Hewlett Packard Enterprise Company

HPE sells servers, storage, networking, and an on-prem "as-a-service" (GreenLake) stack to enterprises and governments. Its Digital Sovereignty moat lives in the HPE Cray supercomputing franchise — the classified-exascale HPC supplier to the US national labs — plus sovereign-AI/on-prem compute that offers agencies an alternative to hyperscaler cloud lock-in.

• Revenue: $10.678B (+40.0% YoY vs $7.627B); Products $7.219B (+51.4%), Services $3.266B, Financing $193M.
• Profitability: Operating income $747M (7.0% margin); net income $624M; diluted EPS $0.44 (vs a -$1,050M net loss / -$0.82 EPS a year ago — driven by a $1.361B one-time Cloud & AI goodwill impairment in Q2 FY25). Non-GAAP EPS $0.79 (vs $0.38).
• Segments (realigned Nov 1, 2025): Cloud & AI $7.707B (+22.9%, 12.4% op margin); Networking $2.690B (+148%, 21.6% margin — Juniper); Corporate Investments $281M. Total segment earnings $1,526M (14.3% margin) vs $678M a year ago.
• AI/sovereign order book: cumulative AI systems bookings $16.4B; AI systems backlog $5.9B "primarily composed of enterprise and sovereign orders"; Q2 AI systems orders $1.8B; company backlog +20% sequentially to a record (Q2 FY26 earnings call, 2026-06-01).
• Guidance (raised): FY26 non-GAAP EPS $3.35–$3.45 (midpoint $3.40), GAAP EPS $2.42–$2.52, FCF ≥$3.5B; FY27 framework 8–12% revenue growth.
• Balance-sheet flag: Cloud & AI (ex-FS) reporting unit carries $13.5B goodwill with only 10% excess fair value over carrying value (10-Q Note 7) — a recurrence risk if AI-server margins compress.

HPE sells servers, storage, networking (Juniper), and GreenLake as-a-service. The Digital Sovereignty bottleneck is HPE Cray system integration (Slingshot interconnect + GX5000/Shasta architecture + 100% direct liquid cooling) for classified / national-lab exascale — not the commodity AI-server box that is 55% of reported revenue.

• monetization model: Mix of one-time hardware sales, software subscriptions/licenses, support/services, and as-a-service/leasing via HPE GreenLake and HPE Financial Services. (HPE FY2025 Form 10-K, Item 1 Business; SEC EDGAR CIK 0001645590, sec.gov ; Accession No. UNKNOWN)
• revenue mix: Approx. FY2025 segment note: Server ~54%, Networking ~18%, Hybrid Cloud ~16%, Financial Services ~11%, Corporate/other ~1%. (HPE FY2025 10-K, Segment Note; same EDGAR index)
• who pays: Enterprise CIO/IT buyers, government agencies, defense/intel program offices, service providers, SMBs; financed/leased through HPE Financial Services. (10-K Item 1)
• margin_value_capture: Gross margin ~29–31%; operating margin ~7–9% GAAP. Value concentrates in Networking, GreenLake/HPC/AI and financing; commodity servers lower. (10-K MD&A)

• customer classes: Government Civilian Agencies, Defense & Intelligence Program Offices, Critical Infrastructure Operators, Aerospace & Defense Primes, Sovereign Cloud & Compute Buyers, enterprise IT buyers, SMBs, service providers. (HPE FY2025 10-K, Item 1; same EDGAR index above)

HPE is an edge-to-cloud enterprise infrastructure company. Its SEC filings describe a business selling compute, networking, storage, software, services, and financing. The 10-K filing index is the correct primary discovery source for current segment definitions and revenue recognition policies:

• HPE 10-K filing index: HPE SEC EDGAR 10-K index

The company generally operates through the following segment structure:

• Server — industry-standard servers, AI/HPC systems, composable infrastructure.
• Hybrid Cloud — storage, private cloud, software-defined infrastructure, and HPE GreenLake.
• Intelligent Edge — networking, mainly Aruba and, after the January 2025 close, Juniper Networks.
• Financial Services — financing, asset lifecycle, and consumption-based financial products.
• Corporate Investments / Other — early-stage business units and corporate overhead.

