Hewlett Packard Enterprise HPE

EARNINGS BANNER — Q3 FY2026 reports 2026-09-02, 29 days away

Hewlett Packard Enterprise's fiscal year ends 31 October. Third-quarter results are due 2026-09-02 on consensus non-GAAP EPS of $0.921 and revenue of $11,927 million — an 11.7% sequential increase on the April quarter's $10,678 million. The last print, on 2026-06-01, beat the consensus EPS estimate by 47.7%. The verdict below is set with that timing as an explicit input: this dive declines to buy into a print that lands four weeks from now on a stock that has risen 82.5% in three months, and the upgrade conditions in Section 10 are written to be resolved by exactly that report.

Reference table

1. What the business is now, and the acquisition that changed it

Hewlett Packard Enterprise is an enterprise technology company headquartered in Spring, Texas, with 67,000 employees, run by president and chief executive Antonio Fabio Neri. Its fiscal year ends 31 October.

The segment structure changed on 2025-11-01, and the vendor payload still reports the old one

From the 10-Q, verbatim:

> "Effective November 1, 2025, HPE implemented an organizational change by (i) merging the Server, Hybrid Cloud, and Financial Services business segments into a new segment named Cloud & Artificial Intelligence ("AI") and (ii) transferring the Telco and Instant On businesses from the Networking segment to the Corporate Investments and Other segment. As a result, the Company's organizational structure consists of the following segments: (i) Cloud & AI; (ii) Networking; and (iii) Corporate Investments and Other. The Company has reflected these changes to its segment information retrospectively."

The vendor's seg_prod for FY2025 reports the OLD five-segment structure — Server Segment, Hybrid Cloud, Networking, Financial Services, Corporate Investments. It is not wrong for the year it labels, but it cannot be reconciled to anything the company now reports, and any comparison built across it will be comparing different things. All segment figures in this dive come from the 10-Q.

Segment results — the table that carries the thesis

Three months ended 30 April, $M, filing-verified:

Six months ended 30 April, $M:

Four things to take from these two tables, in order of importance.

First, the margin arithmetic is the whole investment case. Networking earns 21.6% and Cloud & AI 12.4%. Networking is 25.2% of quarterly revenue but 38.1% of segment earnings. Every point of revenue mix that moves from Cloud & AI to Networking is worth roughly nine points of incremental operating margin on that dollar. HPE describes the Juniper rationale in its own 10-K as advancing "HPE's portfolio mix shift toward higher-growth solutions and strengthen its networking business" — half-weighted as management framing, but the segment table is not framing.

Second, the Networking growth is acquisition, and it should not be read as organic. Juniper closed 2025-07-02, so the year-ago quarter contains none of it. A 148% increase against a base that excludes a $13.6 billion acquisition is arithmetic, not momentum. The first genuinely comparable quarter will be the one reporting on 2026-09-02, which includes Juniper on both sides.

Third, the organic engine grew 9.8% in the half. Cloud & AI — the old Server, Hybrid Cloud and Financial Services businesses, still 70.3% of quarterly revenue — rose 22.9% in the quarter but only 9.8% across the six months. That is respectable and it is not the 40% headline.

Fourth, and genuinely good: Cloud & AI margin nearly doubled. From 6.6% to 12.4% in the quarter and 7.5% to 11.4% in the half. The server business became roughly twice as profitable per dollar of revenue in twelve months, which is what pricing power in a supply-constrained AI-systems market looks like. It is also 11-12%, which is what assembling other people's silicon pays.

The Juniper acquisition, in the 10-K's own numbers

$13,485 million of the $13,625 million consideration — 99.0% — landed on the balance sheet as goodwill and intangibles. The 10-K states the cash was funded "through cash on hand, including proceeds and term loan drawdowns from the financings in fiscal 2024, and commercial paper issuances" — which is where the mandatory convertible preferred stock in Section 4 comes from. Outstanding Juniper equity awards were converted at an exchange ratio of approximately 2.1 HPE shares per Juniper award, with a total fair value of $927 million, of which $239 million sat in consideration and the balance runs through future compensation expense. HPE recorded $192 million of stock-based compensation on those assumed awards in fiscal 2025 alone.

The $6,219 million of intangibles is what produces the $323 million of quarterly amortisation that drives the GAAP-to-non-GAAP gap in Section 3.

Geography — seg_geo is clean

This block reconciles exactly to consolidated FY2025 revenue of $34,296M and correctly identifies the Americas as the largest market. Against eight broken seg_geo blocks in twelve names in the previous batch, including one that named the wrong country as the largest market, this is worth recording as clean. The growth is concentrated in the Americas at 19.7% while Asia Pacific and Japan grew 3.5% — consistent with a United States-led AI infrastructure build.

2. The cash-flow inflection, which is the strongest fact in this file

From the 10-Q's own free-cash-flow reconciliation, verbatim table, $M:

A $3,347 million year-on-year swing in half-year free cash flow, on the company's own reconciliation, published in a filed document. The operating cash-flow line alone moved $3,439 million, from a use of $851 million to a source of $2,588 million.

The composition, from the same statement, is worth naming because it is not all quality. The half included an inventory build of $2,956 million — cash out — offset by an accounts-payable increase of $3,562 million — cash in. A company financing a $3.0 billion inventory build with $3.6 billion of supplier credit is a company whose working capital swing flatters the current period and will reverse. The half also carried $305 million of provisions for inventory and credit losses, a $244 million gain on the sale of a business, and $97 million of H3C divestiture-related severance.

On the house rule, free cash flow is recomputed as operating cash flow less filed capital expenditure, without the company's add-backs. Trailing twelve months to 2026-04-30: operating cash flow of $6,358 million (FY2025's $2,919M, less H1 FY2025's −$851M, plus H1 FY2026's $2,588M) less capital expenditure of $2,369 million (FY2025's $2,292M, less H1 FY2025's $1,075M, plus H1 FY2026's $1,152M) equals $3,989 million — a 5.75% free-cash-flow yield.

The vendor reports a trailing free-cash-flow yield of 8.13%, which is 41% too high, and Section 7 explains exactly why.

H3C — a disposal in progress

The 10-Q's subsequent-events disclosure: "the Company closed on the sale of the remaining 5.2% of the total issued share capital of H3C for approximately $370 million" on 2026-05-28, after an earlier tranche that produced a $244 million gain and $210 million of proceeds in the half. HPE has now exited its Chinese joint venture entirely. The proceeds are real cash and the severance costs of $97 million are the price of it. No vendor field records any of this.

3. Earnings quality — the 77% gap, and where it comes from

Consolidated income statement, three and six months ended 30 April, $M, filing-verified:

Research and development rose 70.7% to $922 million in the quarter — a genuine investment step, and the largest percentage increase in any expense line.

