International Business Machines Corporation (NYSE: IBM) — The Price Reset Held; the Proof Still Hasn’t Arrived

Report date: August 30, 2026
Ticker: NYSE: IBM | Sector: Information Technology — Software & Services
Price reference: $235.59 (August 28 close) | Shares: 942.1M | Market capitalization: $222.0B | Core enterprise value: $262.7B
TTM / FY2026 guide: Revenue $69.1B | GAAP EPS approximately $11.44 | IBM-defined free cash flow $14.65B TTM | FY2026 IBM-defined free cash flow approximately $15.7B

⚡ Claude’s Take

This block is Claude’s subjective opinion. It is not investment advice or a solicitation. The institutional analysis that follows (§1–§15) contains no recommendation and no price target; only this clearly labeled block expresses a directional view or valuation zone.

Verdict: HOLD and require proof. The reset repaired the valuation, but the evidence has not repaired the thesis. I would become more constructive around $210–220 or after two quarters of organic Software growth above 5% accompanied by clean cash growth. I would not short the captive franchise at today’s price. Conviction: medium.

IBM at $235.59 is 28% below its June closing peak and offers a 7.1% yield on the company’s $15.7B free-cash-flow guide. That is a very different proposition from the momentum story the market briefly valued above $320. The mainframe and Transaction Processing estate remains one of enterprise technology’s strongest customer-captivity franchises; Red Hat remains a credible hybrid-cloud control layer; and the company still has a plausible path to mid-single-digit revenue growth with margin expansion.

But nothing operating has changed since Q2. The latest filed quarter still shows Software up 5% reported but flat organically, Transaction Processing down 8%, and first-half free cash flow essentially flat after removing the Financing-receivable effect. Management retained the full-year cash guide, which now requires $10.94B in the second half—10.6% more than a year earlier. The explanation that enterprise software transactions slipped rather than disappeared is plausible, and management said 30–40% had already closed in early Q3. It remains a claim awaiting a filed quarter.

The post-August 10 news is strategically coherent but does not close that evidence gap. A $240M Together AI cloud agreement is tangible demand; the OpenAI partnership broadens IBM Consulting’s toolkit; a dual-architecture processor could make IBM Z more accessible; and cryogenic and HRL milestones advance the quantum roadmap. None discloses enough revenue, margin, exclusivity, adoption, or return data to prove that IBM’s acquired growth engine can compound per share. Indeed, partnering with leading external model providers reinforces the view that IBM’s defensible role is governed integration into regulated workflows—not proprietary frontier-model leadership.

Capital allocation is the central tension. IBM spent an economically adjusted $10.56B of enterprise cash on Confluent and recognized $7.24B of goodwill. The transaction needs roughly $850–956M of annual after-tax earnings merely to clear an 8–9% return hurdle. Meanwhile diluted shares rose almost 5% from 2021 through 2025, no open-market buyback offsets employee issuance, and management’s current long-term incentives no longer include an acquisition-inclusive return metric. A sound strategy is not the same as a proven return.

My framing is therefore “captive core, contested growth, adequate price.” The likely catalyst is not another AI or quantum announcement; it is Q3 evidence that delayed Software deals converted, ARR held, clean operating cash rose, and the second-half guide became mechanically credible. The downside case does not require technological obsolescence. It requires only 1–2% organic growth, continued dilution, and repeated purchases of growth at returns near the cost of capital.

📈 Stock Price Action — Five-Year Event Map

IBM’s adjusted close rose from $110.07 on August 27, 2021 to $235.59 on August 28, 2026, a 114% total gain and approximately 16.4% annualized. The adjusted closing low was $96.92 on November 26, 2021; the adjusted intraday low that day was $95.87. The closing peak was $326.89 on June 2, 2026 ($329.23 unadjusted), with a $332.46 intraday high. The current price is 28.4% below that closing peak.

At August 28, the shares sat 0.7% above the 21-day exponential moving average of $233.92, 1.6% below the 50-day average of $239.36, and 6.3% below the 200-day average of $251.47. Three-, six-, and twelve-month total returns were approximately -26.0%, -0.2%, and -1.6%. Since the August 7 baseline close, the unadjusted price fell 0.7%, but the August 10 dividend made total return roughly flat.

This is now a company-specific repair setup. A third-party factor model attributes only about 22% of recent variance to its market, technology, cloud, dividend, momentum, and infrastructure factors, with approximately 43.5% stock-specific volatility. Cloud and technology exposures strengthened while IBM lagged them; momentum exposure turned negative. Those are model-dependent observations, not valuation evidence, but they argue against calling the decline a generic technology rotation or the rebound a crowded momentum trade. The five-year series comes from the AZI daily-price feed; factor context comes from FactorsToday.

🔄 Changes Since the August 10, 2026 Report

There has been no new operating quarter, guidance update, or filed segment result. The thesis tests therefore have not moved. What changed is IBM’s external validation, technology roadmap, funding, and insider record.

The Together AI agreement is the strongest incremental evidence because it names a dollar amount and deployable infrastructure. At $240M over multiple years, however, the headline represents less than 0.4% of one year of IBM revenue before timing and associated costs. The company release does not say whether IBM owns the accelerators, what utilization is guaranteed, or what gross margin the arrangement earns.

