 52.3%, which would simultaneously validate the FY2027 ramp and prove the pricing regime is durable rather than a spike; or a disclosed multi-year nearline supply agreement with a hyperscale customer that converts price from a spot outcome to a contracted one. Either would justify paying a growth multiple for a cyclical.
• Negative surprise that would break the thesis: gross margin below 46.5% in the September quarter, reversing two quarters of the expansion; exabytes shipped falling sequentially while revenue holds, which would show price carrying the whole load; any of the three HDD players announcing capacity additions; or the FY2026 10-K, due imminently, disclosing customer concentration materially above the 10% last reported. Any of these hits a stock 446% higher over twelve months with a 2.10 beta and no buyback support.

Synthos fair values

All three anchors are multiples of the consensus adjusted-EPS distribution, cross-checked across both usable forward years. FY2027E: mean $35.393, low $27.296, high $42.869 (11 analysts). FY2028E: mean $56.990, low $40.003, high $79.892 (10 analysts).

• Bear ~$490 — 18x the FY2027 consensus LOW of $27.296 ($491). Cross-check: 12.2x the FY2028 consensus low of $40.003; 6.6% below the 200-day moving average of $524.55, which is where a cyclical unwind typically finds its first real support. The scenario: the pricing regime breaks in calendar 2027, gross margin retraces from 52.3% toward the high 30s, hyperscale customers digest, and the multiple compresses as the market re-prices peak earnings. This is not a catastrophe scenario — FY2023 was a catastrophe scenario and it involved a NEGATIVE operating margin. −42.0%.
• Base ~$885 — 25x the FY2027 consensus MEAN of $35.393 ($885). Cross-check: 15.5x FY2028E; corrected enterprise value of $202.1B on FY2027E revenue of $18.348B is 11.0x sales, which is demanding for a hardware manufacturer and is disclosed as such. Sensitivity, stated openly: 22x gives $779 — 7.9% BELOW spot — and 28x gives $991. The whole answer lives in a 22-28x band on FY2027 and we claim no more precision than that. The scenario: FY2027 lands roughly on the consensus revenue line, the implied margin proves too ambitious by several hundred basis points, and the multiple compresses toward the mid-teens on FY2028 as the market prices the cycle rather than the growth. +4.7%, plus about 0.4% of shareholder yield.
• Bull ~$1,360 — 17x the FY2028 consensus HIGH of $79.892 ($1,358). Cross-check: 38.4x FY2027E mean — expensive-looking on the near year, which is the point: this bull case requires no re-rating at all, only that the top of the analyst range is right. The scenario: nearline demand from AI-era data retention proves structural rather than cyclical, no player adds capacity, the flash cost-per-terabyte gap widens on continued NAND scarcity, and Seagate's HAMR-based Mozaic platform extends the areal-density lead. +60.9%.

Base is 4.7% above spot; asymmetry roughly 1.45:1 to the upside (42.0% down, 60.9% up). That ratio looks acceptable and it is misleading, because it treats a 42% decline and a 61% advance as equally likely. They are not: the 61% requires an unprecedented margin to be sustained for two years, and the 42% requires only that this cyclical behave the way it has behaved in every prior cycle in this file. What makes this a Watch rather than a Hold is that the base case sits within 5% of spot, 22% below the street, and rests entirely on a consensus estimate whose implied margin we cannot corroborate from any source in this archive.

7. Knowledge base — one claim, and it is a warning

Raw hits naming Seagate: 1. Entity matches: 1. Text-only matches on the ticker or company name: 0. Used as name-level conviction: 1 (as a caution). Discarded for collision: 0 on the name; 153 on the broad "storage" sweep.

The primary search covered STX, Seagate, HDD, hard disk, hard drive, HAMR, nearline, mass capacity, exabyte, areal density, disk drive, data storage, Dave Mosley, Mosley, Starbox and Western Digital across all 51,928 distilled claims. The STX and Seagate sweep returned exactly one entity match and zero text-only matches — the anticipated ticker collisions with Starbox and STX Entertainment did not materialise; the Starbox sweep returned zero hits of any kind.

Here is the entire name-level lane, verbatim.

> 2026-05-12 · neutral · conviction 65 · horizon: principle · entities: MU, SNDK, STX, WDC · categories: memory/HBM, cyclicals · channel: compound_and_friends · SPEAKER: Josh Brown · role: not management · skill weight 1.0

> "Don't mistake memory names' single-digit PEs for cheapness — they're the most cyclical names on the planet with wildly swinging operating margins, so they deserve a discount."

That is it. One claim, one named speaker, one date, and it is a direct argument against the bull case for this stock.

What the lane is, stated without inflation. The claim is an entity match, so the attribution discipline is satisfied — STX genuinely appears in the entity list. It carries a named speaker, which is the stronger of the two sourcing tiers under the 4-lane attribution policy, and no management speaker_role, so the half-weight management discount does not apply anywhere in this lane. It is neutral in stance and 65 in conviction, and its horizon is principle, meaning it is a durable frame rather than a dated call.

It is also, unusually, directly applicable to the arithmetic in Section 6. Seagate's FY2028 multiple is 14.8x — a low-teens forward multiple. Brown's claim says that is not cheapness. Our base case is, in effect, an independent derivation of the same conclusion from the filings, and we note the agreement rather than claiming credit for it.

The one place we depart from the claim is its classification. Brown groups STX with MU, SNDK and WDC as memory names. Seagate is a disk-drive manufacturer whose own 10-K names Micron, Samsung, SK hynix, Kioxia and Sandisk as competitors. The cyclicality point survives that correction intact — Seagate is, if anything, more cyclical than the memory names on the operating-margin measure — but the demand driver is different and we say so in 6c.

The adjacent lane: five Western Digital claims, used as complex-level colour only

Western Digital is the only direct comparable in the knowledge base, and the WDC lane is more populated than the STX lane. We use it for context and explicitly NOT as conviction on Seagate.

> 2026-06-23 · bearish · conviction 62 · horizon: fact · entities: MU, WDC, SNDK · channel: compound_and_friends · no named speaker

> "Memory/semi stocks severely overextended — semi index 68% above its 200-day MA, most stretched in 25 years; Micron 161% above; pullback severely overdue, need to sober up."