The GAAP-to-non-GAAP gap, computed here rather than taken from a bridge the 8-K did not carry. Net earnings attributable to HPE for the quarter were $624 million. Preferred dividends run at approximately $28.6 million a quarter (30 million shares at $0.953125), leaving roughly $595 million to common on 1,335 million weighted-average basic shares, or approximately $0.45 of GAAP basic EPS. The reported non-GAAP figure is $0.79. That is a 77% gap.

The three components are all disclosed:

• Amortisation of intangible assets: $323 million in the quarter, against $37 million a year earlier. Nearly all of it is Juniper's $6,219 million of acquired intangibles running through the income statement. This is a non-cash charge against cash that was genuinely spent — $13.4 billion of it — and it is the single largest reconciling item.
• Stock-based compensation: $218 million in the quarter, per the statement of stockholders' equity. This is a real economic cost transferred from the income statement to the share count.
• Acquisition, disposition and other charges: $78 million in the quarter, $195 million in the half.

The honest position, stated once. We use consensus non-GAAP EPS for the forward multiples in Section 5 because that is the basis every analyst estimate in the file is struck on, and mixing bases produces nonsense. But a 77% adjustment is large, two of its three components are permanent features of an acquisitive hardware company, and the reader should hold both numbers: 13.0x FY2027 on the non-GAAP basis is roughly 20-23x on a GAAP basis, and the truth of the business is somewhere between.

The beat record

Five consecutive beats, accelerating, with the most recent at 47.7% on a 9.2% revenue beat. The consensus is visibly behind a business it is struggling to model through an integration — which cuts both ways, and Section 5c takes it seriously in both directions.

4. Capital structure — the preferred stock the payload records as zero

Balance sheet at 2026-04-30 against 2025-10-31, $M, filing-verified:

Debt is falling and that is the correct headline: down $1,119 million in six months, with short-term borrowings down $1,600 million. Commercial paper capacity is $5.75 billion and the revolving credit facility $5.25 billion, with combined borrowings capped at $5.75 billion.

The mandatory convertible, in the 10-K's own words

> "In September 2024, the Company issued 30 million shares of 7.625% Series C Mandatory Convertible Preferred Stock, par value $0.01 per share (the 'Preferred Stock'), for an aggregate purchase price of $1.5 billion, less issuance costs of $38 million. As of October 31, 2025, 30 million shares of the Preferred Stock were outstanding. Unless converted earlier in accordance with its terms, each share of Preferred Stock will automatically convert on the mandatory conversion date, which is expected to be September 1, 2027, into between 2.5352 shares and 3.1056 shares of common stock..."

The arithmetic, and it matters. Thirty million preferred shares converting at between 2.5352 and 3.1056 common shares each produces between 76.06 million and 93.17 million new common shares. The minimum conversion rate applies when the common price is high, which it emphatically now is. The 10-Q confirms the operative figure: the dilutive effect of the preferred in the non-GAAP diluted share count is 76 million shares for both the quarter and the half.

76 million shares on 1,323 million outstanding is 5.75% dilution, contractually scheduled for 2027-09-01, and it is unavoidable — that is what "mandatory" means. At the current $52.39 those shares are worth $3.99 billion against the $1.5 billion HPE raised. The preferred holders will have roughly tripled their money at the common holders' expense, which is the price HPE paid in September 2024 for financing an acquisition it had not yet closed.

The dividend cost while it waits: $0.953125 per preferred share per quarter, declared by the board on 2026-08-04 — the day of this dive — payable 2026-09-01 to holders of record 2026-08-15. Thirty million shares gives $28.6 million a quarter, $114.4 million a year. This is why FY2025's eps reads −$0.0446 in the payload against positive net income of $57 million: the preferred dividend exceeded the profit.

None of this is visible in the vendor payload, whose preferredStock field reads $0 in FY2023, FY2024 and FY2025. It is technically defensible — the preferred is carried at $0.01 par and the $1.5 billion sits in additional paid-in capital — and it is economically useless. Section 7 records it as the most consequential omission in this file.

The correct diluted market capitalisation

Reported market capitalisation: $69,375M on 1,324.20 million shares at $52.39. Adding the 76.06 million minimum conversion shares gives 1,400.26 million and a fully-diluted market capitalisation of $73,360M — 5.75% higher. Adjusted enterprise value is approximately $89,314M against the vendor's $85,329M. Every per-share and enterprise-value figure in this dive that uses the reported count is labelled; the fair values in Section 5 are struck on consensus EPS, which already uses the 1,432 million diluted count and therefore already carries the dilution.

5. Valuation — priced in or room?

At $52.39 (market cap $69.38B, 1,324,203,521 shares):

Estimate coverage is good on the near years. FY2026 and FY2027 each rest on 11 analysts for EPS and 14 for revenue, and the FY2027 EPS range is $3.878 to $4.167 — a 7.4% spread, which is unusually tight and reflects a street that agrees about the destination. FY2029 rests on 2 analysts and is excluded from every conclusion. FY2028 rests on 4 and is used only as a cross-check.

est.ebitdaAvg and est.ebitAvg are REJECTED for a fixed-ratio fabrication signature. From FY2023 through FY2029, ebitdaAvg is exactly 16.23% of revenueAvg in every year and ebitAvg exactly 7.50% — with the single exception of the FY2022 row (22.49% and 4.26%), which locates the boundary. Seven consecutive years at constant margins for a company whose segment margin moved from 8.9% to 14.3% in twelve months is not an estimate. All forward valuation uses epsAvg.

Enterprise value — VERIFIED CLEAN, including the noncontrolling interests, and this is worth stating. enterpriseValueTTM of $85,329M less market capitalisation of $69,375M implies net debt plus other adjustments of $15,954M. The filing gives total debt of $21,246M less cash and restricted cash of $5,354M = $15,892M, plus noncontrolling interests of $61M = $15,953M. A one-million-dollar difference on an $85 billion figure. The vendor used the current filed balance sheet AND added the noncontrolling interests — the omission this programme found on seven of twelve names in the last batch does not occur here. The residual is the $1.5 billion of mandatory convertible preferred, which is arguably equity and arguably not; we treat it as equity, add the 76.06 million conversion shares to the share count instead, and give an adjusted enterprise value of approximately $89,314M.

Peer context. The vendor peer set is largely unusable: ASE Technology, Ciena, Credo, Ericsson, Fiserv (a payments processor), Keysight, Nokia, "Everpure, Inc" at the PSTG ticker (a naming error — PSTG is Pure Storage), Super Micro and "Sandisk Corporation" priced at $1,427.62 with a $211 billion market capitalisation, which is not a figure this dive can reconcile to anything. The genuine comparables — Dell, Cisco, Arista, Lenovo, IBM — are absent entirely. No peer multiple comparison is drawn, and the peer block is flagged in Section 7.