The OpenAI partnership is strategically sensible. Large regulated customers increasingly need a neutral integrator to choose models, connect proprietary data, implement controls, and manage change. Yet Microsoft, Accenture, and many systems integrators can offer similar services. The announcement proves ecosystem participation, not scarcity.

The dual-architecture processor, cryogenic milestone, and HRL completion reinforce a long-cycle engineering culture. They deserve option value only after commercial milestones are visible. Finally, the Canadian-note prospectus establishes roughly C$119.25M of annual coupon cost at a 4.34% weighted rate; it does not establish what management will buy.

1. Executive Summary

International Business Machines is a $69B enterprise-technology company composed of businesses with very different economics. Transaction Processing software and the installed IBM Z estate are a captive annuity. Red Hat is a credible but contested hybrid-cloud platform. Automation and Data are increasingly acquisition-built software portfolios. Consulting is a low-barrier people business with valuable client relationships. Infrastructure is a high-value but cyclical product franchise. Financing facilitates sales and complicates consolidated cash-flow interpretation.

The five-year transformation is genuine. From 2021 through 2025, revenue rose from $57.35B to $67.54B, a 4.2% compound rate; gross profit rose from $31.49B to $39.30B; gross margin expanded roughly 330 basis points to 58.2%; and R&D increased from $6.49B to $8.32B. Kyndryl’s separation removed a declining, low-margin managed-infrastructure operation. Software reached 44% of FY2025 sales and carries approximately 83% gross margins.

The transformation has not yet proved attractive acquisition-inclusive returns. Diluted shares rose from 904.6M in 2021 to 948.7M in 2025. Cumulative 2021–2025 R&D was $35.6B, acquisition spending net of cash acquired was $22.3B, and dividends consumed $30.3B. In the first half of 2026 alone, acquisition spending was $10.48B and dividends were $3.17B versus IBM-defined free cash flow of $4.76B. Goodwill and acquired intangibles reached $88.55B at June 30, versus $34.54B of equity and negative $54.01B of tangible equity.

Q2 is the fulcrum. Revenue of $17.16B grew 1.1%, Software grew 5.1% reported but was flat organically, Consulting grew 0.2% reported and 1.1% at constant currency, and Infrastructure declined 7.4%. Red Hat grew 11%, Software ARR reached $24.6B and rose 8%, distributed infrastructure grew 37%, and margins benefited from productivity. But IBM cut constant-currency Software guidance from at least 10% to 6–8% and company revenue guidance to 4–5%. The Q2 earnings exhibit and 10-Q are controlling evidence; management’s delayed-deal explanation is a hypothesis until Q3.

Cash quality is better than a bear would imply and weaker than the headline suggests. H1 GAAP operating cash flow was $7.77B, but $2.26B came from the change in Financing receivables. Removing that effect produced $5.50B; subtracting $743M of net capital expenditure yielded $4.76B of IBM-defined free cash flow, down 1.0% year over year. The FY2026 guide remains approximately $15.7B. That guide is achievable, but it requires a steep second-half conversion and should not be confused with strict owner cash after all capitalized software and acquisition reinvestment.

At $235.59, equity value is $221.96B and core enterprise value is $262.72B after separating debt associated with Financing. The shares trade at approximately 20.8 times filing-based TTM GAAP EPS, 15.2 times TTM IBM-defined free cash flow, 14.1 times guided free cash flow, 3.8 times core EV/revenue, and 15.3 times vendor-defined EBITDA. Those are supportable if organic Software resumes, cash converts, and acquired businesses clear their cost of capital. They are not asset-supported: a no-growth earnings-power stress is well below market value.

The central question is therefore not whether IBM will survive AI. It is whether a captive, slow-changing core can fund a higher-growth portfolio without management paying away the economics. The next answer comes from organic Software, clean cash, acquisition disclosure, and per-share discipline—not from another total-addressable-market slide.

2. Business Overview

IBM sells software, consulting services, infrastructure, and financing to large enterprises and public institutions. Its value proposition is strongest where downtime is intolerable, data is regulated, applications are old and interdependent, and a client wants one accountable supplier across hardware, software, cloud, and services.

Software: the economic center

Software is organized around Hybrid Cloud, Automation, Data, and Transaction Processing. Red Hat supplies RHEL, OpenShift, and Ansible. Automation includes products assembled through acquisitions such as Apptio and HashiCorp. Data includes watsonx, Db2, DataStax, and now Confluent. Transaction Processing includes the operating, database, messaging, and middleware layers required to run mission-critical workloads on IBM Z.

Approximately 80% of Software revenue is recurring. IBM’s ARR annualizes recurring revenue recognized in the quarter; it is not contracted backlog, net retention, or remaining performance obligations. ARR of $24.6B and 8% growth are useful direction indicators, but they do not eliminate the transactional exposure that caused Q2’s miss. The remaining roughly 20% includes term transactions and capacity-related purchases that can shift between quarters.

The segment earned an 82.6% Q2 gross margin and a 32.2% segment pretax margin. The gap reflects selling, R&D, acquisition amortization, and other operating costs. Transaction Processing is likely much more profitable than the average, while acquired Data products are still absorbing integration and amortization. Segment-level profit therefore overstates the maturity of every component and understates the economics of the captive core.

Consulting: strategic adjacency, average industry

Consulting embeds IBM inside client transformations and can pull through Red Hat, automation, data, mainframe modernization, and external products. Backlog was $30.8B, signings were $5.0B, and trailing book-to-bill was approximately 1.05 at Q2. Its economic moat is modest: relationships, domain knowledge, security clearances, accumulated implementation assets, and brand reduce client risk, but projects can be rebid and global labor can be substituted.

