International Business Machines Corporation (NYSE: IBM) — The Price Reset Arrived Before the Proof

Prepared as: Independent investment research Report date: 2026-08-10 Ticker: NYSE: IBM | Sector: Information Technology — Software & Services (GICS: IT Consulting & Other Services) Price (ref.): $237.28 (August 7 close) | Shares outstanding: 942.1M | Market cap: $223.5B | Core EV: ~$264.3B TTM / FY2026 guide: Revenue $69.1B · GAAP EPS $11.31 · IBM-defined FCF ~$14.65B TTM · FY2026 FCF ~$15.7B

⚡ Claude’s Take

The author’s subjective opinion; general information, not investment advice. The analytical body below carries no recommendation. Only this block expresses a view or valuation zone.

Verdict: HOLD / watch for proof. The 29% fall from the June peak repaired the price, but Q2 damaged the evidence. Do not automatically buy the old $230–235 entry range; require either Software organic growth back above 5% or a wider margin of safety around $210–220. Not a short. Conviction: medium.

IBM is still a better company than the pre-Kyndryl conglomerate. It has a genuine mainframe and Transaction Processing captivity moat, a credible Red Hat hybrid-cloud platform, $24.6B of Software ARR, a 58% consolidated gross margin structure, and a full-year free-cash-flow guide near $15.7B. At $237.28 the equity offers roughly a 7.0% yield on that guide and trades at 14.2x guided FCF, versus 17x in June. The valuation reset is meaningful.

The problem is that the operating evidence deteriorated at exactly the point the June thesis said it must not. Q2 revenue grew only 1%; Software grew 5% reported but was flat organically; Transaction Processing fell 8%; management cut constant-currency revenue guidance to 4–5% from at least 5% and Software guidance to 6–8% from roughly 10%. Clients shifted late-quarter capital budgets toward scarce servers, storage and memory, and IBM says 30–40% of delayed software deals have since closed. That may be timing. But a $400–500M miss concentrated in the recurring-quality segment is too large to dismiss as noise.

Cash does not yet rescue the bull. First-half GAAP operating cash flow rose $1.7B to $7.8B, but the filing shows that financing-receivable runoff also increased by $1.7B. Excluding that runoff, operating cash was essentially flat; IBM-defined FCF was $4.76B versus $4.81B. The unchanged full-year guide therefore requires a sharply stronger second half. Meanwhile IBM spent $11.6B for Confluent, recognized $7.24B of goodwill, and carried $48.9B of non-financing debt. The acquisition may be strategically right; the burden is now to prove returns, not merely revenue contribution.

Framing: a captive cash franchise with a contested acquired growth engine. The tag is “cheaper, but the organic engine just failed its test.” I would become constructive if recurring Software ARR remains high-single-digit, delayed transactions convert, organic Software returns above 5%, and cash flow ex-financing receivables rises. I would turn bearish if full-year growth slips below 4%, the $15.7B FCF guide is cut, or ARR decelerates while acquisition amortization and stock compensation keep rising.

🔄 Changes Since the June 10, 2026 Report

The June report’s recommendation was HOLD / accumulate below ~$230–235. Its bull case required at least ~5% organic growth through the z17 down-cycle and Software organic growth above ~5%; its bear case would be falsified by at least 5% organic growth, high-single-digit FCF growth and improving GAAP operating cash flow. The update changes the conclusion in four ways:

The old bull test is broken for the quarter because organic Software was below 5%. It is not permanently falsified: management attributes most of the miss to delayed enterprise buying, and ARR accelerated to 8%. The old bear test remains alive because neither organic growth nor clean cash growth cleared its hurdle. Accordingly, the revised stance is more demanding despite the cheaper price.

📈 Stock Price Action — Five-Year Event Map

Adjusted-close data from AZI’s market feed run from August 2021 through August 7, 2026. IBM rose from $112.32 adjusted to $237.28, a roughly 16% annualized return, but the path now contains a 37.5% maximum drawdown. The stock peaked at $332.41 on June 2, 2026 and is 28.6% below that level. It is above its 21-day EMA ($234.11), below its 50-day ($245.10) and 200-day ($256.03), with negative six- and twelve-month relative strength.

The tape’s message is useful but not dispositive. Factor analysis shows 0.75 market beta, negative Momentum exposure, positive Dividend Yield and Cloud exposures, only 22% explained variance, and approximately 43.5% specific volatility. IBM is no longer a crowded momentum winner, yet it is not statistically behaving like a stable quality factor either. The next re-rating must be earned by company-specific execution.

1. Executive Summary

International Business Machines is a $69.1B TTM enterprise-technology company whose economic quality varies sharply by segment. The installed mainframe and Transaction Processing stack is a captive annuity; Red Hat is a durable but contested hybrid-cloud platform; Automation and Data are increasingly acquisition-built; Consulting is a fragmented people business; Infrastructure is a profitable product-cycle franchise; Financing supports customer purchases but complicates cash-flow interpretation.

The five-year transformation remains real. Kyndryl’s 2021 separation removed low-margin managed infrastructure; Software reached 45% of FY2025 revenue; consolidated gross margin reached 58.2%; and Software ARR rose to $24.6B in Q2 2026. What changed is the confidence interval around the growth engine. Q2 revenue was $17.16B, up 1.1%; gross margin fell 100bp to 57.7%; net income fell slightly to $2.17B; diluted EPS was $2.27. Software grew 5.1% reported but flat organically, Consulting was roughly flat reported / +1.1% constant currency, and Infrastructure fell 7.4% as IBM Z declined 42% against its launch quarter.

There were offsetting positives. Red Hat grew 11%; Software ARR grew 8% to $24.6B; Distributed Infrastructure grew 37% and ended with nearly $500M backlog; Consulting signings grew 6% constant currency; and productivity allowed operating pretax margin to expand 30bp despite the revenue miss. IBM says roughly one-third to 40% of the delayed software transactions had already closed by the earnings call and almost none were lost. The next quarter can therefore distinguish timing from demand.

The balance sheet and cash evidence impose discipline. At June 30, cash and marketable securities were $8.18B, total debt $61.99B, and non-financing debt $48.94B. Goodwill plus acquired intangibles reached $88.55B versus $34.54B of equity. Confluent accounted for most of the step-up. H1 diluted shares rose 0.6% to 952.7M; stock compensation rose 19.6% to $987M; no buyback offset exists. IBM is creating enterprise growth faster than per-share growth.

At $237.28, the stock is no longer priced as though execution is effortless. Market capitalization is $223.5B and core EV is roughly $264.3B after separating $13.05B of Financing debt. That equals 20.98x TTM GAAP EPS, 14.2x guided IBM-FCF and roughly 16.8x core EV/guided FCF. The own-history composite remains elevated at the 77.6th percentile because sales and book multiples are high; P/B is a poor anchor given negative tangible equity. The institutional sections below contain no recommendation or price target; they frame the expectations and evidence.