> 2026-07-07 · bullish · conviction 40 · horizon: thesis · entities: MU · channel: compound_and_friends · no named speaker

> "Despite the memory-name unwind (SanDisk −30%, WDC −28%, Micron −22%), guessing this isn't the ultimate top — plenty of dip buyers ready to step back in."

> 2026-07-28 · bearish · conviction 50 · horizon: fact · entities: MU, SNDK, WDC · channel and SPEAKER: Josh Brown · role: independent

> "A disgusting but necessary unwind in the memory/semi names they own; $46B of inflows into semiconductor ETFs this year is unsustainable, and investors are sick of the circular [a frontier AI developer]/Nvidia financing deals." [The speaker named a specific AI laboratory; the name is redacted per house style.]

> 2026-06-17 · bullish · conviction 75 · horizon: thesis · entities: MU, SNDK, WDC · channel: jordi_visser · SPEAKER: Jordi Visser · skill weight 2.0

> "Memory is a 5-year-minimum bull market — DRAM took off since September, humanoids will need lots of memory; Micron, SanDisk, Western Digital have gone up 3-10x." Test: "memory names remain a bull market over the next 5 years"

> 2026-01-24 · bullish · conviction 80 · horizon: thesis · entities: MU, WDC, SNDK · channel: jordi_visser · SPEAKER: jordi_visser · skill weight 2.0

> "High-bandwidth/DRAM memory is scarce versus surging AI demand — Micron, SanDisk, Western Digital, SK Hynix, Samsung in 'banana zone'; a race to buy at size."

How we read that adjacent lane. It is a two-sided complex-level record with a clear time structure: bullish in January and June 2026 from Jordi Visser (skill weight 2.0), turning to a documented unwind in late June and July from the compound_and_friends channel and Josh Brown. That structure corroborates the technical picture in Section 9 exactly — a violent advance into roughly June 2026 followed by a sharp drawdown, which is why STX sits 22.7% below its 52-week high while still being up 446% over twelve months. It is genuinely useful for understanding where in the cycle we are standing, and it is emphatically not a view on Seagate.

One further adjacent claim is worth quoting because it is the only independent technology view in the file:

> 2024-02-01 · bullish · conviction 72 · horizon: thesis · channel: gavin_baker · SPEAKER: Gavin Baker · role: independent

> "All hard drives will be replaced by flash storage; storage is the layer below something like Snowflake, and the old EMC-style 'how to invest in data storage' opportunity is coming again."

That claim is thirty months old and has not been right yet — HDD is currently taking price rather than losing share — but it is the correct long-run frame and it is the reason the exponential score is a 5 rather than a 7. It also happens to be the same substitution risk Seagate discloses through its own competitor list.

And one on cyclicality that generalises:

> 2024-08-09 · neutral · conviction 60 · horizon: principle · channel and speaker: bill_gurley · role: independent

> "From covering PCs: demand was over- and under-estimated five times; commodities like DRAM and hard drives had boom-bust shortage cycles out of step with reality."

What we discarded and why

• The storage sweep returned 1 entity match and 153 text-only matches. Every one was discarded. The entity match was tagged "battery storage" — a 2019 grid-economics claim from a biotech channel. The 153 text-only matches decompose into: grid-scale and residential battery storage (the largest group by far), self-storage and marina storage real estate, cold storage of crypto private keys, grain and fertiliser storage, oil and floating storage, tool storage financing, datacentre storage software and accelerator hardware, and decentralised web3 storage. Not one concerned a disk-drive or storage-hardware equity. The brief flagged this collision in advance and it is worse than flagged.
• The memory sweep returned 7 entity matches and 508 text-only matches and was abandoned as unusable: "memory" collides with human memory, model context windows, institutional memory and muscle memory. No claim was taken from it that was not already captured by the WDC and SanDisk sweeps.
• The hyperscaler capex sweep returned 39 text-only matches, all basket or theme claims about AI capital expenditure with no storage-hardware specificity. Discarded as non-name-level.
• One claim was excluded on quarantine grounds: a 2025-10-10 entity match on Nvidia/Micron/ASML/Western Digital carrying the flag QUARANTINED:SUMMARY_OF_VISSER. Under the attribution policy a summary-of-a-speaker claim is not the speaker, and we do not use it.
• Zero claims in any lane carried a management speaker_role. There is no company-voice contamination in this lane — which, given that the lane has one claim, is a small mercy.

Conclusion, stated as a finding rather than dressed up. The Synthos knowledge base has essentially no coverage of Seagate Technology. It has one claim, from one named independent voice, and that claim is a caution against precisely the "it's only 15x FY2028" argument that makes this stock look interesting. There is no bullish name-level conviction to weigh against it, and there is no basis in the knowledge base for a conviction premium in the fair value. The conviction rating is Low, breadth 1, net conviction neutral-cautionary, and the emptiness of the lane is a direct input to the Watch verdict.

8. Data integrity — what we tested, what we rejected, and why

Fourteen findings. Several are verified-clean checks, which are reported as findings in their own right.

1. The freshness asymmetry runs BACKWARDS on this name, and it must be stated before any number is used. The vendor's annual figures are FRESHER than any filing in this archive. inc_a[0], bal_a[0] and cf_a[0] all cover fiscal 2026, the year ended 2026-07-03, and carry filingDate: 2026-07-28 — the date of the results 8-K. The newest filing in this archive is that same 8-K, and it does not contain the financials. The 8-K filed 2026-07-28 is present in full text and is NOT a bare cover page — it carries Item 2.02, Item 7.01 and Item 9.01 in prose, including the declaration of a $0.74 quarterly dividend payable 7 October 2026. But it states, verbatim: "The press release is attached to this Current Report on Form 8-K as Exhibit 99.1." Exhibit 99.1 was not extracted and is absent from this archive. The file therefore contains no FY2026 financial statements of any kind. The FY2026 10-K had not been filed as of this dive (the FY2025 10-K was filed 2025-08-01, so the FY2026 equivalent is due imminently). The most recent FILING-VERIFIED balance sheet in this dive is 2026-04-03, from the 10-Q filed 2026-04-29. Every FY2026 annual and June-quarter figure in this dive comes from the vendor payload, cross-checked against earn_cal: Q4 revenue actual $3,629,000,000 matches inc_q exactly, and the four quarterly revenue actuals sum to $12,195M, matching inc_a FY2026 to the dollar. This is stated so that no reader mistakes the FY2026 balance sheet for a filing-verified figure.