5a. What today's price assumes (the inversion)

At $52.39 — 15.3x FY2026 consensus non-GAAP EPS, 13.0x FY2027, 2.77x book — the price embeds:

• Revenue reaches $44,969 million in FY2026. (Consensus; 14 analysts.) The first half delivered $19,979 million, so the second half needs approximately $24,990 million — a 25.1% sequential step. The Q3 consensus of $11,927 million implies Q4 of roughly $13,063 million against Q2's $10,678 million. This is the most fragile assumption in the price, and it resolves on 2026-09-02.
• Non-GAAP EPS reaches $3.415 in FY2026 and $4.018 in FY2027. (Consensus; 11 analysts each.) The half delivered $1.44 ($0.65 plus $0.79), so the second half needs $1.975, of which Q3 consensus is $0.921. Given five consecutive beats culminating in 47.7%, the FY2026 figure looks beatable; the FY2027 figure requires 17.6% growth on top of a year that already grew 31%.
• The Networking margin holds near 21.6% as Juniper anniversaries. (Our derivation; the filing supplies the margin.) Note that Networking's margin FELL year on year, from 25.0% to 21.6% in the quarter and 27.4% to 22.6% in the half — the pre-acquisition Networking business was more profitable than the combined one. Whether that is integration cost or acquired mix is the question the 2026-09-02 print begins to answer.
• The market keeps paying roughly 13-15x forward non-GAAP earnings. (Our number.) At 11x FY2027E the stock is $44.20; at 17x it is $68.31. Note what this multiple is being applied to: a number 77% above GAAP.
• The 2027-09-01 mandatory conversion is absorbed without a de-rating. (Our number; the terms are the company's.) 76.06 million shares, 5.75% of the count, arriving on a scheduled date thirteen months from now. Consensus EPS already uses the 1,432 million diluted count, so the earnings arithmetic is correct — what is not priced anywhere is the supply of stock itself, and mandatory convertibles typically produce hedging pressure into the conversion window.

5b. The return bridge (why the multiple moves)

Expected return over the next twelve months decomposes as: non-GAAP EPS growth (+17.6%, FY2026E $3.415 to FY2027E $4.018) + multiple drift (assumed COMPRESSION, 15.3x on the forward year to about 14.9x on the then-forward year) + dividend yield (+1.06%) ≈ +14% to +16%.

Our base assumes the multiple compresses slightly, and the reason is the dilution and the earnings quality rather than the growth. Almost the entire base-case return is earnings growth — which is the return shape this programme prefers — and it is available at 13.0x because the market has not yet decided whether the Juniper-era HPE is a networking company at 20x or a server assembler at 10x.

The honest statement of where the answer lives. At 11x FY2027E consensus the stock is $44.20 (−15.6%); at 17x it is $68.31 (+30.4%). A six-turn multiple band on a business whose two segments earn 21.6% and 12.4% is not analyst indecision — it is a real, unresolved question about what kind of company this is, and the segment mix disclosed each quarter is how it gets answered.

5c. Variant perception (where we differ, what would surprise)

• We are materially BELOW the street and we say so. Our base of $60 is 13.4% below the consensus $69.27, and the lowest published target, $62, is already 18.3% above spot. Not one covering analyst has a target below the current price, while the rating distribution is 18 buy, 20 hold, 1 sell with the vendor consensus reading HOLD. That combination — every target above the price, a majority of ratings at Hold — usually means targets were raised in response to the move rather than ahead of it, and it is a caution rather than a comfort.
• Our disagreement is about timing and dilution, not about the business. We think the segment table is genuinely good, the margin arithmetic genuinely favourable, and 13.0x FY2027 genuinely not expensive. What we decline to do is pay it four weeks before a print that has to validate a 25% sequential revenue step, on a stock 76.3% above its 200-day average, with 76.06 million shares of scheduled dilution thirteen months out. That is a position on entry, not on value, and it is stated as one.
• We think the consensus is more likely too low than too high, and we still do not act on it. Five consecutive beats, the last at 47.7%, and a FY2027 EPS range of only 7.4% suggest a street that is behind and clustered. Watchable number: the FY2027 consensus non-GAAP EPS, currently $4.018. If it is above $4.50 after the 2026-09-02 print, the multiple at spot falls to 11.6x and this becomes a Buy on arithmetic alone.
• We think the Networking margin decline is under-discussed. Networking earned 25.0% in the year-ago quarter and 21.6% now; 27.4% in the year-ago half and 22.6% now. The market is celebrating the revenue and has not asked why the margin on the acquired business is lower than the margin on the business that acquired it. Watchable number: Networking segment operating margin on 2026-09-02. Above 23% and the integration is accretive to quality as well as scale; below 20% and the $13.6 billion looks like revenue bought at a discount to its own history.
• Positive surprise that would force a re-rate: a Q3 print delivering revenue above $12.5 billion with Cloud & AI organic growth above 15% and Networking margin above 23%, which would confirm both legs at once; or a guided FY2027 above the current $4.018 consensus.
• Negative surprise that would break the thesis: Q3 revenue below $11.5 billion, which would make the FY2026 consensus unreachable; Cloud & AI margin falling back below 10%, indicating AI-server pricing has peaked; any goodwill impairment against the $23,770 million carried — the April 2025 quarter already carried $1,361 million of impairments, so this is precedent, not speculation; or an acceleration of preferred conversion.

Synthos fair values

All three anchors are multiples of the FY2027 consensus non-GAAP EPS distribution (mean $4.018, low $3.878, high $4.167, 11 analysts), which already uses the 1,432 million fully diluted share count including the preferred conversion.

• Bear ~$40 — 10.3x the FY2027 consensus LOW of $3.878, and 11.7x FY2026E. Cross-check: 9.1x FY2028E; 101% above the 52-week low of $19.87 and 14.2% below the 50-day average. The scenario: the second-half ramp does not arrive, AI-server margins compress back toward 8%, Networking integration costs persist, and the market re-rates HPE as the hardware assembler it was for a decade. −23.7%.
• Base ~$60 — 14.9x the FY2027 consensus MEAN of $4.018. Cross-check: 17.6x FY2026E; 13.6x FY2028E; 3.17x book. Sensitivity, stated openly: 13x FY2027E gives $52.23 — spot, to within 0.3% — and 17x gives $68.31. The stock is currently priced at almost exactly 13x, so the entire base case is a four-turn re-rating, and Section 5b says which way and why. The scenario: the FY2026 second half lands close to consensus, Networking margins stabilise near 22%, the mandatory conversion passes without incident and the market settles on a mid-teens multiple for a mixed hardware-and-networking business. +14.5%.
• Bull ~$80 — 19.2x the FY2027 consensus HIGH of $4.167. Cross-check: 18.2x FY2028E; 42.5% above the 52-week high of $56.15; equal to the highest published street target. The scenario: the AI-systems cycle runs longer than modelled, Cloud & AI margins hold above 13%, Networking grows organically at Arista-like rates, and the market prices HPE as a networking company with a server attachment. +52.7%.