AI creates both demand and deflation. Clients need process redesign, data preparation, controls, model selection, and systems integration. The same tools can reduce hours required per engagement and invite clients to demand lower prices. IBM’s Q2 Consulting segment margin rose 160 basis points to 12.1% despite only 1.1% constant-currency revenue growth. That is evidence of productivity, not yet proof that savings will remain with shareholders rather than being competed away.

Infrastructure and Financing: reinforcing, cyclical, and easy to misread

IBM Z mainframes handle high-volume, high-reliability workloads for banks, governments, insurers, and airlines. New system launches create pronounced cycles: hardware revenue jumps, then falls as the installed base digests capacity. Associated software capacity, maintenance, and future upgrades create a much longer profit stream. Q2 Infrastructure fell 7.4% as IBM Z declined 42% against its launch comparison, partly offset by 37% growth in distributed infrastructure.

Financing has $13.05B of debt paired with receivables and customer transactions. For valuation, subtracting all debt without recognizing those earning assets exaggerates leverage. For cash flow, leaving receivable movements in consolidated operating cash can exaggerate conversion. The appropriate treatment is to separate Financing debt in enterprise value and show both GAAP operating cash and operating cash excluding the change in Financing receivables.

No customer is disclosed as material to consolidated revenue. Economic concentration instead lies in product ecosystems and large-enterprise budgets. A bank cannot easily abandon its transaction stack, but it can negotiate a large incremental agreement. A cloud-native customer can choose among Red Hat, hyperscaler-native tools, and other platforms. A Consulting client can rebid the project. IBM’s revenue quality is thus a portfolio of embedded technical dependencies, not a single contractual lock.

3. Industry Dynamics

IBM spans three capital cycles: a harvested mainframe ecosystem, a crowded hybrid-cloud and AI software buildout, and an oversupplied consulting market. A single technology-sector multiple misses those supply-side differences.

Mainframe and transaction processing: harvested supply, durable captivity

Mainframe is a small industry with extraordinary barriers. Recreating the processor architecture, operating system, middleware compatibility, service network, security certifications, and decades of customer code would require enormous sunk cost. The addressable market is not growing fast enough to attract a rational new full-stack entrant. Clients could migrate workload away over many years, but the migration itself is costly and operationally risky. That creates customer captivity rather than classical product differentiation.

This is attractive under a capital-cycle lens. Supply does not rush in when returns are high because the market is too specialized, installed-base network effects are one-sided, and buyers demand proven reliability. IBM can harvest economics while spending enough to preserve compatibility, security, performance, and regulatory confidence. The risk is not a new mainframe vendor. It is gradual workload abstraction, cheaper distributed alternatives, and AI-assisted code modernization reducing migration cost at the margin.

Regulation can reinforce and burden the position simultaneously. European supervisors designated IBM among the critical ICT third-party providers under DORA, alongside firms such as Microsoft, Oracle, AWS, SAP, Accenture, Kyndryl, and TCS. The designation corroborates IBM’s embedded role in financial infrastructure while increasing oversight, resilience, reporting, cyber, and concentration obligations. See the official provider list and designation methodology.

Hybrid cloud, data and AI: large opportunity, abundant capital

The hybrid-cloud control layer is structurally attractive: high gross margins, subscriptions, switching costs, and a persistent need to operate across on-premises systems and multiple public clouds. Red Hat’s open-source distribution, developer ecosystem, certifications, and neutrality create a defensible lane. IBM disclosed 2024 Red Hat revenue of approximately $6.5B, including OpenShift ARR near $1.4B and Ansible ARR near $0.5B at its 2025 investor day.

The competitive supply response is intense. Microsoft’s FY2026 cloud revenue reached $214.4B and Azure grew 41% in its June quarter. AWS Q2 sales rose 37% to $42.2B and operating income reached $16.6B. Oracle reported 39% cloud growth while conventional free cash flow turned deeply negative under infrastructure investment. These companies are willing to spend tens of billions on data centers, accelerators, networking, and models. Venture-funded specialists compete at every data, inference, orchestration, observability, and security layer.

IBM should not try to win raw compute scale or proprietary frontier models. Its viable position is to integrate heterogeneous systems, govern data and models, provide portability, and accept accountability in regulated environments. Open source can improve adoption and reduce client fear of lock-in; it also makes differentiation harder to retain. The new OpenAI and Together AI relationships fit this lane. They do not create exclusive supply.

The European AI Act began new enforcement and transparency requirements on August 2. Tighter governance can increase demand for controls, auditability, model risk management, and experienced integrators. It also raises IBM’s own compliance burden and slows client deployment. The European Commission notice supports a demand tailwind for trusted implementation, not a guaranteed profit pool.

Consulting: excess supply and AI price discovery

Global consulting has low entry barriers at the labor layer and moderate barriers at the account-relationship layer. Accenture’s fiscal Q3 2026 revenue rose 3% in local currency, operating margin was 17%, and bookings of $19.3B were below the prior year’s $19.7B; its full-year growth guide was only 3–4%. That is consistent with a broad market where AI demand is real but discretionary transformation remains cautious. The Accenture release is a useful external check on IBM’s slow Consulting demand.