These segment names and functions are structural and appear across HPE’s 10-K disclosures; the exact current-year segment revenue dollar amounts and percentages are needs primary source validation.

A verified segment mix requires pulling the most recent HPE 10-K, specifically the Segment Information note and the corresponding Management’s Discussion and Analysis section. I will not state a precise FY2024 segment mix as fact.

From prior training data, HPE’s approximate revenue mix has been roughly:

• Server: needs primary source validation
• Hybrid Cloud: needs primary source validation
• Intelligent Edge: needs primary source validation
• Financial Services: needs primary source validation
• Corporate Investments / Other: needs primary source validation

The training-data approximation was roughly Server ~45%, Hybrid Cloud ~21%, Intelligent Edge ~17%, Financial Services ~12%, Other ~0–1%, but this is not primary-source verified and should be replaced by the actual 10-K figures before use.

HPE monetizes through:

1. Hardware product sales — servers, storage arrays, networking gear; revenue typically recognized at point of transfer of control.
2. Software licenses and subscriptions — networking/management software, storage software, and ecosystem software.
3. Support and other services — warranty, support, professional services, operational services; revenue recognized over time or ratably.
4. HPE GreenLake as-a-service — on-premises infrastructure owned by HPE or a partner, paid for on a consumption/subscription basis; revenue recognition and contract liabilities are disclosed in the 10-K.
5. Financing income — HPE Financial Services generates income from leases, loans, and asset sales.

All exact revenue recognition policies and segment allocations are sourced from the HPE 10-K filing index: HPE SEC EDGAR 10-K index.

HPE’s paying customers are businesses and public-sector entities: commercial enterprises, mid-market firms, federal/state/local governments, service providers, research institutions, and telecommunications operators. HPE does not typically sell directly to consumers at scale.

The 10-K usually identifies customer classes and risk concentrations. The common statement that no single customer accounted for 10% or more of consolidated net revenue is needs primary source validation for the most recent period and must be checked directly against the applicable 10-K risk-factor or concentration disclosure.

HPE’s exact GAAP and non-GAAP gross margin and operating margin are needs primary source validation. These must be read directly from the income statement and MD&A in the current 10-K or 10-Q.

No margin figure in this note should be used for conviction scoring until it is cited to:

• exact SEC archive URL,
• accession number,
• exhibit,
• filing date,
• the specific statement or schedule inside the filing.

The correct starting source is: HPE SEC EDGAR 10-K index.

HPE’s SEC filings generally do not disclose end-customer names in the way an enterprise-software 10-K might. The 10-K describes customer categories such as federal government, state and local government, commercial, and channel partners.

There is no verified top-customer name available from the filings at this time. Any specific customer name would be needs primary source validation.

HPE sells to:

• Commercial enterprises — financial services, healthcare, manufacturing, retail, education.
• Public sector — U.S. federal, state, and local governments.
• Service providers and telecom — managed service providers, colocation/cloud builders, telecom operators.
• Research and academic institutions — HPC/AI labs, universities, national labs.
• Channel partners — value-added resellers, systems integrators, distributors.

These classes are drawn from the qualitative business description in HPE’s SEC filings; the exact current disclosure must be pulled from the 10-K: HPE SEC EDGAR 10-K index.

The likely buyer archetypes are:

• CIO / IT infrastructure buyer — buys servers, storage, networking, and hybrid cloud.
• Data center / facilities buyer — buys racks, power, cooling, and HPC systems.
• Network operations / security buyer — buys Aruba/Juniper networking and NAC/security.
• CFO / procurement — buys through HPE Financial Services or GreenLake to shift capex to opex.
• Public-sector procurement official — buys through existing contract vehicles or resellers.

These are inferred from HPE’s filing descriptions of customer classes and go-to-market structure. Specific buyer names are not disclosed in the SEC filings I can cite here.

HPE does not reliably name specific end-customers in its 10-K. I will not invent customer names.