2. Business Overview

IBM sells integrated enterprise technology — software, consulting services, and computing infrastructure — to large organizations (governments, banks, insurers, telcos, retailers, healthcare). Founded in 1911 (incorporated as the Computing-Tabulating-Recording Co.), headquartered in Armonk, New York, with ~264,000 employees, it is one of the oldest continuously operating technology franchises in the world. The 2021 spin-off of its managed-infrastructure-services business as Kyndryl was the pivotal structural act of the modern era: it removed ~$19B of low-margin, declining revenue (FY2020 revenue of $73.6B fell to $57.4B in FY2021) and concentrated IBM on higher-margin software and consulting.

Segment structure and FY2025 revenue (external):

Source: IBM Q4 2025 earnings release (SEC Form 8-K, ex-99.1, filed 2026-01-28).

Q2 2026 operating snapshot:

Software’s segment margin is not gross margin; it is pretax segment profit divided by segment revenue. Its Q2 gross margin was 82.6%, down 130bp because acquired mix and Transaction Processing weakness diluted it. Consulting gross margin rose 140bp to 28.9%. Infrastructure gross margin fell 310bp to 58.4% because high-margin z17 and associated software declined while lower-margin distributed equipment grew. These mix movements explain why consolidated revenue alone is not an adequate quality measure.

How IBM makes money. The economic engine is Software, which carries an ~83.5% gross margin and represents roughly two-thirds of segment profit despite being ~45% of revenue. Within Software, the highest-quality dollars come from two sources: (1) Transaction Processing — the mission-critical software stack (CICS, IMS, Db2, MQ, z/OS middleware) that runs on IBM Z mainframes, an exceptionally sticky, high-margin annuity tied to the installed base; and (2) Red Hat (acquired 2019 for $34B), now an ~$7.5–8B run-rate hybrid-cloud platform (RHEL, OpenShift, Ansible). The Automation and Data sub-segments are increasingly built through acquisition (Apptio, HashiCorp, StreamSets/webMethods from Software AG, DataStax, Confluent).

Recurring vs. non-recurring. Software ARR reached $24.6B in Q2 2026, up roughly $2B. IBM calculates ARR by annualizing the quarter’s subscription, SaaS, term-license and support revenue; it is not the same as contracted backlog or remaining performance obligations. About 80% of Software revenue is recurring, while the remaining 20% is transactional and caused most of Q2’s miss. Consulting is project- and contract-based: backlog was $30.8B, signings $5.0B and trailing book-to-bill 1.05. Infrastructure is the most cyclical: mainframe revenue surges in the first several quarters of a new z cycle and decays thereafter.

Customer concentration is low at the consolidated level; no single customer is disclosed as material. Bargaining power is not uniformly low. A regulated bank is captive to a core transaction stack but can negotiate a large incremental enterprise agreement; a hyperscaler or large enterprise can multi-source infrastructure and data tools; Consulting projects can be rebid. Revenue quality therefore comes from millions of embedded technical and process decisions rather than one long non-cancellable contract.

The Financing segment should be treated as a small captive lender, not a generic debt-funded appendage. Its $13.0B of debt supports financing receivables and customer transactions. Separating that debt from core EV avoids overstating operating leverage, but the receivable runoff affects consolidated GAAP cash flow. An investor must therefore use segment-adjusted debt and the company’s ex-Financing cash reconciliation together; using only one produces a misleading picture.

Verdict: A coherent, well-segmented enterprise-IT franchise anchored by a genuinely unique mainframe/Transaction-Processing annuity and a credible hybrid-cloud platform (Red Hat), diluted by a large but structurally average consulting business. The mix is improving in IBM’s favor every year. The business is understandable and its cash flows are real — but the reported revenue line blends a high-quality software annuity, a cyclical hardware franchise, and a competitive services business, which makes the headline growth rate a poor proxy for underlying quality.

3. Industry Dynamics

IBM competes across three quite different industry structures, and conflating them is the single biggest analytical error in evaluating the company.

(a) Enterprise infrastructure software & hybrid cloud (the good industry). The market for operating systems, container platforms, automation, integration, and data infrastructure software is large, growing (high-single to low-double-digit), and structurally attractive: high gross margins (80%+), recurring revenue, meaningful switching costs, and consolidation toward platform leaders. IBM/Red Hat is a genuine leader in the hybrid/multi-cloud control layer — the “Switzerland” position of running consistently across on-prem, AWS, Azure, GCP, and private cloud. However, this is also where capital is flooding in (Marathon capital-cycle lens): hyperscalers (AWS, Microsoft, Google), data/AI platforms (Snowflake, Databricks, Confluent before IBM bought it), and a swarm of venture-funded AI-infrastructure startups are all competing for the same enterprise AI/data budgets. High returns are attracting capital; pricing pressure and commoditization risk rise accordingly. IBM’s defensible niche is enterprise control, governance, security, and openness — not raw scale, where it is dwarfed by the hyperscalers.

(b) Mainframe systems & associated software (the great, misunderstood industry). This is a monopoly in structural decline that throws off extraordinary cash. IBM is effectively the sole supplier of mainframe hardware and the z/OS software ecosystem to the world’s largest banks, insurers, airlines, and governments. New entrants are zero — capital has entirely exited the category (the Marathon “left-for-dead” cash cow). The installed base shrinks slowly at the margin but is astonishingly durable: ~70% of global transaction value by dollar still touches a mainframe, COBOL migration projects routinely fail, and the cost/risk of re-platforming mission-critical workloads is prohibitive. Each z-cycle (z15→z16→z17) has been stronger than the last, and z17 (2025) drove IBM Z to its highest revenue in ~20 years. The industry is shrinking in unit terms but the profit pool is defended by an absolute moat. This is the highest-quality structural position IBM holds, and it is the one the market most reflexively discounts.

© IT consulting / systems integration (the bad industry). Consulting competes against Accenture, Tata Consultancy Services, Infosys, Cognizant, Capgemini, Deloitte, and the offshore majors. It is a people business: gross margins in the 25–30% range, limited switching costs, intense price competition (management explicitly cited “a very aggressive pricing environment”), and direct exposure to the deflationary risk of GenAI — the same technology IBM sells can compress the billable-hours model. There are no meaningful barriers to entry beyond brand and relationships; market-share is fluid. Structurally, this is an average-to-poor industry, and IBM is not the share leader (Accenture and TCS are larger and more profitable in services).

Capital-cycle synthesis. The three pools sit at different points in the supply cycle. Mainframe supply has exited: no rational entrant will recreate IBM’s hardware, operating system, tools and support ecosystem for a slowly changing installed base. That lack of entry lets an old franchise retain strong economics. Consulting is oversupplied with globally mobile skilled labor and common third-party tools; AI may increase project demand while simultaneously lowering the labor needed per outcome. Enterprise AI/data software is the most ambiguous. Demand grows rapidly, but Microsoft, AWS, Google, Oracle, ServiceNow, Salesforce, open-source communities and venture-backed specialists all subsidize overlapping control, data and agent layers from different home profit pools.