Generative AI initially expands consulting workloads because clients lack clean data, governance, process design, and integration skills. Over time, it can automate code migration, testing, documentation, and configuration. Supply rises faster than differentiated demand unless a provider owns reusable assets or a captive platform. IBM’s attached Software and mainframe estate give it an advantage in specific accounts, but the industry will likely pass a meaningful share of productivity gains to customers.

Industry conclusion: IBM’s best economics come from a harvested, hard-to-enter installed base; its growth opportunities sit in crowded markets attracting abundant capital; and its largest people business operates in structural oversupply. Durable value depends on using the captive core to deepen switching costs, not subsidize undifferentiated expansion.

4. Competitive Position

IBM’s moat is real but narrow. It is strongest where the cost of error and switching exceeds the cost of the product, weaker where technical standards create interoperability, and weakest where labor can be rebid.

The mainframe moat passes the demand test

A moat should show stable share, customer captivity, and an economic reason a competitor cannot profitably reproduce it. IBM Z passes. Clients run systems where outages, reconciliation errors, fraud, and regulatory failures are far costlier than annual license fees. Applications are intertwined with data schemas, middleware, operational procedures, security rules, and staff knowledge. Re-platforming can consume years before it removes the last system dependency.

The architecture also creates a two-sided reinforcing loop inside IBM. New hardware capacity drives software charges, services, support, and future refreshes; the software estate justifies continuing the hardware platform. This is not an open network effect where more users attract more users, but a closed-system scale economy: IBM alone can spread specialized engineering and support across the installed base.

Q2’s 42% IBM Z decline and 8% Transaction Processing decline do not disprove captivity. They demonstrate transaction timing and cycle sensitivity. Captive customers can postpone incremental capacity and negotiate terms. The proper test is whether workload and installed-base economics persist over a full product cycle, not whether every quarter grows.

Red Hat is a platform advantage, not a monopoly

Red Hat gives enterprises a consistent operating and container layer across on-premises and public clouds. Its certification ecosystem, support, security response, and developer familiarity raise switching cost. The “open” positioning can be especially valuable to customers wary of hyperscaler lock-in. IBM’s global accounts and Consulting reach expand distribution.

But OpenShift competes with managed Kubernetes, hyperscaler-native platforms, VMware alternatives, and other open-source stacks. Customers can multi-cloud without using IBM, and source code availability limits proprietary control. Red Hat’s 11% Q2 growth is strong evidence of product relevance; it is not evidence of unconstrained pricing power. Its moat is narrower than Transaction Processing and depends on execution, ecosystem health, and continued technical neutrality.

Acquisition-built breadth can deepen captivity—or mask weak returns

Apptio, HashiCorp, DataStax, and Confluent expand IBM’s ability to manage cloud cost, infrastructure provisioning, data streaming, and AI-ready data. The strategic map makes sense: control infrastructure, observe cost, move real-time data, govern it, and deploy models. Cross-selling into an installed base could reduce customer acquisition cost and make the suite harder to remove.

The same breadth risks portfolio assembly without product integration. Customers may prefer best-of-breed tools. Open-source communities may resist commercial changes. IBM may report acquired revenue growth while the legacy organic base stalls. Confluent’s Q2 contribution helped Data grow 19% even though total Software was flat organically. Until IBM discloses cohort revenue, margin, retention, and return, strategic adjacency is a hypothesis rather than moat evidence.

Brand and regulation amplify, but do not create, the moat

IBM’s brand conveys longevity, security, accountability, and willingness to support systems for decades. In public-sector and regulated procurement, those attributes reduce perceived implementation risk. DORA designation provides independent evidence that supervisors view IBM as systemically embedded. Yet Microsoft, Oracle, AWS, Accenture, and other designated providers share that credibility. Brand opens doors; switching costs and operating performance retain the customer.

Competitive conclusion: the captive mainframe/Transaction Processing core deserves franchise value; Red Hat deserves a smaller platform premium; Consulting and greenfield AI do not. The investment case fails if those distinctions are blurred into one “AI company” narrative.

5. Growth History and Forward Opportunities

IBM’s reported growth improved after Kyndryl, but acquisitions, currency, and product cycles obscure the organic rate.

Revenue compounded at 4.2%; gross profit grew faster as mix shifted to Software and productivity improved. R&D intensity rose from 11.3% to 12.3%, supporting the argument that IBM has not simply harvested its installed base. Diluted shares grew 4.9%, so per-share participation lagged enterprise growth.

Opportunity 1: recurring Software and delayed-deal conversion

The nearest growth opportunity is not a new market. It is conversion of the existing recurring and transactional Software pipeline. ARR was $24.6B and grew 8% in Q2. Red Hat grew 11%. Management said roughly one-third to 40% of delayed Q2 transactions had closed during the first three weeks of Q3 and almost none were lost. If the remaining transactions close without discounts, Q2 will look like budget timing around scarce hardware rather than demand destruction.

This is also the clearest falsifiable test. Organic Software was flat, failing the prior greater-than-5% hurdle. One quarter above 5% would show normalization; two consecutive quarters above 5%, with ARR holding high-single-digit growth, would repair confidence. If ARR decelerates while transactional revenue stays weak, the delay explanation fails.

