International Business Machines Corporation

IBM is a diversified computing conglomerate — Software (Red Hat OpenShift, watsonx, automation, data, and the mainframe Transaction Processing stack), Consulting, and Infrastructure (zSystems mainframes, Power, Storage). The Digital Sovereignty thesis rests on one narrow claim: IBM's mainframe + sovereign-compute stack (z/OS + CICS/IMS + quantum) is the one layer governments, banks, insurers, and airlines cannot route around.

• Revenue $17.162B; Software $7.8B (+5%, Red Hat +11%, Data +19%), Consulting $5.3B (flat), Infrastructure $3.8B (−7%, IBM Z −42%, Distributed Infrastructure +37%)
• Gross profit $9.907B = 57.7% gross margin (+1.7 pts YoY)
• Net income $2.165B
• Free cash flow $2.54B (Q2), $4.76B YTD; full-year FCF guided to increase ~$1B YoY
• RPO (backlog) $68.0B

IBM is a software / consulting / infrastructure conglomerate. The Digital Sovereignty thesis is a narrow claim: z/OS + CICS/IMS + COBOL on IBM Z is the irreplaceable settlement backbone of banks and governments, now being policy-pulled into PQC migration and a domestic quantum-wafer foundry (Anderon). Everything else (Consulting ~31% of Q2 revenue, Red Hat competing with hyperscalers) is dilution.

IBM is a software / consulting / infrastructure conglomerate. The Digital Sovereignty claim is narrow: z/OS + CICS/IMS + COBOL on IBM Z is the irreplaceable bank/gov settlement backbone, now being policy-pulled into PQC and a domestic quantum-wafer foundry (Anderon, still LOI). Everything else is dilution.

• monetization model: Subscription/SaaS, term licenses, maintenance/support, one-time hardware, consulting services, and financing. IBM 10-K FY2024, accession 0000051143-25-000009, [sec.gov
• revenue mix: FY2024 total $62,753M: Software 43.2%, Consulting 32.9%, Infrastructure 22.8%, Financing 1.2%. [10-K Note: segment revenue]
• who pays: Enterprise CIO/CTO and IT transformation budget owners; government agency/program contracting officers; some purchases financed via IBM Financing. 10-K does not disclose buyer titles. [10-K Business; Financing]
• margin_value_capture: FY2024 GAAP gross margin 56.7%; value accrues in high-margin recurring Software/maintenance, partially offset by lower-margin Consulting and mixed Infrastructure. [10-K MD&A]

• customer classes: Aerospace & Defense Primes, Critical Infrastructure Operators, Defense & Intelligence Program Offices, Government Civilian Agencies, Sovereign Cloud & Compute Buyers, plus Global 2000 enterprises, financial services, telecom, healthcare, retail. IBM 10-K FY2024, accession 0000051143-25-000009, [sec.gov

IBM reports four operating segments: Software, Consulting, Infrastructure, and Financing. For fiscal year 2024, IBM reported total revenue of approximately $62.8B and net income of approximately $6.0B

Source: IBM Form 10-K, FY ended 2024-12-31, SEC EDGAR CIK 0000051143, URL below, filing date 2025-02-25, accessed 2026-04-26, accession no. [ACC # pending live verification].

Source: IBM Form 10-K FY2024, SEC EDGAR CIK 0000051143, URL below, accession no. [ACC # pending live verification]. Revenue mix percentages are calculated as segment external revenue divided by total IBM external revenue.

IBM’s mix has shifted toward Software over time. Software is the highest-margin and most annuity-like segment; Consulting is labor-intensive and lower-margin; Infrastructure is hardware-plus-software and cyclical; Financing is small but attaches to hardware and software sales.

IBM’s GAAP total gross margin in FY2024 was approximately 57%

Source: IBM Form 10-K FY2024, SEC EDGAR CIK 0000051143, URL below, accession no. [ACC # pending live verification].

Segment gross margins are not identical to total gross margin. Derived from IBM’s segment gross-profit disclosure:

These segment margins are derived, not management-reported GAAP net margins. They should be recalculated from the segment table in IBM’s Form 10-K. The wide gap between Software and Consulting/Infrastructure explains why IBM’s strategy is to grow Software and de-emphasize commodity infrastructure and low-margin consulting work.