The peer cross-read clarifies the supply response. Oracle is committing extraordinary capital to AI infrastructure and accepts negative conventional FCF to gain scale. ServiceNow is spending almost $9B for security adjacencies while hyperscalers can bundle similar governance. Cisco is seeing AI-networking orders grow faster than its recurring annuity, but merchant silicon and open standards pressure the new pool. Accenture’s entire services cohort repriced on fear that AI compresses labor revenue. IBM is exposed to all four dynamics, yet does not need to win raw cloud, frontier models or AI hardware. Its defensible position is the control plane for complex hybrid estates and the mission-critical transaction layer. The attractive strategy is to deepen that bounded market; the dangerous strategy is paying growth-company prices to compete everywhere.

Regulation cuts both ways. Data sovereignty, audit, model governance, cyber remediation and sector-specific controls favor trusted vendors with on-premise deployment and long histories in banks and governments. IBM can spread certification and compliance costs across a global base. Interoperability mandates, open-source standards and customer demands for model portability weaken proprietary lock-in. Regulation supports IBM when it makes existing captivity safer; it does not automatically create captivity in new AI workloads.

Verdict — structurally good or bad industry? Mixed, weighted good in the installed base and poor at the growth frontier. Two-thirds of segment profit comes from attractive Software/mainframe positions; Consulting is average; greenfield AI/data markets are capital-rich and contested. Industry growth can be strong while IBM’s incremental return is mediocre. The investment case depends on using captivity to sell adjacencies without paying away the advantage through acquisition premiums.

4. Competitive Position

IBM’s moat is real but bifurcated — extremely strong in two places, thin in a third. Naming the mechanism precisely (Greenwald taxonomy) matters:

1. Mainframe / Transaction Processing — customer captivity + switching costs (the genuine moat). This is the strongest competitive position IBM owns and one of the better moats in all of enterprise technology. The mechanism is captivity: a global bank running its core ledger on z/OS with decades of bespoke COBOL/PL-I/CICS code faces a re-platforming project measured in hundreds of millions of dollars, multi-year timelines, and existential operational risk (failed migrations have bankrupted the ambitions of more than one institution). The result is pricing power that shows up directly in the financials — Transaction Processing software carries some of the highest margins in the company and grows or holds even as the underlying hardware cycles. The financial tell that this is a true moat: if it weren’t, IBM could not raise TP prices through a hardware down-cycle and hold the revenue (which it does — TP was flat in 2025 off a +10% prior year, vs. flat off a −9% in the z16 cycle, evidence of a rising floor). z17’s in-line AI inferencing (Spyre/Telum II) and watsonx Code Assistant for Z (which refactors COBOL and removes the “hard to modernize” objection) deepen the captivity rather than relieving it.

2. Red Hat / hybrid cloud — scale economies + open-source community (a moderate, contested moat). Red Hat’s advantage is a combination of the RHEL standard (the default enterprise Linux, with an installed-base/ecosystem lock-in), the OpenShift container platform (~$2B ARR, +30%), and the credibility of an open, cloud-agnostic stack. The moat mechanism is a mix of switching costs (certified application stacks, operational tooling) and a community/ecosystem network effect. It is real — Red Hat has more than doubled since acquisition (~$3.2B → ~$7.5–8B) — but it is contested: the hyperscalers bundle competing managed-Kubernetes and Linux offerings, and “hybrid/multi-cloud neutrality” is a value proposition that erodes if one hyperscaler wins decisively. We rate this a durable but not impregnable advantage.

3. Consulting — brand and relationships only (weak/no durable moat). IBM Consulting wins on brand, incumbency, domain expertise, and its unique ability to sell alongside IBM software (the “client zero”/integrated-value pitch). These are advantages, but not a moat in the Greenwald sense: a competitor (Accenture, TCS) can and does win the same work; switching costs between SIs are low; and the business is exposed to GenAI-driven deflation of the labor-arbitrage model. The financial signature confirms it — ~25–30% gross margins and revenue that went negative in 2024 before inflecting back to ~1–2% growth in 2025.

4. Automation/Data software — acquired switching costs (moderate). Apptio (FinOps), HashiCorp (infrastructure provisioning/Terraform/Vault), webMethods/StreamSets (integration), and now Confluent (data streaming) each carry respectable product-level switching costs once embedded in enterprise workflows. The moat here is being assembled by acquisition rather than grown organically, which raises the bar on integration execution and price discipline.

Greenwald share-stability and return tests. Mainframe passes the strongest practical test: repeated technology shifts and cloud migrations have not displaced IBM’s effective hardware/OS position, program revenue has improved across z16 and z17, and customers continue expanding aggregate MIPS. Transaction Processing’s Q2 decline does not invalidate the moat; a moat reduces switching and protects long-run economics, not quarterly purchasing variability. The test would fail if workloads and attached software contracts migrated away persistently, MIPS declined, or pricing could no longer offset unit attrition.

Red Hat passes a narrower test. RHEL and OpenShift remain standards in enterprise Linux and hybrid Kubernetes, Red Hat grew 11% in Q2, and OpenShift ARR reached $2.2B. Yet share cannot be cleanly verified from IBM’s disclosures, and hyperscalers provide credible substitutes. The evidence supports demand captivity and bounded scale, not monopoly.

Consulting fails the stable-share/excess-return test. IBM’s 28.9% Q2 gross margin and 12.1% segment margin are respectable but below software economics, revenue growth remains around 1%, and customers can shift work among global firms. Brand, regulated-industry references and software adjacency lower selling cost; they do not stop entry. Data/Automation is not yet testable as one moat because IBM does not publish organic share or cohort returns after acquisitions.

The combined-company ROIC signal is obscured by goodwill. FY2025 ROE above 35% reflects thin book equity, while return on assets was only around 7%. A proper strategic ROIC retains acquisition goodwill because IBM chose to buy the franchises. The proxy’s relative-ROIC modifier is directionally aligned, but public reporting does not show enough to determine whether recent deal cohorts clear the cost of capital. Sustained excess return on the full $11.6B Confluent price—not product quality—is the economic moat test for the acquired portfolio.

Direct competitor comparison: Against Accenture, IBM is a weaker pure-services competitor but a stronger product-plus-services bundle. Against Microsoft/AWS/Google, IBM is vastly sub-scale in raw cloud but differentiated on hybrid/on-prem governance and the mainframe-adjacent estate the hyperscalers cannot reach. Against Oracle, IBM is similarly positioned as a legacy-enterprise incumbent monetizing a captive installed base, but Oracle’s OCI/database cloud has re-accelerated faster. Against Snowflake/Databricks, IBM is a challenger in data/AI, buying its way to relevance (Confluent, DataStax).

Verdict — durable advantage, or crowded market with weak differentiation? A durable advantage in the mainframe/TP core and a solid one in Red Hat; weak differentiation in Consulting and contested differentiation in AI/data software. The blended moat is above-average and, importantly, cash-generative and defended in exactly the places that produce the profit. The risk is that the market is paying for the growth in the contested areas while the quality sits in the defended cash cow.

5. Growth History and Forward Opportunities

The historical record is one of stagnation-then-stabilization-then-reacceleration:

Source: EDGAR XBRL (us-gaap:Revenues), IBM 10-Ks.

The 2025 acceleration looked like the start of durable mid-single-digit growth. Q2 2026 shows why the bridge must be decomposed. First-half revenue was $33.08B, up 5.0% reported and 3.4% constant currency. Acquisitions contributed materially to Software; the strongest organic hardware comparison reversed; and Consulting remained around flat. Management now expects 4–5% full-year constant-currency growth. The low end is below the June bull hurdle and the high end only meets it.