2. km_ttm.netDebtTTM — TESTED AND CLEAN. This is the defect class that produced a $45.7B error on META and a sign inversion on AAPL. On Seagate it does not apply, for a documented reason: the company holds no short-term investments. shortTermInvestments is $0 in all three balance-sheet years, and the 10-Q lists exactly three sources of liquidity — "approximately $1.1 billion in cash and cash equivalents", operating cash flow, and "$1.3 billion available for borrowing" — with no marketable-securities line, adding that cash equivalents are "maintained in investments with remaining maturities of 90 days or less." Net debt of $1.861B (total debt $3.565B − cash $1.704B) is correct and we use it unchanged. A verified-clean check on the defect that has broken more dives than any other.

3. The market-capitalisation share count is one corporate action stale — REBUILT. Market cap $189,551,985,150 ÷ $845.35 = 224,229,000 shares, matching the 10-Q cover page's "As of April 27, 2026, 224,228,992 of the registrant's ordinary shares... were issued and outstanding" to within eight shares. But the 8-K filed 2026-05-28 records the issuance of 2,023,124 additional ordinary shares on 26-27 May 2026 in exchange for $185.908M of exchangeable notes. The vendor's market capitalisation is understated by $1.710B. Corrected: 226,252,116 shares, $191.262B market cap, $193.123B enterprise value. Every valuation figure in this dive uses the corrected count.

4. Three mutually inconsistent share counts sit in one payload — reconciled and labelled. weightedAverageShsOut (basic) is 219M; weightedAverageShsOutDil is 229M; the market-cap-implied count is 224.229M; the 10-Q cover count at the balance-sheet date (3 April 2026) was 224,408,070. All four are individually correct and they measure different things: basic and diluted are FY2026 weighted averages across a year in which roughly 12.3M shares were issued in note exchanges; the cover count is a point-in-time figure. ratios_ttm uses the 219M basic count throughout — revenuePerShareTTM $55.685, netIncomePerShareTTM $14.539 and bookValuePerShareTTM $9.895 all divide by 219.0M — while priceToEarningsDilutedRatioTTM of 60.82x uses the 229M diluted count. Both appear in the same block without labels. We use diluted for earnings multiples and the corrected point-in-time count for market capitalisation, and we say which is which everywhere.

5. cf_a FY2026 capital expenditure is ZERO and free cash flow is consequently overstated by roughly $543M — REJECTED and rebuilt from the filing. investmentsInPropertyPlantAndEquipment: 0, capitalExpenditure: 0, freeCashFlow: $3,674M — identical to operating cash flow. The 10-Q states "payments of $382 million for the purchase of property, equipment and leasehold improvements" for the nine months to 3 April 2026, and "paid $161 million" in the March quarter alone. The vendor's own km_ttm block independently contradicts its cf_a block: capexToRevenueTTM, capexPerShareTTM and capexToDepreciationTTM all imply $348M, and evToFreeCashFlowTTM of 57.55 implies free cash flow of $3.326B, not $3.674B. The filing wins: nine-month capex alone ($382M) exceeds the vendor's implied full year ($348M). We estimate FY2026 capex at approximately $543M (nine months at $382M plus a fourth quarter at the March run-rate) and free cash flow at approximately $3.131B, both labelled as derived. The otherInvestingActivities line of −$554M in FY2026, against +$26M in FY2025, is where the misfiling shows.

6. est.ebitdaAvg and est.ebitAvg are FABRICATED — REJECTED in full, with the signature documented. ebitdaAvg is exactly 19.7062% of revenueAvg, and ebitAvg is exactly 15.7328% of revenueAvg, in every year from FY2024 through FY2030 — seven consecutive rows, identical to four decimal places. These are a fixed ratio applied to the revenue forecast, not estimates. Two independent confirmations: FY2027 ebitdaAvg of $3.616B is below the FY2026 actual EBITDA of $4.370B while revenue is forecast 50.5% higher; and it is below the same row's netIncomeAvg of $8.342B, which is arithmetically impossible. All forward valuation runs on epsAvg and revenueAvg. (The data contract records a related fabrication at 50.000%/33.000% appearing in five files; this is the same phenomenon at different constants and is logged as a new instance.)

7. est.netIncomeAvg implies impossible net margins in the outer years — those years REJECTED entirely. The netIncomeAvg rows imply net margins of 45.5% (FY2027), 48.3% (FY2028), 59.2% (FY2029) and 71.9% (FY2030) for a hardware manufacturer. FY2029 rests on TWO EPS analysts and FY2030 on ONE. Both years are excluded from every conclusion. FY2027 and FY2028 are used, with the implied 44.2% and 52.5% non-GAAP net margins reported explicitly in Section 6a as the most fragile assumption in the price rather than smoothed over. Method validation: the FY2026 estimate row implies 28.3% against a 29.2% actual, so the derivation is sound.

8. The earnings ramp was tested for one-off inflation and is CLEAN — a verified-clean finding. The earn_cal sequence 2.59 → 2.61 → 3.11 → 4.10 → 5.71 survives four tests: (a) the effective tax rate ROSE from 2.9% (FY2025) to 13.7% (FY2026), a headwind of roughly $400M, not a tailwind; (b) the disclosed valuation-allowance release and share-based-compensation excess benefits totalled $65M across nine months, under 3% of pre-tax income; (c) the March quarter absorbed a $105M legal settlement and $7M of restructuring above the operating line and $141M of net loss from debt transactions (of which $120M was non-cash induced-conversion expense) below it, all of which DEPRESS earnings; (d) the June quarter's GAAP-to-adjusted gap was eleven cents ($5.60 versus $5.71), the cleanest quarter in the file. The honest trailing multiple is 60.8x GAAP and 54.4x adjusted, and no downward adjustment to reported earnings is warranted.

9. Currency — TESTED AND CLEAN. inc_a[0].reportedCurrency is USD; profile.currency is USD; every balance-sheet and cash-flow row reports USD. The 10-K states: "References to '$' and 'dollars' are to United States dollars." No currency mixing. Despite the Irish incorporation, the Singapore headquarters and the Cayman and Singapore debt issuers, there is a single reporting currency and the quote is struck in it. This was checked because the ASML and TSM failures made it mandatory, and it passed.