Opportunity 2: Red Hat as the neutral enterprise layer

Enterprises rarely move every workload to one public cloud. Cost, latency, sovereignty, regulation, application age, and bargaining power encourage mixed environments. Red Hat can be the supported operating and orchestration layer across those environments. Ansible adds automation; HashiCorp adds provisioning and secrets; Apptio adds technology financial management.

The opportunity is durable because hybrid complexity is not disappearing. The risk is that customers use cloud-native tools where workloads originate and reserve Red Hat for legacy estates, limiting wallet expansion. Growth should be judged by organic revenue, renewal and product-specific ARR—not a broad hybrid-cloud TAM. IBM’s investor-day estimate of a $1.7T hybrid IT opportunity is too broad to establish achievable economics.

Opportunity 3: governed enterprise AI and real-time data

IBM’s most credible AI proposition is not training a frontier model. It is making enterprise data usable, moving it in real time, selecting and governing models, and integrating outputs into critical workflows. Confluent’s streaming platform and DataStax’s data capabilities fit that stack. The Together AI agreement adds an open-model inference path; OpenAI adds premium external models and coding tools; watsonx adds governance and IBM offerings.

This can create pull-through across Software, Consulting, and Infrastructure. It can also commoditize the model layer and leave bargaining power with NVIDIA, hyperscalers, and model providers. IBM must show that integration produces recurring IBM revenue at attractive incremental margins. A growing “book of business” that includes services signings and future commitments is less informative than organic Software revenue and cash.

Opportunity 4: mainframe modernization without mainframe exit

Customers want modern interfaces, languages, analytics, and AI around systems they do not want to replace. IBM can sell modernization that preserves the core. The future dual-architecture processor could run Arm-native Linux applications alongside IBM Z/LinuxONE workloads, reducing friction for developers and expanding addressable applications. AI coding tools can help document and refactor legacy code.

The opportunity contains its own threat. Better tooling can make workloads more portable. IBM wins if modernization increases usage and extends platform life; it loses if it accelerates migration. The February price reaction to AI-assisted COBOL modernization illustrates how quickly the market can switch between those interpretations.

Opportunity 5: quantum optionality

IBM targets at least 1,000 programmable qubits in 2027 and the Starling fault-tolerant system in 2029. The modular cryogenic design and HRL acquisition address packaging, sensing, materials, and silicon-spin capabilities. A proposed U.S. quantum foundry and more than $10B of announced commitment indicate serious scale.

Commercial timing, unit economics, standards, and eventual industry structure are unknowable. Quantum should be treated as funded research with option value, not a near-term revenue forecast. Milestone discipline matters: logical error rates, reproducible workloads, paying customers, external validation, and incremental cash requirements should precede material valuation credit.

Growth conclusion

The plausible medium-term path is 4–5% revenue growth: high-single-digit Software, low-single-digit Consulting, cyclical Infrastructure, and modest mix improvement. Six percent or better requires sustained organic Software growth and cross-sell rather than acquisition accounting. One to two percent is possible if recurring indicators follow the transactional slowdown and AI productivity is competed away. Growth creates value only above the acquisition-inclusive cost of capital and after dilution.

6. Financial Quality

IBM’s financial statements combine strong recurring gross profit, volatile GAAP earnings, multiple cash definitions, financing assets, capitalized software, pensions, and a growing acquisition balance. The company is cash-generative; the analytical task is measuring how much is distributable without buying the next growth increment.

Earnings and margins

Net income from 2021 through 2025 was $5.74B, $1.64B, $7.50B, $6.02B, and $10.59B. The series is not a clean operating trend. In 2022, a pension settlement created an approximately $4.4B after-tax charge. In 2024, another pension-related transaction produced an approximately $3.1B pretax and $2.4B after-tax charge. In 2025, pretax income was $10.33B but the income statement showed a $242M tax benefit. A normalized tax rate is required for earnings-power analysis.

Gross-margin expansion is higher-quality evidence. Consolidated gross margin rose roughly 330 basis points from 2021 to 2025 as Software mix and productivity improved. Q2 2026 margin fell 100 basis points to 57.7%, with Software gross margin down 130 basis points because Transaction Processing weakened and acquired mix increased. Consulting gross margin rose 140 basis points, while Infrastructure fell 310 basis points on product mix. These movements show why group margin depends not just on growth, but on which revenue grows.

Three cash views

The FY2025 gap between IBM-defined free cash flow and strict cash reflects a cash use from Financing-receivable growth/funding, not a runoff. In H1 2026, the direction reversed: the change in Financing receivables released $2.26B into GAAP operating cash. Removing it gives $5.50B of operating cash, versus $5.47B a year earlier. IBM-defined free cash flow was $4.76B versus $4.81B, down approximately 1%.

Both views are useful. Excluding Financing receivables is sensible when evaluating the operating franchise because debt and earning assets support that segment. The strict cash view is useful when testing consolidated distributable cash and funding capacity. Neither alone captures acquisitions, which are classified as investing but are economically the largest reinvestment expenditure.

The retained $15.7B FY2026 guide requires $10.94B of second-half IBM-defined free cash flow, versus $9.89B in H2 2025. That 10.6% increase is possible given seasonal collections and delayed transactions, but the burden is now explicit. The clean H1 series did not pre-fund it.

Balance sheet and accounting quality

At June 30, cash and securities were $8.18B, total debt was $61.99B, Financing debt was $13.05B, and non-Financing debt was $48.94B. Net non-Financing debt was $40.76B. Working capital was negative $7.51B and the current ratio was approximately 0.79. IBM’s recurring revenue, collections, market access, and maturity ladder make a liquidity crisis unlikely, but the acquisition program has consumed balance-sheet capacity.