Reported and organic Software separated. Q2 Software revenue was $7.76B, +5.1% reported and +4.6% constant currency, but management said organic growth was flat. Hybrid Cloud grew 11.2%, led by Red Hat at 11%. Data grew 18.9% because Confluent and other acquisitions were included; management acknowledged the acquisition contribution while analyst questioning implied organic Data declined mid-single-digit. Automation grew 3.6%. Transaction Processing fell 8.1%. A portfolio can report healthy growth while the owned-before-acquisition base stalls; that is the current fact pattern.

The counter-signal is recurring demand. ARR rose roughly $2B to $24.6B and accelerated from 7% to 8% growth. About 80% of Software is recurring, and management expects that pool to grow around 10% in the second half. OpenShift ARR reached $2.2B. HashiCorp posted record bookings and Confluent was said to track plan. If those metrics convert, Q2 can prove a transactional air pocket rather than a structural break. ARR is not contracted revenue, however: IBM defines it as the current quarter’s recurring revenue multiplied by four, and cautions it may not represent future recognized revenue.

Mainframe cycle. IBM Z fell 42% against the exceptional launch quarter, worse than management expected because customers redirected capital toward supply-constrained distributed infrastructure. The full z17 program remains near 130% of z16 on a program-to-program basis, according to management, with roughly $1B more hardware and $3B more attached software than z16. Distributed Infrastructure grew 37% and built nearly $500M of backlog. The installed base reportedly carries 140M MIPS; 85% of workloads are stable or growing. Those facts support captivity and a still-healthy cycle, but they do not remove quarterly cyclicality.

Consulting. Q2 revenue was $5.33B, flat reported and +1.1% constant currency. Signings rose roughly 6% constant currency to $5.03B, trailing-twelve-month book-to-bill was 1.05, and backlog was $30.8B. GenAI represented roughly half of signings and more than 30% of backlog. This supports near-term stability. It does not answer whether AI productivity is retained as margin, passed to customers as price deflation, or reduces future billable effort. The Q2 outcome—1% growth and 160bp segment-margin expansion—is consistent with a smaller revenue-per-outcome model that can still create profit if IBM keeps the productivity.

Forward opportunities (and our skeptical weighting):

• GenAI book of business reached >$12.5B cumulative ($10.5B Consulting, $2B Software) before IBM retired the metric after Q4 2025 — a disclosure change that conveniently arrives just as the comparison base gets harder. The opportunity is real (enterprises want governed, hybrid, on-prem-capable AI), but “book of business” (cumulative signings) is not revenue, and Consulting GenAI is lower-margin services work.
• watsonx / Granite / Orchestrate — IBM’s enterprise AI platform and agentic stack. Differentiated on governance, smaller/efficient models, and on-prem/sovereign deployment (“sovereign AI” is a genuinely well-positioned angle). But IBM is a challenger here against hyperscaler-scale model ecosystems; watsonx is not a foundation-model leader and partners (Anthropic Claude, Mistral, Meta, Hugging Face) for frontier capability.
• Mainframe AI inferencing (Spyre/Telum II) — embedding AI in-transaction on Z is a credible, defensible expansion of the captive base and arguably IBM’s most under-appreciated organic growth vector.
• Confluent integration — real-time data streaming is strategically adjacent to watsonx, Red Hat and agents. The filed purchase price was $11.602B, not the lower estimate used in the June report. IBM assigned $7.238B to goodwill and $3.834B to identifiable client, technology and trademark intangibles. Strategic fit is not an investment return; organic cross-sell and post-amortization cash accretion are the tests.
• Project Lightwell — a subscription clearinghouse for remediated open-source packages. Management cited early adoption by major banks and payment firms and a $5–15B TAM. Treat that TAM as a management hypothesis. The favorable mechanism is unusually clear: mainframe/Red Hat trust plus regulated-client captivity can create a paid security layer; the adverse mechanism is that vulnerability remediation becomes a bundled open-source or hyperscaler feature.
• Quantum — IBM disclosed a Department of Commerce letter of intent supporting an Anderon quantum foundry, with up to $1B of CHIPS incentives and a matching IBM commitment, and plans to invest more than $10B in quantum over five years. This is long-duration optionality and a potentially meaningful capital use, not present earnings power. No quantum value is included in base valuation.

Verdict — high- or low-quality growth? The installed recurring pool is high quality; current incremental growth is mixed quality. ARR and Red Hat are organic positives, distributed hardware is cyclical, Consulting growth is thin, and Data growth is substantially purchased. IBM can still deliver 4–5% consolidated growth, but Q2 moved the burden of proof from “show durability” to “repair organic Software conversion.”

6. Financial Quality

Margins are resilient, but the consolidated line now mixes productivity with acquisition cost. Q2 GAAP gross margin fell 100bp to 57.7%; operating gross margin fell 70bp. Yet adjusted EBITDA margin rose 20bp and operating pretax margin rose 30bp because IBM reduced operating expense and continued its AI/automation productivity program. Software segment margin rose 110bp to 32.2% and Consulting rose 160bp to 12.1%; Infrastructure fell 150bp to 21.8% because of mix. Margin control was strong in a weak revenue quarter.

Five-year financial trend (the picture in one table):

Source: IBM 10-Ks and earnings releases. Kyndryl makes 2020–2021 revenue comparisons imperfect. IBM-FCF is a company-defined non-GAAP measure.

First-half 2026 bridge:

This reconciliation resolves the June report’s open question. IBM-defined FCF does not benefit from financing-receivable runoff because the company subtracts the change. GAAP OCF does. Therefore the H1 GAAP OCF acceleration should not be treated as operating improvement, while IBM-FCF is the more useful management definition for this period. It is still non-GAAP and includes net capital expenditure conventions, but the main concern is now the second-half ramp rather than an undisclosed definitional windfall.

TTM IBM-FCF is roughly $14.65B: FY2025 $14.7B less H1 2025 $4.808B plus H1 2026 $4.760B. Management retained guidance for about $1B of year-over-year growth, or roughly $15.7B. Achieving it requires approximately $10.9B in H2, versus about $9.9B in H2 2025. Seasonality is strongly second-half weighted, but the hurdle is real after flat H1.

Cash generation — strong, but with quality caveats:

• IBM-defined free cash flow was $14.7B in 2025 (+16%, +$2B), the highest in over a decade, at the highest FCF margin in IBM’s history. This is the metric management runs the company on, and it is impressive on its face.
• But GAAP operating cash flow was $13.2B and declining (2023 $13.9B → 2024 $13.5B → 2025 $13.2B). Less ~$1.1B capex, GAAP FCF is ~$12.1B — roughly $2.6B below the headline $14.7B. The gap is the difference between IBM’s definition (which nets out the change in Global Financing receivables — currently a source of cash as the financing book runs off from $15.1B to $12.8B — and excludes certain items) and a plain GAAP calculation. As the financing book shrinks, that tailwind is finite. Open question: how much of the $14.7B is durable software cash vs. financing runoff and working-capital/pension timing?
• Pension: IBM carries a large legacy defined-benefit pension. In 2022 it recognized a ~$5.9B one-time pension settlement charge (the reason FY2022 GAAP NI collapsed to $1.6B) when it transferred US obligations to insurers. The overfunded plan now contributes non-cash pension income to operating results — a legitimate but low-quality, market-sensitive earnings input that investors should normalize out.