10. seg_prod is EMPTY — stated as a gap, not estimated around. The vendor supplies an empty array. Seagate has ONE reportable segment (10-K Note 15: "the Company has concluded that its manufacture and distribution of storage solutions constitutes one operating segment"), so there is nothing to populate. The consequence is that the mass-capacity versus legacy revenue split — the most important mix question about this business — is unavailable from any source in this file, because the 10-K's disaggregation table is a numeric table and the numbers are stripped from the extracted text. No product-mix figure appears anywhere in this dive.

11. seg_geo sums perfectly and is mislabelled in substance — used, with the label corrected. The four lines sum to $9,097M (FY2025), $6,551M (FY2024) and $7,384M (FY2023), matching reported revenue to the dollar in every year. No double-counting, no vanished region — the arithmetic is clean and we verified it. But the 10-K footnote reads: "Revenue is attributed to countries based on bill from locations." These are invoicing entities, not customer geographies. Read as demand data — which is how a screen would read them — they would falsely suggest a 13-point demand shift from Singapore to the United States between FY2024 and FY2025. We present the table with the label corrected and draw no demand conclusion from it. The series also stops at FY2025; there is no FY2026 geographic mix in this file.

12. quote.yearHigh/yearLow disagree with tech.hi52/lo52 — we use tech and say so. The quote block reports a 52-week high of $1,145.00 and low of $144.75; the computed technical block reports $1,094.04 and $147.27. The discrepancies are 4.5% and 1.7% — larger than the KLAC equivalents and material to any percentile calculation. We use the tech figures throughout, because they come from the same six-year close series that generates the moving averages, drawdown, RSI and relative returns, and mixing sources within the technical section would make the percentages mutually inconsistent. Note that tech.hi52 and the drawdown-from-peak reading coincide at −22.73%, which is the expected behaviour when the 252-day maximum is recent, and is NOT a defect. The same applies to the moving averages: quote.priceAvg50 of $890.37 against tech.sma50 of $892.66, and priceAvg200 of $518.32 against sma200 of $524.55 — small differences from different window conventions; we use tech.

13. Balance-sheet component fields are zeroed in FY2026 — the equity composition is unusable. In bal_a[0], retainedEarnings, additionalPaidInCapital, accumulatedOtherComprehensiveIncomeLoss, commonStock and treasuryStock are ALL $0, with the entire $2.167B lumped into otherTotalStockholdersEquity. The FY2025 and FY2024 rows carry the proper decomposition (FY2025: retained earnings −$8.151B, additional paid-in capital $7.706B, AOCI −$8M). The FY2026 equity composition cannot be analysed from this file and no conclusion in this dive depends on it. capitalLeaseObligations, deferredRevenue, intangibleAssets, shortTermInvestments and longTermInvestments are also $0 across all years — for short-term investments this is verified true (finding 2); for the others we treat the fields as not-populated rather than as zero. Related: the depreciation series ($397M, $451M, $513M, $264M, $251M, $276M) halves between FY2023 and FY2024 in a way normal depreciation does not; the FY2023 figure almost certainly contains impairment or accelerated depreciation and we make no comparison across that boundary.

14. lastDividend: 2.94 is the trailing-twelve-month total, not the last declared dividend — corrected from the filing. The 8-K filed 2026-07-28 declares "a quarterly cash dividend of $0.74 per share, which will be payable on October 7, 2026." Forward run-rate $2.96 annualised, a 0.35% yield — coincidentally almost identical to the trailing $2.94, because the dividend grows about 3% a year. The label is wrong; the number happens to be nearly right. It is corrected here because on a fast-growing dividend the same error would be material.

Also tested and found clean, or stated as reported: the FY2026 quarterly revenue actuals in earn_cal sum exactly to the annual revenue line; interestCoverageRatioTTM of 14.4x is consistent with operating income of $4.094B against interest expense of $284M; effectiveTaxRateTTM of 13.71% reconciles to $506M on $3.690B of pre-tax income; stockBasedCompensationToRevenueTTM of 1.30% reconciles to $213M on $12.195B; incomeQualityTTM of 1.154 correctly reflects operating cash flow of $3.674B against net income of $3.184B; cashAtEndOfPeriod of $1.705B in cf_a against cashAndCashEquivalents of $1.704B in bal_a differs by $1M, consistent with a restricted-cash reconciling item; and the 8-K financing events were checked against every statement in the file, with the result set out in the next paragraph.

The 8-K financing scan — the class-1 test, run explicitly. All three 8-Ks were read in full. The 2026-05-28 8-K (exchange of $185.908M of notes for cash plus 2,023,124 shares, consummated 26-27 May 2026) PRE-dates the FY2026 balance-sheet date of 2026-07-03 and is therefore captured in bal_a[0] — but it POST-dates the 10-Q cover-page share count the vendor used for market capitalisation, which is finding 3. The 2026-06-12 8-K (redemption notice calling ALL remaining 2028 Notes for redemption on 2026-09-08) creates an obligation that resolves AFTER every financial statement in this file. That is the class-1 event, and unlike the CRM and GOOGL cases it is a known, dated, quantifiable one rather than a hidden liability: $185M–$214M of cash out and approximately 2.03M–2.35M shares in, by 8 September 2026. The filing wins and the dive says so.

Off-balance-sheet items — hunted for and found to be small, which is itself a finding. The 10-K, Note 14: "As of June 27, 2025, the Company had unconditional long-term purchase obligations of approximately $77 million, primarily related to purchases of inventory components", and "unconditional long-term commitments of approximately $46 million, primarily related to purchases of equipment." Against $12.195B of revenue these are trivial — a striking contrast with the $4.83B of purchase commitments disclosed by KLA in the same batch. The 10-K does note "certain long-term market share based inventory purchase commitments" without quantifying them. The one genuinely material off-balance-sheet-adjacent item is the supplier-finance programme: $368M inside accounts payable at 3 April 2026, up from $20M at 27 June 2025, with no assets pledged and no guarantees given.