Goodwill of $74.60B plus acquired intangibles of $13.96B equals $88.55B, 2.6 times reported equity. Tangible equity is negative $54.01B. Negative tangible equity does not mean the franchise lacks value; internally developed code, client relationships, support capability, and R&D are expensed. It does mean book value supplies no loss-absorption or valuation anchor and acquisition impairments could make GAAP equity volatile.

IBM capitalizes some software costs, reports non-GAAP measures excluding acquisition amortization and other items, and records significant pension effects. These are not inherently aggressive choices, but they make reconciliations essential. Acquisition amortization is non-cash in the current period but represents payment for a finite-lived asset. Stock compensation is non-cash in operating cash flow but real through dilution. Pension income or settlements can obscure operating comparisons. The 2025 annual report and Q2 10-Q provide the controlling bridges.

Profitability and ROIC

Using normalized pretax operating income of approximately $12.49B, a 15% normalized tax rate, and a consistent average capital base of debt plus equity less cash/securities yields NOPAT of approximately $10.62B and normalized ROIC near 12.65%. This is an analytical estimate, not IBM’s reported measure. It is respectable, but below the 15% threshold that would provide strong company-level evidence of a durable high-return franchise.

The consolidated figure averages very different businesses and acquisition vintages. Transaction Processing likely earns exceptional returns on a largely expensed asset base; Consulting earns lower returns; recent acquired Software may not yet cover the price paid. Management’s former incentive ROIC modifier excluded Red Hat goodwill, making it unsuitable as an acquisition-inclusive test. The right next disclosure is cohort return using full purchase consideration, integration costs, and stock compensation.

Financial-quality conclusion: IBM produces real cash and has improved gross-profit quality. The principal risks are definition spread, acquisition reinvestment outside free cash flow, a second-half conversion burden, and a balance sheet dominated by goodwill rather than tangible equity.

7. Capital Allocation

IBM’s capital-allocation policy is coherent: fund R&D, buy software capabilities, protect the dividend, maintain investment-grade access, and avoid material repurchases. The unresolved question is whether the acquired capabilities earn enough to justify their price and financing.

Deployment record

From 2021 through 2025, IBM spent $35.63B on R&D, $22.31B on acquisitions net of acquired cash, and $30.26B on dividends. Physical capital requirements were modest: PP&E additions fell from $2.06B in 2021 to $1.09B in 2025. The economic reinvestment rate is nevertheless high because R&D is expensed and acquisitions sit below the free-cash-flow line.

The dividend remains protected. Annual payments have been approximately $6.3B and increases are token rather than growth-oriented. That suits income holders and signals confidence in baseline cash, but it leaves less room for debt reduction after large deals. IBM has not made open-market repurchases since 2019 despite a $4B residual authorization. Avoiding buybacks at elevated prices was sensible; failing to offset employee issuance means enterprise progress leaks away from each share.

Diluted shares increased 4.9% from 2021 through 2025. In H1 2026, diluted shares rose another 0.6% year over year, and the cash-flow statement’s stock-compensation addback was $1.00B, up about 19%. Stock compensation can align employees and preserve cash, but it is not free. The appropriate charge is either a cash-equivalent expense or modeled dilution; excluding it from both creates false owner cash.

Confluent: strategically rational, financially unproven

The Q2 filing gives enough data to repair the headline purchase price. IBM paid $11.54B of cash consideration and settled $1.10B of assumed convertible debt, then acquired $1.17B of cash and $0.92B of securities. The resulting enterprise cash burden was approximately $10.56B. Adding $65M of pre-acquisition award value produces approximately $10.62B of accounting invested capital.

IBM allocated $7.24B to goodwill, $2.12B to client relationships with a 12-year life, $1.59B to technology with a seven-year life, and $122M to trademarks with a five-year life. Those items total $11.07B, or 95.4% of consideration. Only about 1% of goodwill is tax-deductible. Initial annual amortization of finite-lived assets is approximately $428M.

At an 8–9% after-tax hurdle, the $10.62B capital base must produce roughly $850–956M of annual after-tax operating earnings. An $800M contribution would be only about a 7.5% return. Strategic fit does not lower this burden: if real-time data streaming is essential to IBM’s AI stack, the purchase must still create revenue, margin, retention, or avoided cost worth more than the capital consumed.

Confluent may clear the hurdle through stand-alone growth, reduced customer acquisition cost via IBM distribution, cross-selling into Red Hat and watsonx, and better retention of Data customers. It may miss if open-source alternatives pressure pricing, integration slows innovation, sales overlap produces dis-synergies, or IBM purchases revenue that would otherwise have been partnered. Investors need organic revenue, gross margin, retention, cash contribution, and cohort return—not only Data segment growth.

Funding and leverage

At June 30, net non-Financing debt was $40.76B. The August Canadian-note offering raises gross debt and cash by roughly equal amounts at inception, leaving net debt initially unchanged. Its weighted coupon is about 4.34%. “General corporate purposes” is not evidence of acquisition funding; proceeds may refinance maturities, support liquidity, or fund operations.