Balance sheet: At June 30, cash and marketable securities were $8.178B; debt was $61.987B, including $13.047B attributed to IBM Financing. Core debt was therefore $48.940B and net core debt $40.762B. Working capital was negative $7.514B and the current ratio was about 0.79. Those figures are manageable for a recurring, investment-grade cash generator, but less forgiving than June’s snapshot because cash funded Confluent. Goodwill was $74.599B and acquired intangibles $13.955B against $34.541B equity, leaving deeply negative tangible common equity.

Dilution & SBC — the quiet value leak: H1 diluted shares were 952.7M versus 946.7M, +0.6%; stock compensation rose 19.6% to $987M. Period-end basic shares were 942.1M. The five-year share trend remains upward and no material repurchase offsets grants. Investors should subtract SBC or model dilution explicitly rather than treating IBM-FCF as entirely distributable owner cash.

Verdict — do economics improve with scale? Productivity is visible in pretax and segment margins, so operating scale exists. Cash growth ex-Financing receivables was absent in H1, and acquisition goodwill rose faster than earnings. The company passes the margin test but has not yet passed the incremental-return test.

7. Capital Allocation

Capital allocation is now the decisive issue. The strategy—use mainframe and recurring software cash to buy strategic software, fund R&D and preserve the dividend—is coherent. The price paid for growth, financing structure and lack of per-share offset have raised the hurdle.

H1 2026 sources and uses: IBM produced $4.76B of company-defined FCF, paid $3.166B of dividends and spent $10.480B on acquisitions. It financed the gap through cash, marketable securities and debt. Common-stock issuance added $1.284B to equity; no meaningful buyback offset existed. The dividend consumed 67% of H1 FCF, though annual seasonality makes that ratio unusually high.

Confluent underwriting. IBM closed Confluent on March 17. The filed gross purchase price was $11.602B, including a $11.537B cash equity price at $31 per share and assumed awards. Acquired cash and securities totaled $2.082B, putting the price net of those balances around $9.52B before transaction and integration costs. The preliminary allocation assigned $7.238B to goodwill, $2.122B to client relationships with a 12-year life, $1.590B to technology with a seven-year life, and $122M to trademarks. Only 1% of goodwill is expected to be tax deductible. IBM also assumed or issued roughly 3.0M stock awards valued at $665M.

The strategic rationale is credible: Confluent’s data-streaming layer can feed Red Hat, watsonx, Automation and transaction workloads. The economic case requires more. If normalized after-tax cash earnings eventually reach $800M, the net purchase price earns roughly 8%; at $600M it earns roughly 6%. Those are illustrative hurdles, not forecasts. IBM must disclose or demonstrate organic cross-sell, retention, margin and cash accretion sufficient to clear a cost of capital around 8–9%. Reported Data growth alone cannot do that because the acquisition is inside the numerator.

The 10-Q also discloses a definitive July 22 agreement for another business that would enter Infrastructure; the target and price were not disclosed in the filing. That adds prospective integration and funding risk before Confluent’s returns are visible. Separately, planned quantum investment above $10B over five years may include R&D, facilities and partnerships; until IBM gives a cash bridge, it should be treated as a capital claim with optional rather than base value.

Dividend — sacred but slow: IBM is a dividend stalwart (~30 consecutive years of increases), paying ~$6.3B/yr at a ~2.9% yield. But raises are token (~1%/yr; $5.80B in 2020 → $6.26B in 2025), explicitly subordinated to funding M&A. Payout is ~52% of GAAP EPS and ~43% of IBM-FCF — safe, but the de-minimis growth rate signals the dividend is a maintenance obligation, not a growing return of capital. Critically, with the share count rising, aggregate dividend dollars grow while the per-share raise barely outpaces nothing.

Buybacks — absent: IBM has not run a material repurchase since 2019 ($1.36B in 2019, then effectively zero), having redirected capital first to deleveraging the Red Hat acquisition and now to the M&A program. The absence of buybacks while issuing ~1.4%/yr in SBC means stock comp is a pure dilution cost with no offset — the opposite of best-practice capital allocation at a mature, cash-rich company.

Incentives and insider evidence. The 2026 proxy says the 2023–2025 PSU score weighted revenue 40%, operating EPS 30% and FCF 30%, with a relative ROIC modifier of plus or minus 20 points; the modifier was zero for that cycle. Newer awards retain operating measures plus relative-TSR or ROIC modifiers. The presence of ROIC is better than the June report stated, but it is a modifier rather than a primary weight, and the board awarded the CEO’s 2025 annual incentive at 150% of target.

Our complete 24-month Form 4 parse covered 178 filings and 351 reported transactions. It found five open-market purchases totaling 3,620 shares, all by independent directors: Alfred Zollar, David Farr twice, Michelle Howard and Michael Miebach. Two open-market sales totaled 30,143 shares, one by an SVP and one by another SVP; neither filing was marked under a Rule 10b5-1 plan. The remaining activity was primarily grants, option exercises and tax withholding. Director purchases show some alignment but are too small to offset the absence of CEO/CFO open-market buying or the aggregate dilution signal.

Verdict — has management allocated capital intelligently? Kyndryl and Red Hat support management’s strategic record. Confluent, the undisclosed Infrastructure deal, quantum commitments and continuing dilution make the next phase more difficult. Capital allocation is coherent but not yet proven value-creative on a fully burdened per-share basis.

8. Changes and Headwinds — Last Two Years

The last two years moved IBM from portfolio repair to proof of organic compounding. HashiCorp, DataStax and Confluent broadened Automation and Data; z17 validated the mainframe refresh; productivity expanded pretax margins; and ARR rose. Those are genuine advances. Q2 revealed that the integrated portfolio is not yet self-propelling: reported Software growth depended on acquisitions, the transaction-sensitive 20% of Software created a large miss, and organic Software was flat.

The late-quarter buying shift is the most important new fact. Management says clients redirected capital toward servers, storage and memory to secure supply ahead of price increases; cyber concerns also distracted buyers. The shift hurt both IBM Z and attached Transaction Processing/data software. Because enterprise budget envelopes are finite, IBM benefited in Distributed Infrastructure while losing higher-margin Software timing. This is not simply demand destruction. It is also not harmless: it shows IBM’s bundle can cannibalize itself across quarters and that even sticky software transactions remain budget-sensitive.

Guidance changed accordingly. Full-year constant-currency revenue is now 4–5%; Software is 6–8%; FCF still rises about $1B. The low end assumes the current buying dynamic persists; the high end assumes more normal pipeline conversion. Management also expects recurring Software growth near 10% in H2 and Transaction Processing to remain pressured as some activity moves into 2027. Those explicit assumptions form the Q3/Q4 scorecard.