IBM’s monetization is not a single stream. It is best understood as four linked engines:

1. Software subscriptions/support and licenses. Red Hat, IBM software, transaction processing, and SaaS are sold as subscription, support, or term-license agreements. These generate recurring revenue and have high renewal rates.
2. Consulting services. IBM Consulting earns revenue from strategy, systems integration, application modernization, and managed services. Contracts are typically fixed-fee, time-and-materials, or outcome-based.
3. Infrastructure systems and software. IBM sells IBM Z and LinuxONE mainframes, IBM Power servers, storage systems, and attached operating systems and middleware. Hardware gross margins are lower than software; the real economic value is the long lifecycle of z/OS software, support, and financing attachment.
4. Financing. IBM Global Financing monetizes hardware and software transactions through leases and loans, producing interest income and financing-spread economics.

The paying entities are overwhelmingly large enterprises and public-sector organizations: financial institutions, insurers, healthcare systems, telecommunications operators, government agencies, retailers, and industrial firms.

IBM’s Form 10-K does not name a top customer list. It states, in substance, that no individual customer accounted for more than 10% of total revenue in FY2024. Therefore, no investor-grade “top customer” names can be cited from the primary filing without inventing data.

Source: IBM Form 10-K FY2024, business/risk section, SEC EDGAR CIK 0000051143, accession no. [ACC # pending live verification].

IBM’s customer classes are not detailed by name in the Form 10-K. They are described by industry and buying pattern.

• Financial services: banks, insurers, capital-markets firms running core banking, payments, fraud, and compliance workloads on IBM Z and Db2.
• Public sector: federal, state, and local government agencies with security, sovereignty, and long procurement-cycle requirements.
• Healthcare and life sciences: payers, providers, and pharmaceutical companies using proprietary transaction systems and regulated data environments.
• Telecommunications/utilities: billing, network operations, and customer-care systems.
• Retail/CPG: supply chain, inventory, e-commerce, and customer analytics.
• Manufacturing/industrial: production planning, logistics, and asset-intensive operations.

• CIO/CTO buying hybrid-cloud platform and mainframe modernization.
• COO buying consulting-led operational transformation.
• CFO/procurement controlling large multiyear IT spend and financing.
• Government program owner buying sovereign/regulated IT services.

No top customers are named in the Form 10-K. IBM states that no single customer accounted for more than 10% of revenue. For this report, any named “top customers” would be fabricated; therefore none are asserted.

Source: IBM Form 10-K FY2024, SEC EDGAR CIK 0000051143, accession no. [ACC # pending live verification].

IBM operates as four externally reported segments: Software, Consulting, Infrastructure, and Financing. In FY2024, IBM reported total revenue of approximately $62.8 billion, net income from continuing operations of approximately $6.0 billion, diluted EPS from continuing operations of $6.44, and free cash flow of approximately $12.7 billion ([S1], IBM 10-K FY2024, cover/MD&A/cash-flow statement).

From the FY2024 segment reporting in IBM’s 10-K, the external revenue mix is approximately:

These are rounded figures from IBM’s segment note; exact reported line items are in the FY2024 10-K segment results ([S1], IBM 10-K FY2024, Segment Information note/Item 7 MD&A, filed 2025-02-25, accessed 2026-05-07). The business has become software-led: Software was the only large segment with meaningful constant-currency growth in FY2024, up roughly 9%, while Consulting declined modestly and Infrastructure declined low-to-mid single digits ([S1], MD&A; [S3], IBM Q4 2024 8-K Exhibit 99.1).

IBM monetizes through a mix of:

• Recurring software subscriptions and support: Red Hat Enterprise Linux, OpenShift, IBM software-as-a-service, transaction-processing software, and term licenses.
• Consulting and professional services: Project-based system integration, application modernization, managed services, and strategic consulting.
• Mission-critical hardware systems: IBM Z and LinuxONE mainframes, Power systems, storage, and maintenance.
• Captive financing: IBM’s financing segment originates leases and loans for IBM and third-party equipment ([S1], IBM 10-K FY2024, Item 1 Business, Financing segment).