Non-equity tripwire — checked and passed. STX is common equity: ordinary shares, par value $0.00001 per share, NASDAQ Global Select, ISIN IE00BKVD2N49, CUSIP G7997R103, with 100,000,000 preferred shares authorised of which "none were issued or outstanding as of April 3, 2026" per the 10-Q. The price of $845.35 is not par-like; beta is 2.102, an order of magnitude above the sub-0.3 of a fixed-income-like instrument; the dividend is variable and discretionary; volume was 4.87M shares (roughly $4.1B of turnover); and the 52-week band of $147.27 to $1,094.04 is a 643% range — the opposite of a narrow band. This is common equity and behaves like the most volatile kind of it.

9. Technicals

• Price $845.35. −22.73% from the 52-week high of $1,094.04, and +474.0% above the 52-week low of $147.27. Position within the annual range: 73.7th percentile — the upper third, not the middle and not the top.
• Moving averages, split decision: −5.3% below a 50-day average of $892.66; +61.2% above a 200-day average of $524.55. The primary uptrend is not merely intact, it is extreme; the intermediate trend is broken.
• RSI 51.5 — dead neutral, which is unusual for a stock 22.7% off its high and tells you the drawdown has already been digested. MACD −15.46 — negative, and the most negative reading of any technical field in this file.
• Maximum drawdown from peak over the trailing year: −22.73%, i.e. the current price is the worst point of the drawdown.
• Relative performance, and it is extraordinary: 3-month +14.46% vs SPY +7.59% and QQQ +7.67%; 6-month +95.25% vs SPY +11.09% and QQQ +15.71%; 12-month +446.06% vs SPY +24.26% and QQQ +30.80%. STX has outperformed the S&P 500 by 422 percentage points over twelve months. That is not a normal equity return; that is a cyclical going from distressed to peak inside one fiscal year.
• Beta 2.102 — the highest in this batch, and consistent with everything above.
• Sentiment: 1 strong buy, 28 buy, 19 hold, 4 sell across 52 analysts; consensus target $1,130.15 (+33.7%), median $1,090 (+28.9%), high $1,600 (+89.3%), low $875 — 3.5% ABOVE spot. Every published target in this file is above the price. That is rare and it is the strongest single argument against our base case.

Today's move and what it does to the entry

STX closed 2026-08-04 at $845.35, up 1.72% or $14.29 from a previous close of $831.06. It opened at $877.55 — a 5.60% gap — traded to a high of $885.00, fell to a low of $833.97, and closed at $845.35, which is 4.48% BELOW the day's high and 1.37% above the day's low, on 4.87M shares against a 4.68M average.

That is a failed gap, and on this particular day it is a relative-strength failure. The whole memory and storage complex was bid: MRVL +12.81%, INTC +10.92%, SNDK +10.79%, LRCX +7.90%, MU +7.62%, AMD +7.00%, KLAC +6.95%, AMAT +5.48%, ASML +4.22%, WDC +4.14%. STX at +1.72% was the laggard of its own group, and it was the one that gave its gap back. No company-specific news for 2026-08-04 appears in this file; the last company event was the 2026-07-28 results release, seven days earlier, which was a strong beat.

The honest read. Elsewhere in this batch a 7% gap means paying up. Here the reader is being offered something more ambiguous and, on balance, worse: a stock that participated in a sector melt-up for one hour and was distributed for the remaining six and a half, on volume only 4% above average. A gap-and-fade of this shape, below a falling 50-day average, with MACD at −15.46 and RSI at a neutral 51.5, is the signature of supply meeting a rally rather than demand initiating one. One day is one day and we do not over-read it — but the same day that confirmed the complex's strength called Seagate's participation in it into question.

And the valuation does not rescue the chart. Even 22.7% below the high, STX trades at 60.8x trailing GAAP earnings, 54.4x trailing adjusted, 23.9x the FY2027 consensus and 15.8x sales, and our base fair value of $885 is only 4.7% above the closing price. The drawdown removed some of the parabola's excess; it did not create a discount.

What we are telling the reader, plainly: we are not asking you to buy a gap that already failed, and we are not asking you to buy at all today. The setups that would change this are the falsifiers:

• A price near $650-700 — roughly 18-20x FY2027E and 11-12x FY2028E, and between the 200-day average of $524.55 and the current price — which would put our base $885 at a 26-36% premium and make the risk/reward genuinely asymmetric. That is 17-23% below today's close and well inside the range this stock has traversed in a single quarter.
• Or the 2026-10-28 print landing above $4.090B of revenue with gross margin at or above the June quarter's 52.3%, which would validate both legs of 6a simultaneously and convert the pricing regime from a spike into a plateau. At that point paying up is defensible, because the uncertainty being paid for has been resolved.

Buying today means paying for a 5.6% gap that was sold, for a 4.7% base-case return, on a 2.10 beta, with no company-specific catalyst for 85 days and a dilution event in 35. That is not a trade with an edge in it.

10. Insiders — eight transactions, all selling, and only one of them means anything

Every insider transaction in this file is from July 2026, in the two weeks following the fiscal-year-end and around the earnings release.

Total open-market sales: 5,196.25 shares, approximately $4.75M at the transaction prices. Zero open-market purchases.

The reading, stated carefully.

The chief executive's four sales are trivial in size and read as a scheduled programme. 3,960 shares in four tranches on a single day, at four prices within $3.04 of each other ($919.64 to $922.68), is the execution signature of a pre-arranged sale, not a discretionary decision. $3.65M against a residual holding of 316,900 shares worth $267.9M at today's close. He sold 1.23% of his position. That is noise, and the residual stake is enormous, long-tenured and unhedged. We read it as neutral.

The chief legal officer's sequence is more notable and we say so. James Lee vested 1,237 restricted share units on 22 July, sold 542.25 shares on 23 July at $905.01, and then sold a further 694 shares on 30 July at $877.45, leaving a direct holding of 346.75 shares — worth approximately $293,000. An executive vice president of a $190B company holding under $300,000 of stock directly is unusual. The 30 July sale is also the only transaction in this file executed after the 28 July earnings release, at a price 3.8% above today's close. We do not over-read a $610,000 sale by a legal officer — this is a small absolute number and legal officers are not the informed traders in a business like this — but the reduction to a near-nil direct holding is the one datum in this table that is not routine, and we log it rather than smoothing it.

What the file does NOT contain is any open-market purchase by any insider, at any price, in the window covered. After a 446% twelve-month move and a drawdown of 22.7% from the high, not one officer or director bought. That is a mildly negative signal and it is reported as mildly negative — the absence of buying into a drawdown carries less information than a sale would, but it is not nothing.