The risk profile also changed through capital commitments. Confluent is now on balance sheet at $11.602B gross purchase price. IBM disclosed a new Infrastructure acquisition agreement and more than $10B of planned quantum investment over five years. Cash fell, core debt rose and negative tangible equity deepened. None creates a liquidity crisis; together they reduce the margin for an integration miss or a FCF guide cut.

Changes verdict: the captive core remains intact and recurring ARR improved, but organic growth, cash conversion and acquisition burden worsened. The June report’s operating bull case failed its first important observation. Valuation improved enough to keep the case investable, not enough to erase the need for proof.

9. Risk Analysis (Risk Matrix)

The risks interact. A transactional delay is manageable if recurring ARR holds and H2 cash converts. It becomes more serious if IBM must protect the dividend, fund integration and quantum investment while core debt remains near $49B. The most plausible downside is not insolvency; it is several years of low per-share growth while acquisition amortization, interest and dilution absorb franchise cash.

10. Valuation Discussion (Embedded Expectations)

No price target; no recommendation. This section frames what the current price requires.

Live capitalization. The August 7 adjusted close was $237.28. Multiplying by 942.134M period-end shares gives $223.55B of common equity. Add $48.940B of non-Financing debt and subtract $8.178B of cash and securities for a core enterprise value of $264.31B. Financing debt is separated because it funds receivables and belongs to the Financing segment’s economics. Using total debt without the paired assets would overstate operating leverage.

These multiples are materially lower than June’s. The own-history signal also cooled, but less dramatically than price alone suggests. AZI’s August 7 composite is 77.56th percentile: P/E 20.98x at the 68.3rd percentile, P/S 3.27x at the 85.7th and P/B 6.85x at the 78.7th. P/B is not economically useful because tangible equity is negative. P/S says the market still capitalizes IBM as a richer revenue mix than its history; the Q2 organic evidence has not yet earned that assumption.

Earnings-power value. A Greenwald EPV asks what normalized distributable earnings are worth with no growth. Using $14.0–15.0B of sustainable levered owner cash—below the $15.7B guide to allow for SBC and cyclicality—and an 8.5–10.0% equity discount rate produces equity EPV of roughly $140–176B, or $149–187 per current share. The range is intentionally conservative: it capitalizes no organic growth and treats IBM-FCF as owner cash after a modest normalization. Current market capitalization exceeds the midpoint by roughly $65B. That excess is the market’s franchise-growth value, not support from current earnings power alone.

Reproduction value is harder and lower. Reported tangible common equity is approximately negative $54B. That is not a liquidation estimate: internally built software, decades of mainframe code, trained enterprise sales/support capacity, certifications and customer integrations are expensed. Capitalizing three years of R&D would add roughly $23–25B before obsolescence; rebuilding client relationships and a global support ecosystem would cost more and still would not recreate customer captivity. Even a generous adjusted asset calculation remains below EPV and far below enterprise value. In Greenwald’s framework, EPV above adjusted asset value confirms a franchise, while market value above EPV requires profitable growth. IBM’s mainframe and Red Hat deserve franchise value; Confluent and greenfield AI growth must earn the remainder.

Reverse expectation. A one-stage Gordon inversion is crude but informative. At $223.55B equity value and $14.65B starting FCF, a 9% cost of equity implies roughly 2.3% perpetual FCF growth; at 8% it implies roughly 1.3%, and at 10% roughly 3.2%. Using the $15.7B guide reduces those requirements. This looks modest, but it ignores acquisition outlays, periodic mainframe cycles and dilution. If the company must continually spend billions on M&A to produce that growth, reported FCF is not the complete reinvestment charge. The stock is inexpensive only if post-acquisition organic cash grows.

Peer framing. Peer reports were re-read rather than mechanically averaged because business models diverge. Accenture recently traded near 10x trailing earnings and a roughly 14% FCF yield after its own AI-services reset, despite net cash and higher clean ROE. Oracle traded near 18x guided non-GAAP EPS but carried negative conventional FCF, a large infrastructure build and last-notch investment-grade credit. ServiceNow traded around 28x strict FCF / 54x after-SBC owner cash for near-20% subscription growth. Cisco traded around 41x TTM FCF after an AI-networking re-rating. IBM sits between a distressed services valuation and premium software/infrastructure valuations. Its 14.2x guided FCF is not demanding; its 3.83x core EV/revenue is demanding unless Software mix and organic growth keep rising. A blended peer average would obscure that tension.

Illustrative operating scenarios—not price targets:

The bear does not require collapse: IBM can remain a sound franchise and generate poor equity returns if growth is 2%, shares rise and the multiple normalizes. The base requires the low end of guidance plus modest margin improvement; it can generate acceptable owner returns from a 7% starting guided-FCF yield. The bull requires evidence absent in Q2—organic Software acceleration and per-share discipline—not merely reported acquisition growth.

Verdict: valuation moved from clearly demanding to conditionally attractive. The current price no longer assumes double-digit cash compounding, but it still values the enterprise above no-growth earnings power and near the upper quartile of its sales history. The margin of safety is in successful H2 conversion, not tangible assets.

11. Variant Perception

Consensus / market belief: Before July 14, the tape treated IBM as a scarcity AI incumbent: captive mainframe cash, Red Hat growth, quantum optionality and expanding margins justified a premium. The 25% preannouncement decline replaced that with a timing-versus-structural debate. The rebound from $199 to $237 suggests investors believe the FCF guide and some delayed-deal recovery, while the stock below its 50- and 200-day averages shows the old compounder consensus has not returned.

Strongest bull case: Q2 was an enterprise budget-allocation accident, not a demand break. Clients bought scarce hardware before price increases, almost no software deals were lost, 30–40% had already closed, ARR accelerated to 8%, Red Hat grew 11%, distributed infrastructure grew 37%, signings rose and margins expanded. The 80% recurring Software pool grows near 10% in H2, Transaction Processing normalizes in 2027, and H2 cash delivers the $15.7B guide. At a 7% guided FCF yield, shareholders need steady mid-single-digit per-share cash growth rather than a heroic re-rating.

Strongest bear case: Management’s timing explanation masks an organic stall. Software was flat organically despite an AI spending boom; Data’s 19% growth was purchased; TP fell 8%; Consulting grew 1%; and H1 FCF was flat. IBM spent $11.6B on Confluent at a large goodwill allocation, plans another Infrastructure acquisition and more than $10B of quantum investment, while stock compensation and shares rise. The FCF guide is an H2 promise and the own-history sales multiple remains elevated. IBM becomes a serial acquirer using captive-mainframe cash to rent growth at low returns.

The 3–5 assumptions that matter most:

1. Delayed-deal conversion: does the 30–40% closure rate reach a majority without discounting or losses?
2. Recurring versus transactional Software: can ARR stay 8–10% while the 20% transactional pool recovers?
3. H2 cash conversion: does IBM produce roughly $10.9B of H2 FCF without another financing/working-capital distortion?
4. Confluent ROIC: do disclosed organic growth, margin and cross-sell justify $11.6B?
5. Per-share discipline: do buybacks or lower SBC stop persistent share growth?