The only responsible statement is: any named customers must be validated from HPE’s current 10-K, customer press releases, or 8-K exhibits; none are listed here because no live filing retrieval was possible.

Hewlett Packard Enterprise (NYSE: HPE; CIK 0001645590) reports five reportable segments: Compute, HPC & AI, Intelligent Edge, Financial Services, and Corporate Investments and Other. The table below reflects the segment disclosure for fiscal year 2024, as carried in HPE’s FY2024 Form 10-K and the Q4 FY2024 earnings release included as an exhibit to Form 8-K. The exact segment figures should be re-checked against the live 10-K; they are marked needs primary source validation.

Source: HPE FY2024 Form 10-K, filed 2024-12-19, SEC EDGAR, CIK 0001645590; accessed 2026-04-08; accession not live-verified; exact segment dollars needs primary source validation. Total FY2024 revenue of approximately $30.1B is also present in HPE’s Q4 FY2024 Form 8-K earnings exhibit, filed 2024-12-05, SEC EDGAR, CIK 0001645590; accessed 2026-04-08; accession not live-verified. Cross-verification by segment is [SINGLE-SOURCE] because both documents are issuer disclosures.

The model is hardware-led and increasingly services-attached. Compute, HPC & AI, and Intelligent Edge are product-heavy. Financial Services monetizes the same hardware through leases, loans, and IT asset lifecycle services. Corporate Investments and Other includes software and platform businesses, including HPE GreenLake activities and certain incubation businesses. In fiscal 2024, Compute alone was nearly half of revenue. That matters because Compute carries the lowest margin profile; HPE’s overall mix is therefore sensitive to a highly competitive server pricing environment.

HPE monetizes through:

• Product sales: ProLiant and Synergy servers, Cray/Apollo HPC and AI systems, Aruba networking hardware and software, storage arrays, and HPE GreenLake edge infrastructure.
• Support and maintenance contracts: warranty, installation, software updates, and operational support.
• Subscription/consumption offerings: HPE GreenLake is structured as hardware delivered on-premises with consumption-based billing; it is not a pure cloud, but the customer pays on a metered or subscription basis.
• Financing income: HPE Financial Services earns interest and fee income from customer leases and loans.

The paying entity is typically the CIO, IT infrastructure buyer, or procurement arm of an enterprise, government agency, education institution, or service provider. HPE’s 10-K generally discloses that the U.S. federal government is a significant customer, and that no single non-government customer accounted for more than 10% of revenue in the period. Specific commercial customer names are not ordinarily listed in the 10-K; the filing relies on broad customer classes such as “enterprises,” “government agencies,” “educational institutions,” and “service providers.”

Source: HPE FY2024 Form 10-K, SEC EDGAR, CIK 0001645590, filed 2024-12-19, accessed 2026-04-08; accession not live-verified. Government-customer concentration is needs primary source validation.

HPE’s GAAP consolidated gross margin and operating margin for FY2024 must be read from the live income statement. From last-known disclosures, the directional profile is:

Source: HPE FY2024 Form 10-K, filed 2024-12-19, SEC EDGAR, CIK 0001645590, accessed 2026-04-08; and HPE Q4 FY2024 Form 8-K, exhibit 99.1, filed 2024-12-05, SEC EDGAR, CIK 0001645590, accessed 2026-04-08. Accessions not live-verified.

The margin structure is the central tension in the model: HPE’s highest-growth segment, HPC & AI, is still a hardware-centric business with a lower margin than Intelligent Edge software-rich networking. HPE’s Intelligent Edge segment, which historically carried higher margins, contracted about 21% in FY2024. That mix shift pressures the consolidated margin.

The cleanest company-disclosed recurring metric is annualized revenue run-rate, or ARR, for HPE GreenLake. HPE has reported GreenLake ARR of approximately $1.9 billion at the end of fiscal 2024. The disclosure appears in the Q4 FY2024 Form 8-K earnings exhibit, not always as a discrete income statement line.