One structural note. All transactions are in "Ordinary Shares", consistent with the Irish plc structure, and none is coded F-InKind — unlike the KLA table in this batch, these are genuine dispositions rather than tax withholding, with the exception of the two M-Exempt lines which are the RSU vesting itself.

11. Verdict, kill-criteria and flip conditions

Watch.

This is the most improved business in this batch and the one where the improvement is most likely to be temporary.

What is genuinely excellent, and none of it is in dispute: revenue of $12.195B in FY2026, up 34.1%, the highest in the company's history; a gross margin of 45.6% against 23.4% two years earlier, with the June quarter at 52.3%; an operating margin of 33.6% for the year and 42.5% in the June quarter, against minus 4.6% for the whole of fiscal 2023; net income of $3.184B; operating cash flow of $3.674B, more than triple FY2025's; total debt down $1.430B to $3.565B and net debt down $2.243B to $1.861B, a leverage ratio of 0.43x against a 6.75x covenant, a completely undrawn $1.3B revolver to January 2030, and interest coverage of 14.4x; shareholders' equity positive at $2.167B after reading minus $453M a year ago; a 36-day cash conversion cycle; a 50.3% return on invested capital; and a beat on the last four consecutive prints, the most recent by 12.0% on adjusted EPS. We tested the ramp for tax and one-off inflation and it survived every test. It is a clean operating result.

What we are declining to pay for: 60.8x trailing GAAP earnings, 54.4x trailing adjusted, 23.9x the FY2027 consensus and 15.8x sales, on a business whose five-year revenue compound rate is 2.7% and whose FY2026 total is 4.6% above its FY2022 peak. A consensus that requires revenue 57.3% above the previous all-time record and non-GAAP net margins of 44.2% and 52.5% against 36.0% in the best quarter ever printed. A base fair value of $885 that is 4.7% above spot and 21.7% below the street's $1,130.15. A knowledge-base lane of one claim, which is a caution against precisely this valuation logic. A share price that opened up 5.6% today, on a day its entire complex was bid, and closed up 1.7%, 4.5% below its own high.

The distinction that matters. We are not saying Seagate is a bad business or that the nearline demand is fake. The nearline demand is real, the areal-density cost curve is real, the three-player structure is real, and the balance-sheet repair is one of the most impressive in this batch. We are saying that roughly half of FY2026's growth was price by the company's own attribution, that price in a commodity with a 10-K section headed "Price Erosion" is the least durable input there is, that the forward estimates embed a margin with no precedent anywhere in six years of history, and that we have no independent basis — from the filings, from the knowledge base or from our own arithmetic — for believing the cycle more than the market does. In the Synthos frame, a name where our fair value sits 22% below consensus and the conviction lane holds a single cautionary claim is a Watch by construction.

Pre-registered KILL criteria — what would take this to Avoid:

• September-quarter gross margin below 46.5% (the March-quarter level), which would reverse two quarters of expansion and convert the pricing regime from a plateau into a spike.
• Exabytes shipped falling sequentially while revenue holds or rises, which would prove price is carrying the entire load and would date the peak precisely.
• September-quarter revenue below roughly $3.6B, or a guide implying FY2027 revenue below roughly $16.0B — the low end of the consensus range. That is the FY2027 ramp failing, and there is nothing beneath a 24x multiple to catch it.
• Any of the three HDD suppliers announcing greenfield capacity, or Seagate's own capital expenditure running above roughly $250M in a single quarter. The oligopoly's pricing power exists only while nobody builds.
• The FY2026 10-K, due imminently, disclosing customer concentration materially above the approximately 10% last reported, which would convert a diversified-looking revenue base into a hyperscale single-point-of-failure.
• A break below the 200-day moving average of $524.55, which on a 2.10-beta name in a cyclical unwind is a genuine trend break rather than noise.

Pre-registered FLIP conditions — what would take this to Buy — Tactical:

• Price in the $650-700 zone (roughly 18-20x FY2027E, 11-12x FY2028E) with no deterioration in the fundamental picture. This is the single most likely path to a Buy and it is only 17-23% below today's close on a stock with a 643% annual range. At $675 the base case is +31% and the risk/reward becomes genuinely asymmetric.
• The 2026-10-28 print beating $4.090B of revenue with gross margin at or above 52.3% — both legs of 6a validated at once, at which point paying 24x forward is defensible because the pricing regime has been shown to hold across two quarters rather than one.
• A disclosed multi-year nearline supply and pricing agreement with a hyperscale customer, which would convert Seagate's revenue from spot-priced commodity to contracted capacity and would justify a structurally higher multiple. This is the single change that would most alter our framework.
• A reclaim of the 50-day moving average at $892.66 on above-average volume, converting today's failed gap into a base — though on its own this would be a technical flip only and would need a fundamental leg with it.
• Genuine name-level knowledge-base conviction appearing — a bullish claim about Seagate rather than a cautionary claim about the complex containing Seagate. At present the lane holds one claim and it points the other way, which alone is worth roughly one tier.

Where STX fits in the Synthos Framework Portfolio. The data-infrastructure / storage-hardware sleeve, at 0% today with a 1-1.5% target on a fill in the $650-700 zone or a confirmed October print. On the batch overlap question: STX and WDC are the two survivors of a three-player industry and now carry near-identical market capitalisations ($189.55B and $189.10B). They are one trade with two expressions, and the knowledge base's coverage sits almost entirely on the WDC side — five claims to Seagate's one — which is a fact about the knowledge base, not about the companies. Anyone holding one should not hold the other. Note also the batch-level irony worth stating plainly: the memory scarcity that is revenue for Micron is a competitive TAILWIND for Seagate, because it raises the cost per terabyte of the substitute product. Two names, the same physical shortage, and the transmission mechanism is completely different. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $845.35.

Single biggest risk: the FY2027 consensus is a peak-cycle extrapolation. Consensus requires revenue to rise 50.5% in the twelve months to July 2027, to $18.348B — 57.3% above the highest revenue this company has recorded in any year of its history — and a non-GAAP net margin of 44.2%, against 36.0% in the single best quarter it has ever printed. The entire valuation architecture rests on it. There is no valuation cushion to absorb a miss. At 24x a forward estimate that embeds an unprecedented margin, a 15% revenue disappointment does not produce a 15% price decline; it produces a de-rating on top of one, which is what the $490 bear case is. Three years ago, in fiscal 2023, this business had revenue of $7.384B and an operating margin of minus 4.6%.