What would falsify each side: The bull is falsified if FY2026 CC growth falls below 4%, the FCF guide is cut, or Software organic remains below 5% while ARR decelerates. The bear is falsified by two consecutive quarters of greater than 5% organic Software growth, clean operating-cash growth ex-Financing receivables, and disclosed evidence that recent acquisitions earn above the cost of capital.

Variant lean: the market may now overstate the permanence of one transaction-heavy miss, but the burden of proof properly shifted to IBM. The interesting variant is not “AI wins” or “mainframes never die.” It is that the captive core can fund satisfactory returns at a lower valuation even if growth is only moderate—provided management stops converting that cash into low-return goodwill. The operating and allocation conditions must be separated.

12. Fact vs. Interpretation

13. Open Questions

1. Of the $400–500M Q2 Software shortfall, how much closes in Q3, at what price, and how much moves to 2027?
2. What were organic growth rates for Hybrid Cloud, Automation and Data after each acquired business?
3. What bridge makes H2 IBM-FCF rise by roughly $1B year over year after H1 was flat?
4. What revenue, gross margin, retention and synergy assumptions underpinned the $11.602B Confluent price?
5. When will IBM disclose realized ROIC by acquisition cohort, including goodwill, integration cost and SBC?
6. What business is covered by the July 22 Infrastructure acquisition agreement, at what price and with what funding?
7. How much of the more-than-$10B quantum plan is incremental cash R&D/capex versus existing spend, and what milestones gate it?
8. Can Consulting hold or improve profit if revenue per engagement falls through AI productivity and competition?
9. Why should recurring ARR be annualized from one quarter rather than disclosed as contractually committed backlog or NRR?
10. When will repurchases at least offset employee issuance and stop per-share leakage?

14. What Must Be True (Bull and Bear)

BULL CASE — what must be true:

1. Recurring Software grows near 10% in H2 and ARR remains at least 8% while transactional revenue normalizes.
2. Organic Software returns above 5% for two quarters, establishing Q2 as timing.
3. Full-year constant-currency revenue stays within 4–5% and the $15.7B FCF outcome is delivered without balance-sheet harvesting.
4. Productivity expands pretax and FCF margins after integration costs rather than only offsetting gross-margin pressure.
5. Confluent and HashiCorp produce disclosed organic cross-sell and post-tax returns above 8–9% while share growth slows.

Falsification test: FY2026 CC revenue below 4%, an FCF-guide cut, or another quarter of Software organic growth below 5% accompanied by ARR deceleration breaks the repaired-growth thesis.

BEAR CASE — what must be true:

1. Q2 exposed an underlying 2–3% organic company funded by cyclical hardware and M&A.
2. ARR eventually follows transactional weakness, showing that the annualized measure lagged demand.
3. H2 working capital cannot close the FCF gap, forcing a guide cut or balance-sheet use.
4. Confluent and further acquisitions fail to clear cost of capital after amortization, interest and dilution.
5. Consulting AI productivity is competed away through price, shrinking both revenue and eventual margin.

Falsification test: Two consecutive quarters of organic Software above 5%, consolidated organic growth near 5%, rising operating cash ex-Financing receivables and measurable acquisition ROIC above 9% would invalidate the serial-acquirer/value-trap case.

15. Source Appendix

Appendix B provides the dated source ledger. The evidence hierarchy was: (1) IBM’s Q2 2026 10-Q and earnings exhibits; (2) FY2025 10-K and 2026 proxy; (3) five years of 10-K/10-Q/8-K filings and two years of raw Form 4 ownership XML; (4) the Q2 earnings call for management explanations; and (5) AZI/ROIC/FactorsToday only for price, standardized financial and factor cross-checks. Filing values control every conflict. The SEC mirror contains 631 locally saved documents, including all five annual reports, fifteen quarterlies, eighty-three 8-Ks and the full Form 4 set since August 2021.

The Q2 preannouncement, release and call were treated as distinct sources. The filing establishes financial facts; the call supplies claims such as delayed-deal closure, mainframe MIPS and future recurring growth. ARR, GenAI book of business, program-to-program z17 comparisons, Lightwell TAM and quantum plans are management-defined measures or hypotheses unless explicitly reconciled. Insider analysis used 178 raw Form 4s from August 10, 2024 onward rather than the SEC’s rendered XSL pages. Market capitalization and core EV were recomputed at the August 7 price because standardized enterprise-value feeds carried period-end prices and included Financing debt.

All public sources were accessed through August 10, 2026.

This memo contains no investment recommendation and no price target outside the clearly labeled “Claude’s Take” block. It is general information, not investment advice.

APPENDIX A — Standard Diligence Questionnaire

IBM (NYSE: IBM) — Standard Diligence Questionnaire Appendix

Supplemental to the research memo. Fact / Interpretation / Assumption labels applied where it matters. Figures as of 2026-08-10.

General

What thoughtful questions have other investors asked about this company? The Q2 call concentrated on whether the $400–500M shortfall was delay or demand, what portion had closed, why Transaction Processing and Data weakened, and whether the 4–5% guide assumes continued capital-budget diversion. Management said roughly one-third to 40% of delayed transactions had closed and almost none were lost. Investors also pressed on recurring Software growth, z17 program strength, distributed-infrastructure backlog and the second-half FCF ramp. The least-answered questions remain Confluent ROIC, organic Data growth, per-share dilution and the cash milestones for quantum investment.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Mixed. IBM Z has moved off its launch peak and fell 42% in Q2, while Distributed Infrastructure grew 37%. Transaction Processing was also weak. Consolidated earnings are below the z17 launch benefit but productivity keeps pretax margins near a high. It is a hardware/mix down-cycle inside a structurally improved profit base.

Driven by the external environment or internal actions? Both. Internal: the deliberate software mix-shift, the $4.5B AI productivity program, and margin expansion are management-driven. External: enterprise AI demand and the z17 refresh cycle are environment-driven tailwinds.

How stable are revenues? About 80% of Software is recurring and ARR is $24.6B. Consulting backlog is $30.8B. Q2 demonstrated that the 20% transactional Software pool plus hardware can still move consolidated growth materially. Stability is high, not SaaS-like.

Outlook for products/services? FY2026 constant-currency revenue is guided to 4–5%, Software 6–8%, and FCF roughly $1B above 2025. Recurring Software is expected around 10% in H2; Transaction Processing remains pressured and some activity moves to 2027.

How big will this market be? Growing. Enterprise hybrid-cloud/AI/automation software is a multi-hundred-billion-dollar, high-single-digit-growth TAM; the mainframe profit pool is structurally shrinking but defended; IT consulting is large, slow-growth, and GenAI-disrupted. Roughly half of revenue is non-US (international + domestic balanced).

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More in AI/data software (hyperscalers + startups flooding in — Marathon capital-cycle warning); stable/monopolistic in mainframe; intensely competitive in consulting.

How profitable is the business (ROIC, ROE)? FY2025 ROE above 35% is flattered by thin equity. Q2 Software gross margin was 82.6%, consolidated 57.7%. Goodwill and intangibles now total $88.6B; acquisition-inclusive ROIC, not ROE, is the decision metric. Public disclosure is insufficient to prove recent cohorts exceed the cost of capital.