Base is 14.5% above spot; asymmetry roughly 2.22:1 (23.7% down, 52.7% up), plus a 1.06% dividend. A 14.5% base with 2.22:1 asymmetry would ordinarily clear the Tactical bar. What holds it back is stated without hedging: the print is 29 days away, the stock is up 82.5% in three months and sits 76.3% above its 200-day average, and Section 10's upgrade conditions are all resolvable on 2026-09-02. Waiting four weeks for information that materially changes the distribution is not caution; it is the correct use of a known date.

6. Knowledge base — 87 raw hits, three survivors, all bullish, all late

Raw hits: 87. After a case-sensitive entity re-run: 3. Name-level claims on HPE: 3, and ALL THREE are BULLISH. Discarded: 84.

The primary sweep ran the entity terms HPE, Hewlett Packard Enterprise, Hewlett-Packard Enterprise, Juniper and Juniper Networks, plus free text on HPC and supercomputing, across all 52,021 distilled claims. It returned 87 hits across 19 channels, of which 44 — more than half — come from a single channel, jensen_huang, and are that speaker's own commentary on accelerated computing. Seven come from eli_lilly_ceo, the store defect this programme has documented as carrying other companies' claims and which is excluded here on content mismatch. A case-sensitive entity re-run collapses 87 to 3.

The three surviving claims, verbatim:

> 2026-06-17 · BULLISH · conviction 65 · horizon: thesis · entities: DELL, HPE, PSTG · channel: jordi_visser · speaker: Jordi Visser

> "AI-factory infrastructure names — Dell, HPE, Pure Storage — are part of the buildout; Dell already reached the 'insane numbers' earnings mode."

> 2026-06-02 · BULLISH · conviction 75 · horizon: thesis · entities: DELL, HPE · channel: andreas_steno · speaker: Andreas Steno · role: independent

> "Computer/server makers have barely started: Dell's blowout AI-server numbers plus spillover to consumer products; HP Enterprise's new Nvidia production-line partnership looks compelling — Dell trading like a meme coin."

> 2026-04-11 · BULLISH · conviction 85 · horizon: thesis · entities: NVDA, GLW, DELL, HPE · channel: jordi_visser · speaker: jordi_visser · role: independent

> "Underinvestment vs massive AI compute/inference demand — not enough compute; semis and hardware keep rising all year, Nvidia way too cheap relative to demand; CapEx receivers' earnings keep growing."

What this lane is, stated without inflation, in four observations.

First, and unusually for this programme: all three carry a NAMED SPEAKER. Jordi Visser twice, Andreas Steno once. The overwhelming majority of claims in this store carry speaker: null and are channel-attributed only — the weakest sourcing the 4-lane policy admits. These do not. That is a genuine quality upgrade over most lanes in this batch and it is why the conviction reads positive-low rather than empty.

Second, breadth is 2, not 3, and we say so. Two of the three claims come from the same channel and the same speaker eight weeks apart. The independent-source count is two: Jordi Visser and Andreas Steno. Treating a repeated view from one voice as two data points is exactly the concentration error a breadth metric exists to prevent.

Third, every one of them is a basket call, not a stock call. HPE appears alongside DELL in all three and alongside PSTG, NVDA and GLW in two. Not one claim distinguishes HPE from Dell, discusses the Juniper acquisition, mentions networking margins, or names a price. The closest to a company-specific observation is Steno's reference to "HP Enterprise's new Nvidia production-line partnership" — which is genuinely HPE-specific and which we hold at the weight of a single unverified secondhand reference, since no filing in this archive discusses it.

Fourth, and most important for weighting: they are all late. The claims are dated 2026-04-11, 2026-06-02 and 2026-06-17. HPE's twelve-month return is 164.5% and its three-month return 82.5%, so all three were written after the stock had already doubled from its 52-week low of $19.87. A bullish claim written after a double is a description of momentum, not a discovery of value, and it is graded accordingly. The highest-conviction of the three — Visser at 85, the highest single conviction figure in this batch — is also the earliest and the most general, being principally a claim about Nvidia and compute scarcity in which HPE is one of four named beneficiaries.

Attribution note. No claim carries a management speaker_role; two carry independent, which per house rule is judged on content and is correct in both cases — these are independent market commentators, not company voices. The eli_lilly_ceo channel appeared 7 times in the raw sweep and is excluded on content mismatch, consistent with the documented store defect in which that channel carries other companies' claims.

Conclusion. Breadth 3 (independent sources 2), claim count 3, net conviction positive-low. The Synthos knowledge base is bullish on Hewlett Packard Enterprise. It is bullish in the way a crowd is bullish after a move — as one of several names in an AI-infrastructure basket, from two voices, all within five months of this dive, none of them naming the acquisition that actually changed the business. That is a real positive lane and it is not conviction, and the distinction is the whole point of reporting it this way.

7. Data integrity — what we rejected and why

Seven findings. HPE's payload has one omission of genuine economic consequence, one capital-expenditure defect, one stale segment structure and one peer block containing an apparent data-entry error — against which the enterprise value, the share count and the geographic segmentation are all clean, which is an unusually good record.

1. THE $1.5 BILLION MANDATORY CONVERTIBLE PREFERRED IS RECORDED AS ZERO — the most consequential omission in this file. bal_a reports preferredStock: 0 for FY2023, FY2024 and FY2025. The 10-Q balance sheet reports 30,000,000 shares of 7.625% Series C Mandatory Convertible Preferred Stock issued and outstanding, raised for $1.5 billion in September 2024, automatically converting on 2027-09-01 into between 2.5352 and 3.1056 common shares each — a minimum of 76.06 million shares, 5.75% of the count — and costing $114.4 million a year in dividends, with the most recent declaration made on 2026-08-04, the day of this dive. The vendor's zero is technically defensible because the preferred carries at $0.01 par with the proceeds in additional paid-in capital, and it is economically useless. The visible consequence: FY2025 eps reads MINUS $0.0446 in the payload against POSITIVE net income of $57 million — an apparent contradiction that only resolves once the reader knows about a preferred dividend the payload does not disclose. Fully diluted market capitalisation is $73,360M, not $69,375M.