Most fragile assumption in the price: that non-GAAP net margin expands from 29.2% to 44.2% by FY2027 and 52.5% by FY2028. Every other assumption is a volume assumption, and volume is observable quarter by quarter in the exabyte disclosure. This one is a price-and-margin assumption made against a company whose own annual report contains a section headed "Price Erosion", whose own attribution of the current expansion is "pricing actions undertaken by the Company", whose sales-incentive lever has already been pulled from 14% to 10% of gross revenue and cannot be pulled much further, and whose effective tax rate is rising off a 2.9% base supported by incentive programmes that begin expiring into fiscal 2036. If margin expands only half as far — to roughly 37% in FY2027 — FY2027 adjusted EPS is nearer $29.60 than $35.39, and at 25x that is $740, twelve percent below today's price.

Provenance & disclosures

• Traceability: 1 tagged knowledge-base claim names Seagate (raw hits 1, entity matches 1, text-only matches 0, discarded for collision 0 on the name; breadth 1, net conviction neutral-cautionary). The single claim is compound_and_friends, 2026-05-12, neutral, conviction 65, horizon principle, SPEAKER Josh Brown, no management role: memory-complex single-digit multiples are not cheapness because operating margins swing wildly and the discount is deserved. It is a caution, not a thesis, and it is the entire name-level lane. Five adjacent Western Digital claims were used as complex-level colour only and are quoted in Section 7: 2026-06-23 (bearish, conv 62, no named speaker, complex "most stretched in 25 years"); 2026-07-07 (bullish, conv 40, no named speaker, on the unwind); 2026-07-28 (bearish, conv 50, speaker Josh Brown, "a disgusting but necessary unwind" — the speaker named a specific AI laboratory in this claim and the name is redacted here per house style, replaced with "a frontier AI developer"); 2026-06-17 and 2026-01-24 (both bullish, conv 75 and 80, speaker Jordi Visser, skill weight 2.0, memory bull market). Two further independent claims are quoted for technology and cycle framing: Gavin Baker, 2024-02-01 (all hard drives will be replaced by flash) and Bill Gurley, 2024-08-09 (DRAM and hard drives have boom-bust shortage cycles out of step with reality). One claim was EXCLUDED on quarantine grounds — a 2025-10-10 Nvidia/Micron/ASML/Western Digital entity match flagged QUARANTINED:SUMMARY_OF_VISSER, which under the attribution policy is a summary of a speaker rather than the speaker. No claim in any lane carried a management speaker_role, so the half-weight management discount does not apply anywhere. Searches run: STX, Seagate, HDD, hard disk, hard drive, HAMR, nearline, mass capacity, exabyte, areal density, disk drive, data storage, Dave Mosley, Mosley, Starbox, Western Digital, SanDisk, storage, memory, hyperscaler capex across 51,928 distilled claims. Discards with reasons: the storage sweep returned 1 entity and 153 text-only matches, ALL discarded — the entity match was tagged "battery storage" and the text matches were grid batteries, residential solar-plus-storage, self-storage and marina real estate, cold storage of crypto keys, grain and fertiliser storage, oil and floating storage, tool-storage financing, datacentre storage software, and decentralised web3 storage; not one concerned a storage-hardware equity. The memory sweep returned 7 entity and 508 text-only matches and was abandoned as unusable (human memory, model context, institutional memory). The hyperscaler capex sweep returned 39 theme-level matches, none storage-specific. The anticipated STX ticker collisions with Starbox and STX Entertainment did NOT occur — both the STX and Starbox sweeps were clean. All quotes are verbatim from the stored claim text except the single redaction noted above.
• Data as-of: income statement, balance sheet and cash flow for fiscal 2026, ended 2026-07-03, sourced from the vendor payload (which carries filingDate: 2026-07-28) and cross-checked against the earnings-calendar actuals — Q4 revenue actual $3,629,000,000 matching inc_q exactly, and the four FY2026 quarterly revenue actuals summing to $12,195M, matching inc_a to the dollar. Most recent FILING-VERIFIED balance sheet: 2026-04-03, from the 10-Q filed 2026-04-29. Estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873600 = 2026-08-04T20:00:00Z ($845.35, +1.72%; open $877.55; day range $833.97–$885.00; 50-DMA $892.66; 200-DMA $524.55; RSI 51.5; MACD −15.46) · knowledge-base claims 2026-08-04. Fiscal/calendar note: Seagate's fiscal year ends on the Friday nearest 30 June; FY2026 ran to 2026-07-03 (53 weeks from FY2025's 2025-06-27) and FY2025 to 2025-06-27. "Q4 FY2026" means the quarter ended 2026-07-03. All figures come from the Synthos vendor data file for STX or from the SEC filings in the STX archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents and its central gap: 10-K filed 2025-08-01 (fiscal 2025, year ended 2025-06-27); 10-Q filed 2026-01-30 (December 2025 quarter); 10-Q filed 2026-04-29 (March 2026 quarter); 8-K filed 2026-05-28 (closing of privately negotiated exchanges of $185.908M principal of 3.50% Exchangeable Senior Notes due 2028 for $185.908M cash plus 2,023,124 ordinary shares, consummated 26-27 May 2026); 8-K filed 2026-06-12 (redemption notice calling ALL remaining 2028 Notes for redemption on 2026-09-08); 8-K filed 2026-07-28 (fourth-quarter and full-year FY2026 results, plus declaration of a $0.74 quarterly dividend payable 7 October 2026). The 2026-07-28 8-K is NOT a bare cover page — it carries Items 2.02, 7.01 and 9.01 in full prose — but it states that "The press release is attached to this Current Report on Form 8-K as Exhibit 99.1", and Exhibit 99.1 was not extracted and is absent from this archive. The file therefore contains no FY2026 financial statements of any kind. No filing in this archive covers fiscal 2026; the FY2026 10-K had not been filed as of this dive. Numeric tables are stripped from the extracted text throughout, which is why the debt schedule, the revenue disaggregation table and the earnings-per-share computation table are reconstructed from surrounding prose rather than read directly. Earnings-call transcripts are not available on the current data plan; management Q&A is absent from this dive.