How profitable is the industry — competitors, barriers? Software: high margins, high barriers (mainframe = absolute barrier; hybrid cloud = moderate). Consulting: 25–30% gross margins, low barriers. Mainframe: monopoly economics, zero new entrants.

Can the business be easily understood? Yes, at the segment level — though the blend of a software annuity, a cyclical hardware franchise, and a services business makes the consolidated growth rate misleading without decomposition.

Can it be undermined by foreign low-cost labor? Yes, in Consulting — directly exposed to Indian-heritage SIs (TCS, Infosys) and offshore labor arbitrage. Software and mainframe are insulated.

Do brands matter? Moderately. “IBM” carries enterprise trust/credibility (security, governance, support) that genuinely aids enterprise and government sales, but brand is not the core moat — switching costs and captivity are.

Nature of competition / customers’ switching costs? High for mainframe/TP (re-platforming is prohibitively risky/expensive) and embedded software; moderate for Red Hat/OpenShift; low for consulting.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? The mainframe installed base and customer relationships (the true moat) are not capitalized; the Red Hat/acquired-software franchises are carried at acquisition cost. Interpretation: economic value of the captive base exceeds book.

Off-balance-sheet liabilities? Operating leases, purchase commitments, and a large defined-benefit pension (currently net-funded, contributing non-cash income) are the items to watch; the pension is market-sensitive (a 2022 $5.9B settlement charge is precedent for volatility).

How conservative is the accounting? Mixed. IBM-FCF excludes the change in Financing receivables, correctly preventing H1 runoff from flattering the measure; GAAP OCF did rise because of that runoff. Operating results exclude acquisition amortization, SBC and restructuring in several presentations. Investors should retain acquisition amortization/interest when judging M&A and subtract SBC or model dilution.

How CapEx-hungry? Low — capex ~$1.1B on $67.5B revenue (asset-light software/services). The real “investment” is M&A and capitalized software, not physical capex.

Capital Allocation & Management

How much FCF, and how is it used? ~$14.7B IBM-FCF (2025). Uses: ~$8.3B M&A, ~$6.3B dividends, ~$0 buybacks. Philosophy: fund a software acquisition roll-up and a sacred (slow-growing) dividend; deleverage between deals; do not repurchase stock.

Significant acquisitions recently? Confluent closed for a filed $11.602B gross purchase price; $7.238B became goodwill. HashiCorp, DataStax, Apptio and StreamSets/webMethods preceded it. Another undisclosed Infrastructure acquisition agreement was signed July 22.

Buying back shares? No — essentially none since 2019.

Issuing large amounts of stock to insiders? Yes — SBC drives ~1.4%/yr dilution (shares 897M→949M over 5 years) with no buyback offset. This is the key per-share concern.

Compensation policy / motivations of management? The 2023–2025 PSU program weighted revenue 40%, operating EPS 30% and FCF 30% and included a relative-ROIC modifier of ±20 points; it was zero in that cycle. ROIC is present but secondary. The CEO’s 2025 annual incentive paid at 150% of target.

Valuation & Market Data

ADR, MLP, or K-1? No — common stock, US domestic filer (10-K/10-Q). Standard 1099 dividend.

Dividend policy? ~30 consecutive years of increases; ~2.9% yield; ~1%/yr raises; ~52% of EPS / ~43% of FCF payout — safe but barely growing.

How profitable is the business? TTM IBM-FCF is roughly $14.65B and guided FY2026 FCF roughly $15.7B. H1 was flat; per-share compounding remains muted by dilution.

Net income vs. cash from operations diverging? H1 GAAP OCF rose $1.695B, but Financing-receivable runoff increased $1.658B. Excluding it, operating cash rose only $38M and IBM-FCF fell $48M. The clean cash evidence was flat.

Risks & Downside

What would cause the stock to decline? Failure of delayed Software deals to convert, ARR deceleration, an FCF-guide cut, poor Confluent returns, another debt-heavy acquisition, or continued multiple normalization from the 77.6th own-history percentile.

Risk of catastrophic loss? Low. IBM is investment-grade, diversified and cash-generative. Beta is roughly 0.75. The realistic downside is low per-share compounding and de-rating, not going-concern loss.

Chance of total loss? De minimis. IBM is a 114-year-old, profitable, IG-rated enterprise with durable cash flows.

Recent News & Events

Has the business environment changed recently? Yes. Clients redirected late-quarter capital toward supply-constrained hardware; Software transactions slipped; IBM preannounced the miss; full-year revenue guidance fell to 4–5%; FCF guidance held. Confluent closed, Lightwell launched, and IBM disclosed large quantum-foundry ambitions.

Significant acquisitions / accounting changes / new markets? Confluent moved into Software/Data with $3.8B of identifiable intangibles. A pending Infrastructure deal remains undisclosed. Lightwell targets trusted open-source remediation, and quantum investment/foundry plans create a new long-duration capital program.

APPENDIX B — Source Appendix

IBM (NYSE: IBM) — Source Appendix

All sources accessed through 2026-08-10. Primary filings and exhibits control. Management transcript claims are identified; quantitative feeds were reconciled to filings and live price.

Primary — SEC filings (EDGAR, CIK 0000051143)

Primary — XBRL financial data

• SEC EDGAR XBRL company facts, CIK 0000051143: https://data.sec.gov/submissions/CIK0000051143.json and issuer filings. Used as a cross-check; period-specific 10-Q/10-K statements control.

Primary — Earnings-call transcripts

• IBM Q2 2026 Earnings Call, 2026-07-22. Used for delayed-deal conversion, organic Software, ARR outlook, program-to-program z17, backlog and Lightwell claims. Transcript claims were not promoted to filed fact.
• IBM Q4 2025 and Q1 2026 calls, 2026-01-28 and 2026-04-22. Used for June-thesis continuity and original guidance.

Secondary — quantitative aggregators (reconciled to filings)

• AZI price history: https://azitrading.com/controls/download-data.php?t=IBM — adjusted closes, moving averages, beta, five-year drawdown and event map through 2026-08-07.
• AZI fundamentals / valuation index: TTM EPS $11.3125 and own-history composite 77.56th percentile (P/E 68.3rd, P/S 85.7th, P/B 78.7th). Statement arrays were not used; filings control.
• ROIC standardized statements: annual/quarterly financial cross-check and Q2 transcript. Market capitalization and EV were rejected where period-end price was stale.
• FactorsToday: base/all-factor loadings, relative strength and explained variance through 2026-08-10. Used only for positioning, not valuation.

Methodology notes

• Core EV uses June 30 period-end shares, non-Financing debt and cash/securities re-keyed to the August 7 price.
• IBM-FCF is reconciled exactly for H1: GAAP OCF less the change in Financing receivables and net capital expenditure.
• Form 4 raw XML URLs were created by removing SEC XSL wrapper paths; transaction codes and 10b5-1 flags were parsed from the XML and checked against filing counts.
• The report applies Greenwald customer-captivity, share-stability, EPV-versus-asset-value and Marathon capital-cycle tests. Management TAM, pipeline and future-return claims remain hypotheses.