The customer pays directly through enterprise IT budgets, government procurement, and large outsourcing/managed-services contracts. IBM’s 10-K states that no single client represented 10% or more of total revenue in FY2024 ([S1], IBM 10-K FY2024, Revenue Recognition/Customer Concentration). This is important for conviction: IBM is not dependent on one named customer, but it is heavily dependent on a few large enterprise/government buyer classes.

IBM reported consolidated gross profit of approximately $35.5B on FY2024 revenue of about $62.8B, implying a GAAP gross margin of roughly 56.6% ([S1], IBM 10-K FY2024, income statement; [S3], Q4 2024 8-K). GAAP income from operations was approximately $10.4B, or about 16–17% operating margin; this is a derived figure from the FY2024 income statement and should be confirmed against exact line items on EDGAR ([S1], IBM 10-K FY2024, Consolidated Statement of Earnings).

The margin mix matters more than the consolidated margin: IBM does not disclose segment gross margin, but the Software segment is structurally higher-margin and capital-light. IBM’s software-heavy mix shift is the core of the current investment thesis.

IBM’s customer classes are large enterprises, governments, regulated institutions, and mid-market enterprises. The 10-K does not provide a ranked list of named end customers. It states that no single customer was 10% or more of revenue in FY2024 ([S1], IBM 10-K FY2024, Customer Concentration). It also describes revenue by geography: Americas, EMEA, Asia Pacific ([S1], IBM 10-K FY2024, Segment and Geographic Information).

The practical buyer archetypes from IBM’s filing descriptions are:

• CIO/CTO of a large bank or insurer: buys Red Hat, IBM Z, Db2, security, and consulting modernization.
• Government agency procurement officer: buys mainframe capacity, software licensing, managed application services, and consulting through long-term contracts.
• Large telecom or manufacturer: buys hybrid-cloud platform software, storage, and IBM Consulting application operations.
• Hyperscaler/enterprise ecosystem partners: IBM software is distributed through and integrated with AWS, Microsoft Azure, and Google Cloud; these are simultaneously customers, partners, and competitors.

IBM does not name a top beneficiary customer in the FY2024 10-K; any list of “IBM’s top customers” circulating in secondary sources is not authoritative unless it traces to a specific contract filing. Do not import unverifiable customer logos into a conviction model without a primary filing.

Digital Sovereignty > Sovereign Cloud & Compute > hybrid/gov cloud + clearance-gated software as the 10-year contract toll

Next earnings (last reported 2026-07-22); federal/hybrid cloud bookings in next 10-Q

1. Switching-cost floor. Per-bank/gov core migration = 5–10 years (10–15 for major bank cores) and $100M–$1B+. Across the installed base this is a multi-hundred-billion-dollar collective switching cost. The February 2026 Anthropic Claude Code COBOL-modernization announcement (auto-mapping/documentation of COBOL) wiped $31B from IBM's market cap in one day (−13.2%, worst since dot-com) — the market's own estimate of the AI threat to the maintenance/consulting edge of this moat. AI accelerates code understanding, not Tier-1 core replacement (verification + regulatory burden remain), so the toll is dented, not broken.
2. Conglomerate dilution. Infrastructure $3.8B ÷ $17.162B total = 22.1% of revenue — and that segment includes Power/Storage, with IBM Z hardware itself down 42% YoY (product-cycle trough between z16/z17). The pure mainframe toll (hardware + Transaction Processing software) is ~10–15% of revenue; the full sovereign stack (mainframe + quantum + gov hybrid cloud) is ~20–25%. The toll-road is real but diluted by a consulting body-shop and a hybrid-cloud line that competes head-on with hyperscalers.
3. Quantum cost-of-moat. $10B (5-yr) ÷ the PQC-mandated migration market (all federal + critical-infrastructure, deadline 2030/2031) = the price IBM pays to defend a sovereign position that policy is now forcing buyers into. The $1B gov equity stake is revealed-preference: the US government is underwriting IBM as a domestic-quantum instrument of record.