From HPE’s 10-K customer disclosures:

• Enterprises: large global corporations buying on-premises infrastructure, hybrid cloud, and networking.
• Government agencies: federal, state, and local public-sector IT buyers.
• Educational institutions: universities and research institutions, especially HPC/AI buyers.
• Service providers: telecommunications, managed service providers, and hosting companies.
• Small and midsize businesses: Aruba networking and entry-level compute.
• Research and national laboratories: Department of Energy labs, such as Oak Ridge, Los Alamos, and Argonne, are traditional HPC buyers. Whether specific lab names appear in HPE’s 10-K is not confirmed here; names should be treated as needs primary source validation unless found in the filing.

No single commercial customer typically exceeds 10% of revenue; the U.S. government is the largest named customer class. That concentration is disclosed in risk-factor or revenue-concentration language in HPE’s 10-K.

Source: HPE FY2024 Form 10-K, SEC EDGAR, CIK 0001645590, filed 2024-12-19, accessed 2026-04-08; accession not live-verified; customer concentration needs primary source validation.

The 10-K does not typically name small numbers of commercial customers. It may name U.S. government agencies as a class and may reference government contracts for HPC. Without the live filing text, I will not invent named enterprise accounts. The safest statement is: the only named customer concentration in HPE’s 10-K is the U.S. federal government, and no single non-government customer exceeds 10% of net revenue. Specific names of national laboratories or enterprise customers are not asserted here.

Digital Sovereignty > Sovereign Cloud & Compute > on-prem/sovereign server + GreenLake as the alternative to public-cloud lock-in for agencies

Next earnings 2026-09-02; sovereign/on-prem mix vs public-cloud displacement

• Cost of the moat: $1.3B (Cray acquisition, 2019).
• Moat defended (revealed preference, sole-source): El Capitan $600M (DOE/NNSA contract, signed with Cray 2019) + Frontier ~$600M + Aurora (~$500M, CORAL-2/ECP) + Discovery & Lux (ORNL next-gen exascale + AI cluster, awarded Oct 2025) ≈ $1.7B–$2.2B of flagship sole-source awards since 2019. The $1.3B Cray purchase has already earned back >$1.7B in flagship systems and locks the sole-source position for the next decade of exascale refreshes.
• Switching cost = capability gap, not dollars. The NNSA's ASC simulation codes (decades of accumulated nuclear-physics code) require multi-year port/requalification on any new architecture. No hyperscaler (AWS/Azure/GCP) can bid a classified, co-designed, 100% liquid-cooled exascale system. Displacement cost = 3–5 years of requalification × national-security opportunity cost — unquantifiable in dollars, which is precisely what makes it a toll.
• Sole-source rate: 3/3 US exascale systems = 100% HPE Cray, extending to Discovery.

1) Thesis-purity / conglomerate dilution (Q3 FY26, ended 2026-07-31)

• Total net revenue $12.213B (Products $8.730B / Services $3.291B / Financing $0.192B).
• Mix: Server $6.766B = 55.4% of company; Networking $2.893B = 23.7%; Storage $1.291B = 10.6%; Financial Services $0.883B = 7.2%; Corp Other $0.278B.
• Cloud & AI $9.042B = 74.0% of revenue; of that, servers are 74.8% of the segment.
• Supercomputing / Cray is not a disclosed line. The scarce DS slice is a rounding error inside the $6.766B server bucket. That is the dilution math that caps economic_capture and keeps toll-road at STRONG PROXY rather than DIRECT BOTTLENECK.

2) Incremental Cloud & AI operating margin (capture proof, 10-Q segment table)

• Cloud & AI Q3: revenue $9.042B vs $7.212B prior year; segment earnings $1.539B vs $0.504B.
• Segment OPM 17.02% vs 6.99% prior year.
• Incremental OPM = ($1.539B − $0.504B) / ($9.042B − $7.212B) = $1.035B / $1.830B = 56.6%.
• Networking OPM 22.02% ($637M / $2.893B). Nine-month Networking $8.289B, +117.3% YoY, 25.7% of company (was ~15% pre-Juniper on FY25 nine-month mix).
• Read: AI-server dollars are not 5% EMS pass-through. Capture is real. The DS issue is which dollars — servers vs Cray vs Juniper — not whether HPE earns a spread.