IBM’s cash generation and market access make leverage manageable. The concern is opportunity cost, not imminent distress. Interest and acquisition amortization reduce the return that common shareholders receive from acquired revenue. Another large transaction before Confluent produces measurable returns would weaken the claim that management is harvesting a captive franchise for owners.

Incentives and insider evidence

The 2025 annual incentive weighted revenue 50% and non-GAAP operating cash flow 50%. Results produced a 140 score, plus 10 points for strategic objectives, for a 150 score; the CEO’s payout was $5.25M. The 2023–2025 performance-share program weighted revenue 40%, operating EPS 30%, and free cash flow 30%, with a relative-ROIC modifier. That modifier was zero for the cycle and, importantly, its ROIC definition excluded Red Hat goodwill.

For the 2025–2027 cycle, relative total shareholder return replaced ROIC as the modifier. TSR alignment is useful, but it does not force management to earn attractive returns on full acquisition consideration. Revenue, adjusted cash, EPS, and relative stock performance can all improve temporarily while acquisition cohorts earn below the cost of capital.

Two years of ownership filings show 177 Form 4s and 350 transactions. Five open-market purchases totaling 3,620 shares were all by directors; three open-market sales totaling 55,143 shares were all by senior vice presidents. There was no CEO or CFO open-market purchase. Robert Thomas’s August 26 sale of 25,000 shares was not identified as a 10b5-1 transaction; Arvind Krishna’s August 27 transaction involved phantom units in an excess savings plan, not an open purchase. Insider records are context, not a standalone signal, but they do not provide conviction confirmation.

Capital-allocation conclusion: the priority order is understandable and the dividend is well covered, yet management has not supplied the acquisition-inclusive scorecard that its strategy demands. The burden has shifted from strategic logic to realized return.

8. Changes and Headwinds — Last Two Years

The last two years contain both structural improvement and a sharp evidence break.

2024–2025: mix improvement and confidence

Software mix, Red Hat growth, productivity, and the z17 cycle improved IBM’s financial profile. FY2025 revenue rose 7.6%, gross margin reached 58.2%, and IBM-defined free cash flow reached $14.7B. The market increasingly viewed the company as a recurring-software franchise with AI and quantum optionality rather than a low-growth hardware and services vendor.

That confidence was not purely narrative. R&D increased, recurring Software expanded, Infrastructure launched a successful system cycle, Consulting protected margins, and gross profit grew faster than revenue. The January 2024 and January 2025 earnings reactions show the tape rewarding operating evidence.

Early 2026: disruption fear and narrative acceleration

In February, AI-assisted legacy-code modernization was interpreted as a threat to the mainframe moat. The concern is conceptually valid but often overstated. Code translation is only one part of migration; data integrity, interfaces, testing, operational procedures, regulation, performance, and accountability remain difficult. Still, better tools can reduce switching cost incrementally and deserve monitoring.

Q1 results were operationally solid: revenue rose 6% at constant currency, Software grew 8%, Infrastructure 12%, Consulting 1%, free cash flow was $2.2B and pretax margin expanded 140 basis points. Software guidance remained at least 10%. The stock nevertheless fell the next day, implying expectations had outrun the result.

From late May through June 2, quantum, open-source, foundry, and technology-investment announcements drove a 46% rally from $225 to an unadjusted $329.23 close. The move did not correspond to a new earnings quarter. It increased the valuation’s dependence on distant optionality and made the subsequent reset more violent.

Q2 2026: the operating break

IBM’s July 14 preliminary letter disclosed a $400–500M shortfall concentrated in Software and Infrastructure. Management said clients redirected scarce capital budgets toward servers, storage, and memory, delaying software transactions. The stock fell 25.2% that day and ultimately 37.5% from the June peak.

The full result confirmed that the problem was mix, not a broad collapse. Red Hat grew 11% and ARR grew 8%, while Transaction Processing fell 8%, IBM Z fell 42%, and total organic Software was flat. Consulting demand stayed weak but margins improved. Company guidance fell to 4–5% constant-currency revenue and 6–8% Software, while the $15.7B cash guide remained.

The bull thesis’s prior test—organic Software above 5% through the hardware down-cycle—failed for this observation. The bear thesis was not confirmed, because recurring growth held and delayed transactions may convert. No post-Q2 operating disclosure is available to adjudicate the question.

Current operating and policy headwinds

1. Enterprise budget competition. AI infrastructure, servers, storage, memory, cybersecurity, and software compete for the same capital envelope. IBM can benefit from several categories while still losing timing within a quarter.
2. Transaction Processing timing. The most profitable pool is captive but not immune to capacity cycles, contract negotiation, and deferred transactions.
3. Consulting deflation. AI expands transformation work and reduces delivery hours. Price discovery determines who keeps the productivity.
4. Acquisition integration. Confluent, HashiCorp, DataStax, Apptio, and other products must become one usable architecture without damaging their communities or sales momentum.
5. Regulatory burden. DORA and AI regulation support trusted vendors but add compliance expense, supervision, concentration scrutiny, and potentially slower deployments.
6. Supply and capital intensity. The Together AI cluster relies on scarce NVIDIA systems. IBM must avoid entering raw infrastructure economics where hyperscalers have far greater scale.
7. Quantum funding. Technical leadership could create a new franchise; premature capital commitments could reduce near-term owner cash without measurable commercial return.

The new August announcements improve strategic plausibility, not near-term earnings visibility. The next earnings filing remains the highest-value evidence event.