• Where the filings contradicted or corrected the vendor (detailed in Section 8): market capitalisation and enterprise value rebuilt from 224,229,000 to 226,252,116 shares ($189.552B → $191.262B; EV $191.413B → $193.123B), because the 8-K of 2026-05-28 records 2,023,124 ordinary shares issued after the 10-Q cover-page count the vendor used; FY2026 capital expenditure rebuilt from the vendor's $0 to approximately $543M, and free cash flow from $3.674B to approximately $3.131B, because the 10-Q states "payments of $382 million for the purchase of property, equipment and leasehold improvements" for the nine months to 3 April 2026 — a figure that alone exceeds the $348M the vendor's own km_ttm block implies for the full year, and which its cf_a block contradicts; est.ebitdaAvg and est.ebitAvg rejected in full on a fabrication signature (exactly 19.7062% and 15.7328% of revenue in seven consecutive years); FY2029 and FY2030 estimate rows rejected entirely (2 and 1 analysts, implying 59.2% and 71.9% net margins); seg_geo relabelled from apparent customer geography to the 10-K's "bill from locations", with no demand conclusion drawn; seg_prod recorded as EMPTY with the consequence — no mass-capacity versus legacy split anywhere in this dive — stated rather than estimated around; lastDividend of $2.94 identified as the TTM total, with the 8-K giving the actual declared quarterly rate of $0.74 ($2.96 forward, 0.35% yield); quote.yearHigh/yearLow ($1,145.00/$144.75) rejected in favour of tech.hi52/lo52 ($1,094.04/$147.27) for internal consistency; book-equity metrics rejected as lenses (P/B 85.4x, ROE 348.2%) because equity was negative until this fiscal year; FY2026 equity-component fields recorded as unusable (retained earnings, paid-in capital and AOCI all zeroed); and the 2026-06-12 redemption notice logged as a financing event post-dating every financial statement in the file, with the resulting $185M–$214M cash outflow and ~2.03M–2.35M share issuance by 2026-09-08 quantified in Section 4. Where vendor and filing AGREED — worth recording, because these are the checks that most often fail: netDebtTTM of $1.861B is CORRECT (Seagate holds no short-term investments; the 10-Q lists only cash, operating cash flow and an undrawn revolver as liquidity sources); the market-cap share count matched the 10-Q cover page to eight shares; reportedCurrency is USD in every statement against a USD quote and a USD-denominated 10-K; seg_geo sums exactly to reported revenue in all three years; the FY2026 quarterly revenue actuals sum exactly to the annual line; and the effective tax rate, interest coverage, stock-compensation intensity and income-quality ratios all reconcile to the underlying statements.
• Basis note: analyst estimates in est are struck on an adjusted (non-GAAP) basis — verified by comparing the FY2026 row's $14.910 against GAAP diluted EPS of $13.90 and the reported adjusted total of $15.53, and the FY2025 row's $7.962 against GAAP's $6.77. All forward multiples in this dive use adjusted EPS. The trailing multiple is shown on both bases (60.8x GAAP, 54.4x adjusted). The gap between the two is 11.7% for FY2026, driven by a $105M legal settlement, $141M of net loss from debt transactions (including $120M of non-cash induced-conversion expense) and restructuring — all of which depress GAAP earnings rather than flattering them.
• Estimate coverage: 15 analysts on FY2027 revenue and 11 on FY2027 EPS; 14 and 10 on FY2028. The FY2029 EPS row rests on TWO analysts and FY2030 on ONE — both are excluded from every conclusion, and their implied net margins of 59.2% and 71.9% are the reason. The ebitdaAvg and ebitAvg rows are discarded outright as fabricated across all seven forward years.
• Peer note: the vendor peer set contains one genuine comparable — Western Digital ($548.62, $189.099B market capitalisation, +4.14% on the day) — alongside nine names that are not (Autodesk, Datadog, Electronic Arts, Fortinet, Infosys, Monolithic Power, Motorola Solutions, NXP Semiconductors, Super Micro Computer). No peer multiples are supplied in the file, so no peer-multiple comparison is drawn. The 10-K's own principal-competitor list is used instead — Kioxia, Micron Technology, Samsung Electronics, Sandisk, SK hynix, Toshiba and Western Digital — and its composition (five of seven are flash and memory manufacturers, not disk-drive makers) is treated as the central structural fact about this business.
• Fair-value caveat: the $490 / $885 / $1,360 anchors are multiples of the consensus adjusted-EPS distribution — 18x the FY2027 low of $27.296, 25x the FY2027 mean of $35.393, and 17x the FY2028 high of $79.892. Stated arithmetic, not a discounted cash flow. The base is deliberately sensitivity-disclosed: 22x gives $779, which is 7.9% BELOW spot; 28x gives $991. The whole answer lives in a 22-28x band on FY2027 and no greater precision is claimed. The base assumes substantial multiple compression (23.9x forward today to roughly 15.5x forward in twelve months) on the explicit reasoning that cyclicals at peak margins de-rate; the bull case assumes no re-rating at all, only that the top of the analyst range is right; the bear case assumes the pricing regime breaks and finds support near the 200-day average. The most fragile input in all three is the implied non-GAAP net margin embedded in consensus — 44.2% in FY2027 and 52.5% in FY2028 against 36.0% in the best quarter this company has ever recorded — and it is disclosed as such rather than buried.
• Timing: fiscal 2026 fourth-quarter and full-year results were released 2026-07-28, seven days before this dive, so the full fiscal year is reflected in the data. The next print is 2026-10-28, 85 days away. A second, harder-dated event sits inside the window: the 3.50% Exchangeable Senior Notes due 2028 are called for redemption on 2026-09-08, 35 days away, producing an estimated $185M–$214M cash outflow and approximately 2.03M–2.35M new ordinary shares. 2026-08-04 was a violent sector-wide semiconductor and memory up-day (MRVL +12.81%, INTC +10.92%, SNDK +10.79%, LRCX +7.90%, MU +7.62%, AMD +7.00%, KLAC +6.95%, AMAT +5.48%, ASML +4.22%, WDC +4.14%); STX's +1.72% was the laggard of that group and the stock closed 4.48% below its own opening gap. No company-specific news for 2026-08-04 appears in this file. Both the timing and the relative weakness are explicit inputs to the Watch verdict rather than oversights.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full.