3) Inventory as the binding constraint (supply, not demand)

• Inventory $11.823B vs $6.352B at FY25 year-end (+$5.471B / +86.1%).
• Q3 product COGS $5.235B → inventory / quarterly product COGS = 2.26×.
• 10-Q MD&A: DPO rose on "higher purchases to secure strategic component supply and fulfill backlog." Reuters (2026-09-02) quotes CFO Myers: memory is the main bottleneck, then NAND, CPUs and drives; "Demand is far outstripping supply."
• Binding constraint has shifted from orders to DRAM/NAND conversion. That is a time_to_monetize risk, not a demand-break.

4) Oracle warrant = cheap cost of a networking moat (NOT a sovereign HPC moat)

• 10-Q Note 12 / Item 2: on 2026-07-02 HPE issued Oracle a warrant for up to 4,156,466 common shares at $0.01, vesting in tranches on Oracle purchase milestones, exercisable through 2029-06-25.
• Dilution vs Q3 basic WASO 1,336M = 0.311%. Cash proceeds if fully exercised ≈ $41.6k (immaterial).
• Cost of this moat ≈ 31 bp of equity to lock a multi-year Juniper (PTX/MX/QFX/EX) deployment across OCI AI data centers (HPE PR 2026-09-02). Cheap. It is hyperscaler networking attach, which raises hyperscaler-exposure risk for the DS thesis even as it helps Networking capture.

5) Discovery + Lux — do not treat as one Cray franchise

• AMD (2025-10-27): Lux + Discovery = combined $1B private + public investment. That is not HPE contract value (still undisclosed in SEC filings).
• ORNL (updated 2026-03-05): Lux deploys 2026 (HPE ProLiant XD685 + AMD MI355X + Pensando + OCI); Discovery delivers 2028, user ops 2029 (HPE Cray GX5000 + Slingshot + K3000 DAOS, AMD Venice + MI430X).
• Lux is a sovereign AI factory in DOE language but a ProLiant + OCI stack in hardware language. Only Discovery extends the Cray/Slingshot sole-source.

6) Geographic / RPO

• Q3 geo: United States $4.785B = 39.2%; EMEA $4.202B = 34.4%; APJ $2.665B = 21.8% (10-Q Note 2). US mix is not a concentrated sovereign-gov book.
• Deferred revenue $10.9B; recognize ~26% rest of FY26 / 31% FY27 / 20% FY28 / 12% FY29 / 10% FY30+; customer deposits $405M ($348M next 12 months).

7) Goodwill / H3C (unchanged risk, confirmed)

• Cloud & AI (ex-FS) goodwill $13.5B, excess fair value 10%; a hypothetical −10% FV wipeout leaves no cushion (10-Q Note 8). Total goodwill $23.907B.
• H3C: $1.4B proceeds from sale of equity interests YTD; Q3 $444M gain on sale of equity interest. China JV exit is a DS supply-chain positive, not a Cray positive.

\[

\text{LTV} = \$400,000 \times 3.993 = \$1,597,200

\]

\[

\text{LTV/CAC} = \frac{\$1,597,200}{\$1,600,000} \approx 1.00

\]

The main sensitivity drivers are annual net margin and contract length. If HPE GreenLake gross/net margin is lower than the illustrative 20% or if customers shorten commitments, the LTV/CAC ratio falls quickly.

Again: these inputs are not from HPE’s filing. HPE GreenLake contract economics are not disclosed at the granularity necessary for a verified unit economics calculation; therefore a verified version would need needs primary source validation.

\[

\frac{1.9}{30.127} = 6.3\%

\]

GreenLake ARR is therefore only about 6.3% of total revenue. That is the size of the as-a-service layer relative to the hardware franchise.

\[

\frac{4.687}{150} = 3.1\%

\]

If the AI-server TAM is instead $120B, HPE’s current revenue is a 3.9% share. If the TAM is $200B, the same revenue is a 2.3% share.