2. capitalExpenditure in the TTM block is roughly 70% understated — the definition-change signature. capexPerShareTTM of $0.5378 on 1,324 million shares implies trailing capex of approximately $712 million. The filed figures are $1,152 million for the first half of FY2026 alone and $1,075 million for the first half of FY2025, against cf_a's FY2025 figure of $2,292 million — which is internally consistent with the halves. Trailing capex is approximately $2,369 million, so the TTM block understates it by roughly 70%. capexToDepreciationTTM of 0.2167 against annual figures running 0.83-1.26 confirms the swing. Consequently freeCashFlowYieldTTM of 8.13% is REJECTED — the correct figure on operating cash flow less filed capex is approximately $3,989 million, a 5.75% yield, 41% lower. Also rejected: freeCashFlowPerShareTTM ($4.225), priceToFreeCashFlowRatioTTM (12.30x), evToFreeCashFlowTTM (15.13x), capexToOperatingCashFlowTTM (0.1129), capexToRevenueTTM (0.0185), capitalExpenditureCoverageRatioTTM (8.855x) and freeCashFlowToEquityTTM ($7,647M). Note the annual cf_a capex figures — FY2022 $3,122M, FY2023 $2,828M, FY2024 $2,367M, FY2025 $2,292M — are all approximately correct; the defect is confined to the trailing block.

3. seg_prod reports a segment structure the company abandoned on 2025-11-01 — REJECTED for current use. The FY2025 row carries Server Segment $17,745M, Networking $6,850M, Hybrid Cloud $5,754M, Financial Services $3,507M, Corporate Investments $776M and Intersegment Eliminations −$333M. Effective 2025-11-01 the company merged Server, Hybrid Cloud and Financial Services into Cloud & AI and moved Telco and Instant On out of Networking, restating retrospectively. The vendor's structure is not wrong for the fiscal year it labels, but it cannot be reconciled to anything the company now publishes, and the FY2023 and earlier rows use three further different structures (Compute / HPC & MCS / Storage / Intelligent Edge in FY2023; Hybrid IT in FY2019; Enterprise Group in FY2017). All segment figures in this dive come from the 10-Q.

4. totalDebt is inflated by lease obligations — CONFIRMED to the dollar. The vendor's FY2025 totalDebt reads $24,077M. The 10-Q's liquidity section states total debt of $22,365M at the same date. The difference is $1,712 million, which is exactly the vendor's own capitalLeaseObligations line of $1,712 million. A 7.7% overstatement, arising entirely from including finance leases the company excludes. The same defect fires in FY2024 ($19,816M vendor against a filed figure implied by the same construction, $1,570M of leases) and FY2023 ($1,160M). This is the operating-lease inflation class, confirmed by exact arithmetic, and it has now fired on six of twelve names in three consecutive batches.

5. est.ebitdaAvg and est.ebitAvg carry a fixed-ratio fabrication signature — REJECTED. From FY2023 through FY2029, ebitdaAvg is exactly 16.23% of revenueAvg and ebitAvg exactly 7.50%, in every year. The FY2022 row (22.49% and 4.26%) sits outside the pattern and marks its start. Seven consecutive years at constant margins for a company whose segment operating margin moved from 8.9% to 14.3% in twelve months. All forward valuation uses epsAvg. The FY2029 row rests on 2 analysts and is excluded.

6. The peer block contains at least two errors and is discarded. The set is ASE Technology, Ciena, Credo, Ericsson, Fiserv (a payments processor, unrelated to enterprise hardware), Keysight, Nokia, "Everpure, Inc" carrying the PSTG ticker — PSTG is Pure Storage, and the company name is wrong — Super Micro, and "Sandisk Corporation" at a price of $1,427.62 and a market capitalisation of $211.4 billion, a figure this dive cannot reconcile to any plausible reading. The genuine comparables — Dell Technologies, Cisco, Arista Networks, IBM, Lenovo — are absent entirely. No peer comparison is drawn.

7. tangibleBookValuePerShareTTM is MINUS $3.263 — not an error, and stated because it is easily misread. Goodwill of $23,770M plus intangibles of $6,368M equals $30,138M against total equity of $24,754M at fiscal year end, of which Juniper contributed $7,272M of goodwill and $6,219M of intangibles. tangibleAssetValueTTM reads MINUS $4,356M. These are arithmetically correct and they mean that on a liquidation basis the common equity is worth less than nothing, which is normal for a serial acquirer and is why book-value anchors are not used in Section 5. bookValuePerShareTTM of $18.945 and shareholdersEquityPerShareTTM of $18.945 are identical, consistent with the preferred being carried at par and confirming finding 1 from a second direction.

Verified CLEAN — recorded because clean checks are findings:

• Enterprise value INCLUDING noncontrolling interests. enterpriseValueTTM of $85,329M less market capitalisation of $69,375M implies $15,954M, against a rebuilt total debt $21,246M less cash $5,354M plus noncontrolling interests $61M = $15,953M. A $1 million difference on an $85 billion figure, using the current filed balance sheet and correctly adding the NCI. This is the omission that fired on seven of twelve names in the previous batch and it does not fire here.
• Share count — exact. The 10-Q cover states 1,324,203,521 shares outstanding at 2026-05-26 against 1,324,204,000 implied by market capitalisation ÷ price. An exact match.
• seg_geo is CLEAN and correctly ordered. Americas $15,848M, EMEA $11,510M, Asia Pacific and Japan $6,938M, summing to $34,296M — exactly consolidated FY2025 revenue — with the Americas correctly identified as the largest market.
• tech.max_dd_from_peak of −6.696% coincides exactly with pct_from_hi. Per the data contract this is correct, not a defect — the six-year peak falls inside the last twelve months, which for a stock up 164.5% is trivially true. Described as the current distance from the high, never as a maximum drawdown.
• quote versus tech. tech.last of $52.39 equals quote.price of $52.39 exactly. quote.yearHigh/yearLow ($64.25/$19.635) against tech.hi52/lo52 ($56.15/$19.87) — a 14.4% discrepancy on the high, the largest in this batch. tech is used throughout, and the quote.yearHigh of $64.25 is not used anywhere.
• dividendPerShareTTM of $0.5575 matches profile.lastDividend of 0.5575. Clean.

Vendor composite rating — B / 3 overall, with debtToEquityScore of 1 out of 5 and priceToBookScore of 4. The debt sub-score points at the right thing and the price-to-book sub-score points at a metric that is meaningless when tangible book is negative. The discountedCashFlowScore of 3 rests on the free-cash-flow figures rejected in finding 2 and carries no weight.