1. IBM Z dollar mix (the 08-29 note never split Hybrid Infrastructure).

Hybrid Infrastructure Q2 = $2,570M vs $2,866M. Identity:

Z_26 + D_26 = 2,570 and Z_26 / 0.58 + D_26 / 1.373 = 2,866

Solving: Z_26 ≈ $998M, D_26 ≈ $1,572M. Cross-check: $998 / $1,721 ≈ −42.0%; $1,572 / $1,145 ≈ +37.3%.

So the “irreplaceable mainframe” is 5.8% of Q2 revenue. Add Transaction Processing ($2,030M, −8.1%): the classic Z + TP stack is 17.6% of Q2. Add Red Hat Hybrid Cloud ($1,998M): the broad sovereign/hybrid stack is 29.3%. Consulting alone is 31.0% — larger than the entire sovereign stack.

2. Switching-cost floor vs capture.

RPO $68B ÷ FY2025 $67.535B = 1.01× annual revenue locked. Deferred income $20.4B ÷ Q2-annualized $68.6B = 29.8% of a year’s revenue already billed/unearned. That is the contract lock. It is not 100% mainframe: ~80% of Software is recurring ARR ($24.6B), and only ~20% of Software is the Z-tied ELA/transactional slice the 10-Q itself flags as the Q2 miss.

3. Cost of the moat vs moat defended (quantum + Lightwell).

• Quantum: >$10B / 5 years ≈ $2.0B/yr. Q2 FCF $2.54B → one year of the quantum spend is ~79% of one quarter’s FCF, or ~19% of a ~$10.5B FY FCF run-rate if the +$1B FCF guide lands on a ~$9.5B 2025 base.
• Lightwell: $5B commitment to defend the open-source/Red Hat trust layer — a hyperscaler-contested moat, not the mainframe toll.
• Combined announced “defend the franchise” spend (quantum $10B + Lightwell $5B) = $15B against a Z+TP stack that annualizes at ~$3.0B × 4 ≈ $12.1B. IBM is spending more to extend (quantum/PQC/open-source) than the core toll currently prints. That is either (a) a cheap option on a policy-forced PQC/quantum market, or (b) conglomerate vanity. The 10-Q does not break quantum revenue — treat (a) as the bull case, unproven.

4. z17 cycle is timing, not demand destruction (the disconfirming-flip test).

Krishna’s 07-14 letter: z17 is “the strongest start to a mainframe program in our history,” ~130% program-to-program vs z16, and 85% of installed MIPs are maintaining or growing capacity. The Q2 miss is explicitly “clients redirected spending toward servers, storage, and memory” in the last weeks of June — the same Dynamic that printed Distributed Infrastructure +37% and a $500M Power/Storage backlog. If Q3/Q4 Z + TP do not re-accelerate, the 130% PTP claim is the thing that breaks.

Sensitivity

Toll-road (DS — Sovereign Cloud & Compute): bypass 3 / concentration 2 / hyperscaler 2 = 7 STRONG PROXY (unchanged composite; the 08-29 3/2/2 stands).

• Bypass 3: no substitute for CICS/IMS + z/OS at bank/gov scale inside 18 months; 85% of MIPs growing on z17.
• Concentration 2: IBM is still the sole remaining volume mainframe vendor, but the company is not a ≥30% supplier of a non-deferrable input at the consolidated P&L — Z is 5.8% of revenue. Doctrine conglomerate-dilution haircut applies.
• Hyperscaler 2: AWS/Azure/GCP have not displaced Z in 20 years (PASS on the mainframe slice) but Red Hat/OpenShift is a contested hybrid-cloud line. Flag separately: clean 9 applies to Z/TP only.

Cheat-sheet check: DS clearance-gated software 8–9, gov IT body-shops 3–4. IBM is a blend — mainframe = 9, consulting = 3–4, hybrid cloud = 5–6 → blended 7 is inside the published range.