Non-equity tripwire — checked and passed, with a note. HPE is common stock, par value $0.01, NYSE-listed, with 9,600,000,000 shares authorised and 1,323,294,768 issued and outstanding at 2026-04-30. Price of $52.39 is not par-like; beta is 1.438; volume was 21.9M shares (~$1.15 billion of turnover, the heaviest in this batch); the 52-week band of $19.87 to $56.15 is a 183% range; the dividend is a regular quarterly common dividend. This is common equity. Note the separately listed 7.625% Series C Mandatory Convertible Preferred Stock, ticker HPEPrC on the NYSE, which is NOT this security and which Section 4 describes in full.

8. Technicals

• Price $52.39. −6.7% from the 52-week high of $56.15; +163.7% above the 52-week low of $19.87. Position within the annual range: 90th percentile.
• Both moving averages are far below the price and the gap is extreme. +12.3% above a 50-day average of $46.66; +76.3% above a 200-day average of $29.72. The 200-day figure is the single most important technical fact in this file: a stock trading at 1.76 times its own one-year average price has no reference level beneath it until $46.66, and then nothing until the mid-thirties.
• RSI 62.2 — strong, not overbought, which after a 164% year is itself notable and indicates the advance has been orderly rather than parabolic. MACD +1.11 — positive.
• Current distance from the six-year peak: −6.7%, identical to the 252-day figure. Per the data contract that coincidence is correct.
• Relative performance, and the numbers are extraordinary: 3-month +82.5% against SPY +7.6% and QQQ +7.7% — outperforming by 75 points in a quarter; 6-month +138.0% against SPY +11.1%; 12-month +164.5% against SPY +24.3% and QQQ +30.8%. This is not a recovery from underperformance in the manner of several names in this programme; it is the largest single-name outperformance in this batch by a factor of more than two.
• Sentiment: 18 buy, 20 hold, 1 sell across 39 analysts, with the vendor consensus label reading HOLD. Consensus target $69.27 (+32.2%), median $69, high $80 (+52.7%), low $62 — 18.3% ABOVE spot. Every published target is above the price while a plurality of ratings is Hold. That pattern is characteristic of targets raised in response to a move rather than in anticipation of one, and it is treated as a caution.

Today's move and what it does to the entry

HPE closed 2026-08-04 at $52.39, up $2.15 or 4.28%, from $50.24. It opened at $52.61, traded $50.39 to $53.42, and closed on 21.9 million shares — approximately $1.15 billion of turnover, the heaviest in this batch. The only filing dated 2026-08-04 is the 8-K declaring the preferred dividend, which is routine and does not explain a 4.28% move.

The honest read: this is a chase, and we decline it. Up 4.28% today, 82.5% in three months, 164.5% in twelve, 76.3% above the 200-day average, 29 days before a print that has to validate a 25% sequential revenue step. A Watch on a business we like at a multiple we consider reasonable is a statement about entry timing, and it is a statement we would rather make explicitly than dress up as a valuation view.

9. Insiders — eight filings, one vesting event, one new director

There is not one open-market purchase and not one open-market sale in the file, and the correct conclusion is that this file contains no signal.

What happened is one routine vesting event and one board appointment. Kirt Karros, an officer, had 40,282 restricted stock units convert to common on 2026-07-20 at a reference price of $44.56, with 16,607 shares — 41.2% — withheld by the company for tax, leaving 23,675 held. The three fractional A-Award entries of 252.05, 151.80 and 93.27 units on 2026-07-15 are dividend-equivalent accruals on unvested awards, not grants. David I. Goulden filed a Form 3 on 2026-07-28 as a newly appointed director and received a standard 3,495-unit initial grant.

Two observations worth making.

The reference price is informative even though the transaction is not. Karros's vesting was priced at $44.56 on 2026-07-20. The stock closed at $52.39 on 2026-08-04 — 17.6% higher fifteen days later. That is the pace of the move this dive is declining to chase.

What the file does not contain is the more interesting fact. No transaction by chief executive Antonio Neri, by the chief financial officer, or by any other named executive officer appears in the eight most recent filings. On a stock that has risen 164.5% in twelve months, the absence of any executive selling is mildly reassuring and the absence of any executive buying is entirely unremarkable — neither is weighted heavily, and the file is reported as uninformative.

10. Verdict, kill-criteria and flip conditions

Watch.

The business case, at full strength and filing-verified. Net revenue +40.0% to $10,678 million in the April quarter. Total segment earnings +125% to $1,526 million. Networking revenue +148% to $2,690 million at a 21.6% operating margin, against Cloud & AI's 12.4% — the mix shift the $13,625 million Juniper acquisition was bought for, arriving on schedule. Cloud & AI's own margin nearly doubled, from 6.6% to 12.4%. Free cash flow, on the company's reconciliation, from MINUS $1,724 million to PLUS $1,623 million year on year in the half. Total debt down $1,119 million to $21,246 million. Research and development +70.7%. Five consecutive earnings beats, the last at 47.7%. Consensus revenue +31.1% in FY2026. At 15.3x FY2026 and 13.0x FY2027 consensus non-GAAP EPS, none of that is expensively priced. And the knowledge base — for the first time in this batch — is unanimously positive from named speakers.

Why this is Watch and not Buy — Tactical, stated as four specific, dated, resolvable objections rather than as hedging.

First, the print is 29 days away and the bar is high. FY2026 consensus revenue of $44,969 million against a first half of $19,979 million requires roughly $24,990 million in the second half — a 25.1% sequential step. The Q3 consensus of $11,927 million implies Q4 of approximately $13,063 million against Q2's $10,678 million. That is a large, specific, imminent test, and the correct response to a large imminent test with a known date is to wait for it.

Second, the entry is the worst in this batch. +164.5% over twelve months, +82.5% over three, +4.28% on the day, 76.3% above the 200-day average of $29.72, at the 90th percentile of the annual range. Nothing about that is a reason the business is bad and everything about it is a reason the price already reflects a great deal.

Third, the earnings quality and the dilution are both real and both scheduled. The 77% gap between GAAP and non-GAAP, on $323 million a quarter of intangible amortisation and $218 million of stock compensation, against negative $3.26 of tangible book value per share and $23,770 million of goodwill on a balance sheet that took $1,361 million of impairments fifteen months ago. And 76.06 million shares — 5.75% of the count — converting automatically on 2027-09-01, for which the vendor payload's disclosure is the single character "0".

Fourth, we are 13.4% below the street and the street's own behaviour is a caution. Every one of 39 published targets is above the current price, the lowest by 18.3%, while 20 of 39 ratings are Hold. Targets raised into a move are not the same thing as conviction ahead of one.

And the counterweight, stated fairly because it is strong. The consensus is more likely too low than too high — five consecutive beats, the last at 47.7%, on a FY2027 EPS range of only 7.4%. If the 2026-09-02 print delivers, the FY2027 consensus moves up, the multiple at an unchanged price falls toward 11x, and this becomes a Buy on arithmetic that requires no re-rating at all. That is precisely why the posture is Watch rather than Avoid, and why every upgrade condition below is designed to be answered by that report.

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

• Q3 FY2026 revenue below $11.5 billion on 2026-09-02, which would make the FY2026 consensus of $44,969 million unreachable and invalidate the entire second-half ramp.
• Cloud & AI segment operating margin below 10%, reversing the 6.6%-to-12.4% improvement and signalling that AI-server pricing power has peaked.
• Networking segment operating margin below 20%, extending the decline from 25.0% a year ago and indicating that the acquired business is structurally less profitable inside HPE than outside it.
• Any goodwill or intangible impairment against the $23,770 million of goodwill and $6,368 million of intangibles. The April 2025 quarter carried $1,361 million of impairments, so this is precedent rather than speculation.
• Total debt rising back above $23 billion, reversing the six-month reduction.
• A break below the 50-day average of $46.66 on company-specific news rather than sector beta, which on a stock 76.3% above its 200-day average would have very little beneath it.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical, all resolvable on or shortly after 2026-09-02:

• Q3 FY2026 revenue above $12.5 billion with Cloud & AI organic growth above 15%, which would confirm the second-half ramp and demonstrate that the growth is not solely the Juniper anniversary.
• Networking segment operating margin above 23%, showing the integration is accretive to quality as well as to scale.
• The FY2027 consensus non-GAAP EPS moving above $4.50, which at an unchanged price is 11.6x and makes the base case work with no multiple expansion whatsoever. This is the single cleanest trigger in the list.
• A price below approximately $47 — the 50-day average — which would put the base case above 27% with the same fundamentals.
• Free cash flow above $2.5 billion for the first three quarters on the company's own reconciliation, confirming that the $3.3 billion half-year swing was structural rather than a working-capital timing effect.
• A named-speaker knowledge-base claim that distinguishes HPE from Dell — at present all three surviving claims treat it as a basket constituent.

Where HPE fits in the Synthos Framework Portfolio. No position today, with an explicit calendar trigger: revisit within one week of the 2026-09-02 print. The technology sleeve would take this name at a target weight of 2% on either of the first two upgrade conditions above, entered in halves. Sizing note, and it is the operative discipline: a stock 76.3% above its 200-day moving average four weeks before a print does not need a smaller position — it needs a later one. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $52.39, with the fair-value anchors, kill criteria and upgrade conditions all gradeable.

Single biggest risk: the second-half ramp the consensus requires. FY2026 consensus revenue of $44,969 million against a filed first half of $19,979 million implies approximately $24,990 million in the second half — a 25.1% sequential step, reported in two instalments starting 2026-09-02. The first half already grew 29.1% and the April quarter 40.0%, so the required trajectory is not absurd; it is simply not yet evidenced, and the market has already paid for it with a 164.5% twelve-month return that leaves the price 76.3% above its own one-year average. If the ramp arrives, the FY2027 consensus rises, the multiple compresses mechanically toward 11x, and the upgrade conditions above trigger. If it does not, a stock with negative tangible book value, $21.2 billion of debt, 5.75% of scheduled dilution and no technical support until $46.66 has a very long way to fall. The distribution is wide, the resolution date is known, it is twenty-nine days away, and the only thing this dive declines to do is guess.