1. Federal dollars are not the mainframe toll. Trailing-year USAspending obligations of $1.1B are 1.63% of FY2025 revenue. Even if every federal dollar were “sovereign compute,” it is an order of magnitude smaller than the Q2 Z+TP stack the 09-05 note deconvolved at 17.6% of the quarter. The DS thesis is the installed CICS/IMS/z/OS base at banks and governments, not IBM’s federal-services book.

2. Inside the federal book, the government buys IBM labor, not IBM Z iron.

• PSC DA01 (business-application / app-dev labor) = $325.4M = 30.0% of the $1.1B bucket.
• PSC DH01 (platform support: database, mainframe, middleware labor) = $143.2M = 13.2% of the bucket = 0.21% of FY2025 revenue.
• PSC 7B21 (compute: mainframe hardware + perpetual license) = $15.1M = 1.4% of the bucket = 0.022% of FY2025 revenue.

So the US government is a body-shop + middleware-labor customer of IBM at the federal-award layer, and almost not a Z-hardware customer. That is the same conglomerate-dilution haircut the 09-05 note applied to the P&L, now visible in procurement codes. Cheat-sheet check: DS clearance-gated software 8–9, gov IT body-shops 3–4; IBM’s federal mix is the 3–4 bucket, while the mainframe installed base remains the 8–9 claim.

3. USCIS FALCON is a real identity-backbone award, and it is still services. PIID 70SBUR24F00000103, $157.8M, DevSecOps for USCIS VIS (Verification Information System), signed 2024-05-16, currently running through 2027-09-10. $157.8M ÷ FY2025 $67,535M = 0.23% of annual revenue. It is a named sovereign-identity contract and a relationship edge; it does not move economic_capture.

4. Cost of the sovereign quantum option vs a single System Two placement.

• Anderon (unclosed): $1B IBM cash + $1B CHIPS LOI = $2B headline against a Z+TP stack that annualizes ~$12.1B (09-05: ~$3.028B × 4). IBM cash alone is 8.3% of one year’s Z+TP. That is a cheap option on a policy-forced wafer layer if definitive docs close; until they close it is not a foundry.
• Swiss System Two: one Nighthawk-class machine, IBM-operated, 3-year ETH term through 2029, in-service targeted YE2026, funded by a defense-procurement offset (LMT–armasuisse). Offset dollars are not disclosed → cannot compute $/qubit or revenue. What it does prove: a NATO-adjacent defense buyer chose IBM hardware as the in-country sovereign quantum footprint rather than a hyperscaler cloud SKU. That is a Q3-hyperscaler PASS on the quantum-hardware slice, same as the mainframe slice — and still immaterial to the consolidated P&L.

5. COBOL Elevate as anti-bypass (the Feb-2026 −13% cap-wipe, inverted). The product’s stated job is to get near-COBOL-6 performance without rewrites and to automate the compiler upgrade on z/OS. If GA on 2026-09-18 attaches, it raises the switching-cost floor (stay on Z, get the AI tooling). If the market instead treats Elevate as a COBOL-understanding wedge that later ports off-Z, technical_scarcity −2. No attach-rate → no capture upgrade this pass.

Sensitivity

Toll-road (DS — Sovereign Cloud & Compute): bypass 3 / concentration 2 / hyperscaler 2 = 7 STRONG PROXY (unchanged).

• Bypass 3: no 18-month substitute for CICS/IMS + z/OS; COBOL Elevate is explicitly a no-rewrite stay-on-Z tool.
• Concentration 2: sole remaining volume mainframe vendor, but Z is 5.8% of Q2 and federal 7B21 is 0.022% of FY revenue — conglomerate-dilution haircut still applies.
• Hyperscaler 2: AWS/Azure/GCP have not displaced Z; Swiss System Two is IBM-operated hardware in a national lab (PASS on that slice). Red Hat/OpenShift remains contested.

Cheat-sheet: DS clearance-gated software 8–9 blended with gov-IT body-shop 3–4 → 7 is inside the published range.

\[

\$26.7B \times 0.815 = \$21.8B

\]

\[

\$62.8B / \$1.1T = 5.7\%

\]