Provenance & disclosures

• Traceability: 3 name-level knowledge-base claims name Hewlett Packard Enterprise out of 52,021 distilled claims (raw hits 87, surviving a case-sensitive entity re-run 3, entity matches 3, discarded 84; breadth 3 but independent sources 2, net conviction positive-low). All three are BULLISH and all three carry a NAMED SPEAKER, which is an unusual sourcing upgrade in this store where most claims carry speaker: null. They are: jordi_visser, 2026-06-17, conviction 65, speaker Jordi Visser, entities DELL/HPE/PSTG — "AI-factory infrastructure names — Dell, HPE, Pure Storage — are part of the buildout; Dell already reached the 'insane numbers' earnings mode."; andreas_steno, 2026-06-02, conviction 75, speaker Andreas Steno, role independent, entities DELL/HPE — "Computer/server makers have barely started: Dell's blowout AI-server numbers plus spillover to consumer products; HP Enterprise's new Nvidia production-line partnership looks compelling — Dell trading like a meme coin."; and jordi_visser, 2026-04-11, conviction 85 — the highest single conviction figure in this batch — role independent, entities NVDA/GLW/DELL/HPE — "Underinvestment vs massive AI compute/inference demand — not enough compute; semis and hardware keep rising all year, Nvidia way too cheap relative to demand; CapEx receivers' earnings keep growing." Two of the three come from the same channel and the same speaker eight weeks apart, so the independent-source count is 2, not 3, and this is stated rather than allowed to inflate breadth. Every one is a basket call naming HPE alongside DELL; not one distinguishes HPE from Dell, discusses the Juniper acquisition, mentions networking margins or names a price. And every one was written after the stock had already doubled from its 52-week low of $19.87 — a bullish claim written after a double is a description of momentum, not a discovery of value, and it is graded accordingly. The 84 discards are dominated by jensen_huang (44), whose own commentary on accelerated computing never names this company, plus eli_lilly_ceo (7), excluded on content mismatch consistent with the documented store defect in which that channel carries other companies' claims, plus 17 further channels caught by the deliberately broad free-text terms "HPC" and "supercomputing". No claim carries a management speaker_role; the two carrying independent are judged on content per house rule and are correctly classified. All quotes are verbatim from the stored claim text.
• Data as-of: fundamentals — balance sheet, segment results, income statement, cash flows, share count and the preferred-stock terms through 2026-04-30, all filing-verified from the 10-Q filed 2026-06-02; the Juniper purchase-price allocation and the mandatory convertible terms from the 10-K filed 2025-12-18; the preferred dividend declaration from the 8-K filed 2026-08-04, the day of this dive · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873599 = 2026-08-04T19:59:59Z ($52.39, +4.28%; 50-DMA $46.66; 200-DMA $29.72; RSI 62.2; MACD +1.11) · knowledge-base claims 2026-08-04. HPE's fiscal year ends 31 October; FY2026 means the year ending 2026-10-31, and every "FY" reference in this dive is on that basis. All figures come from the Synthos vendor data file for HPE or from the SEC filings in the HPE archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents: 10-K filed 2025-12-18 (fiscal 2025 — source of the Juniper purchase-price allocation, the mandatory convertible preferred terms and the acquired-award exchange ratio); 10-Q filed 2026-03-10 (January 2026 quarter); 10-Q filed 2026-06-02 (April 2026 quarter — source of the segment tables, the free-cash-flow reconciliation, the liquidity table, the balance sheet, the diluted share reconciliation and the segment-realignment disclosure); 8-K filed 2026-06-01 (Item 2.02, second-quarter results — the Exhibit 99.1 press release did not extract, so all quarterly comparisons in this dive come from the 10-Q and from earn_cal, and no management guidance figure is quoted because none was recoverable); 8-K filed 2026-07-24 (no extractable item content); 8-K filed 2026-08-04 (Item 7.01 — the board's declaration of a $0.953125 per share dividend on the 7.625% Series C Mandatory Convertible Preferred Stock, payable 2026-09-01 to holders of record 2026-08-15). The archive contains duplicate copies of every document under two filenames; they are byte-identical and were checked. All carry preserved [TABLE] statement data — 87 tables in the April 10-Q.
• Where the filings contradicted or corrected the vendor (detailed in Section 7): preferredStock: 0 against 30,000,000 shares of 7.625% Series C Mandatory Convertible Preferred Stock, $1.5 billion raised, automatically converting 2027-09-01 into a minimum of 76.06 million common shares (5.75% dilution) and costing $114.4 million a year — the omission that also explains why FY2025 eps reads MINUS $0.0446 against POSITIVE net income of $57 million; capitalExpenditure in the TTM block, implying roughly $712 million against filed half-year figures of $1,152 million and $1,075 million and a trailing figure near $2,369 million, which inflates freeCashFlowYieldTTM from a correct ~5.75% to a reported 8.13%; totalDebt of $24,077M for FY2025 against the 10-Q's stated $22,365M, the $1,712M difference being exactly the vendor's own capitalLeaseObligations line; seg_prod, which reports the five-segment structure the company abandoned on 2025-11-01 in favour of Cloud & AI / Networking / Corporate Investments and Other, restated retrospectively; est.ebitdaAvg/ebitAvg, fixed at exactly 16.23% and 7.50% of revenueAvg in every year from FY2023 to FY2029; and the peer block, which contains a payments processor, a company name attached to the wrong ticker (PSTG) and a constituent priced at $1,427.62 with a $211 billion market capitalisation. Where vendor and filing AGREED — recorded, and this name has an unusually good record: the share count is an exact match (10-Q cover 1,324,203,521 at 2026-05-26 against 1,324,204,000 implied); the enterprise value of $85,329M is within $1 million of a rebuild that uses the current filed balance sheet AND correctly adds the $61M of noncontrolling interests — the NCI omission that fired on seven of twelve names in the previous batch does not fire here; seg_geo reconciles exactly to FY2025 revenue of $34,296M with the Americas correctly identified as the largest region; the annual cf_a capex figures for FY2022-FY2025 are all approximately correct, the defect being confined to the trailing block; dividendPerShareTTM matches profile.lastDividend; and tech.max_dd_from_peak of −6.696% coincides with pct_from_hi correctly.
• Basis note — read this before using any multiple in this dive. Analyst estimates and the P/E figures in Section 5 are on non-GAAP diluted EPS, computed by the company on 1,432 million weighted-average shares that already include the 76 million-share dilutive effect of the mandatory convertible preferred. The gap between GAAP and non-GAAP was 77% in the April quarter — approximately $0.45 of GAAP basic EPS after preferred dividends against $0.79 reported non-GAAP — driven by $323 million of intangible amortisation (up from $37 million, essentially all Juniper), $218 million of stock-based compensation and $78 million of acquisition, disposition and other charges. 13.0x FY2027 on the non-GAAP basis corresponds to roughly 20-23x on a GAAP basis. Both numbers are true and the reader should hold both. The prior-year April quarter additionally carried $1,361 million of impairment charges, which is why its GAAP result was a $1,109 million operating loss and why year-on-year GAAP comparisons flatter.
• Free-cash-flow note: the company's own reconciliation, quoted verbatim in Section 2, gives half-year free cash flow of PLUS $1,623 million against MINUS $1,724 million a year earlier — a $3,347 million swing — after adding back proceeds from property sales and the effect of exchange-rate changes. On the house rule (operating cash flow less filed capital expenditure, no add-backs) the trailing-twelve-month figure is approximately $3,989 million, a 5.75% yield. The half's composition includes a $2,956 million inventory build financed by a $3,562 million increase in accounts payable, which flatters the current period and will reverse. The vendor's freeCashFlowYieldTTM of 8.13% and every derived cash-flow ratio are not used anywhere in this dive.
• Estimate coverage: 11 analysts on FY2026 non-GAAP EPS; 11 on FY2027 — the anchor for all three fair values, with a range of $3.878 to $4.167, a 7.4% spread; 14 on revenue in both years; 4 on FY2028 EPS, used only as a cross-check; 2 on FY2029, which is excluded from every conclusion.
• Peer note: the vendor peer set is discarded in full for containing a payments processor (Fiserv), a company name attached to the wrong ticker ("Everpure, Inc" at PSTG, which is Pure Storage), and a constituent priced at $1,427.62 with a $211.4 billion market capitalisation that cannot be reconciled. The genuine comparables — Dell Technologies, Cisco, Arista Networks, IBM, Lenovo — are absent entirely. No peer-multiple comparison is drawn.
• Fair-value caveat: the $40 / $60 / $80 anchors are multiples of the FY2027 consensus non-GAAP EPS distribution — 10.3x the low of $3.878, 14.9x the mean of $4.018, and 19.2x the high of $4.167 — on a consensus share count that already carries the mandatory-convertible dilution. Stated arithmetic, not a discounted cash flow. The base is sensitivity-disclosed: 13x FY2027E gives $52.23, which is spot to within 0.3%, and 17x gives $68.31. The entire base case is therefore a four-turn re-rating, and Section 5b states which way it runs and why. The bull case of $80 is exactly the highest published street target; the base of $60 is 13.4% BELOW the street consensus of $69.27, whose lowest constituent target of $62 is already 18.3% above the current price.
• Timing — and this is a verdict input, not a footnote: third-quarter fiscal 2026 results are due 2026-09-02, twenty-nine days after this dive, on consensus non-GAAP EPS of $0.921 and revenue of $11,927 million. The last print, 2026-06-01, beat consensus EPS by 47.7% and revenue by 9.2%, the fifth consecutive beat. The FY2026 consensus requires second-half revenue of approximately $24,990 million against a filed first half of $19,979 million. The most recent insider filings are dated 2026-07-22 and 2026-07-28 and contain no open-market transaction of any kind. 2026-08-04 carried one company filing — the preferred dividend declaration — which does not account for the day's 4.28% move.
• 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.