9. Risk Analysis

The most realistic downside is not insolvency or rapid obsolescence. IBM is diversified, investment-grade, and supported by mission-critical recurring cash. The realistic loss path is several years of low per-share growth while the market stops paying a software premium. A 1–2% organic company can remain essential to customers and still be a disappointing security if it issues shares and reinvests at low returns.

The catastrophic tail is an operational or security failure that damages confidence in regulated workloads. The moat is built partly on the cost of error; a severe error would attack the reason customers accept captivity. This risk is low-frequency and cannot be captured by near-term valuation multiples.

Risk interactions matter. A cash-guide miss alone is manageable. A miss combined with another large acquisition and ongoing dilution would undermine both balance-sheet and governance confidence. Slow Consulting alone is manageable. Slow Consulting plus weak Software would reveal a company-wide demand problem. A mainframe down-cycle alone is expected. A simultaneous decline in Transaction Processing workload would be more serious.

10. Valuation Discussion — Embedded Expectations

This section contains no price target and no recommendation. It asks what the current capitalization requires.

Live capitalization and multiples

The August 28 close was $235.59. Multiplying by 942.134M June 30 shares gives $221.96B of equity value. Add $48.94B of non-Financing debt and subtract $8.18B of cash and securities to obtain $262.72B of core enterprise value. The post-quarter Canadian notes initially increase debt and cash together, so they do not change net debt at issuance.

The multiple reset is material. It does not make IBM conventionally cheap on sales or tangible assets. A blended enterprise with 45% Software can justify more than a consulting or hardware multiple, but only if the Software growth is organic and the high-margin Transaction Processing pool remains stable.

Own-history percentile comparisons were not used in the conclusion. The underlying vendor history restated prior ranks without corresponding changes in raw ratios, and the authenticated series is not reproducible from the public quote page. The raw ratios are sufficient: price-to-sales around 3.2 times is high relative to the IBM of five years ago because the mix and narrative changed. Whether that change deserves permanence is the analytical question.

Peer framing without false precision

Accenture is the cleanest people-business reference: slower revenue, net-cash economics, and lower structural margins than Software. HPE is a hardware and infrastructure anchor. Cisco combines installed networking, recurring support, and acquisition integration. Oracle has installed-base captivity and faster cloud growth but is spending heavily on capacity. Microsoft and ServiceNow show the valuation ceiling for much faster recurring growth and superior incremental economics.

Exact peer tables can create more precision than comparability. Fiscal dates, lease treatment, stock compensation, capitalized software, acquisition amortization, and net-debt definitions differ. Directionally, IBM’s equity cash multiple is below premium software but its sales multiple is above services/hardware. That gap means mix improvement is already recognized even after the price decline.

Reproduction value and earnings-power value

Reported tangible equity is negative $54.01B. Capitalizing all $22.57B of 2023–2025 R&D with no obsolescence would still leave a mechanical negative $31.44B before assigning value to mainframe code, client integrations, certifications, sales capacity, and global support. Reproduction value therefore cannot be estimated from book assets alone.

The inability to reproduce the franchise cheaply is itself evidence of a moat. A new entrant could spend tens of billions and still lack compatible code, trust, reference customers, and operating history. In Greenwald’s framework, earnings power above adjusted reproduction assets indicates franchise value. IBM clears that qualitative test.

A no-growth earnings-power stress uses $13.5B of normalized levered owner cash—between strict FY2025 cash of $11.46B and TTM/guided IBM-defined free cash flow—and capitalizes it at 9%. The result is $150B of equity earnings power, approximately 32% below the current $221.96B capitalization. This is not fair value and not a target. It deliberately assigns no growth and charges for cash-quality uncertainty. It shows that tangible assets and static earnings do not support the entire quote; profitable franchise growth must supply the difference.

Reverse expectations

Using a simplified perpetuity at a 9% equity capitalization rate, today’s equity value implies approximately 2.4% perpetual cash growth from $14.65B of TTM IBM-defined free cash flow, 1.9% from the $15.7B guide, or 3.8% from $11.46B of strict FY2025 cash. These are not forecasts. They show how much the cash definition changes the apparent hurdle.

The 1.9–2.4% headline looks modest. It becomes less modest if IBM must repeatedly spend acquisition cash to produce it or if diluted shares grow 0.5–1.0% per year. Enterprise cash growth must exceed share growth to create the same per-share result. Conversely, if Q2 was timing and the acquired portfolio cross-sells organically, today’s starting cash yield can carry a reasonable outcome without multiple expansion.

Explicit operating scenarios—not targets

The scenarios intentionally stop short of equity values. They identify load-bearing variables: owner-cash definition, organic Software and ARR conversion, sustainable free-cash-flow margin, acquisition-inclusive ROIC, dilution, and the terminal multiple. The bear does not assume a broken company; the bull requires evidence absent from Q2.

Valuation conclusion: the market correctly prices a durable captive core and no balance-sheet crisis. It may over-credit IBM-defined free cash flow as fully distributable and acquired Data growth as organic. It may under-credit the possibility that a transaction-heavy quarter mean-reverts. The present quote offers an adequate starting yield only if the proof arrives.

11. Variant Perception

The consensus debate is framed too broadly as “legacy IBM versus AI IBM.” The more useful variant separates the protected cash engine from the purchased growth engine.

Prevailing belief

Before July, investors increasingly treated IBM as a scarce incumbent beneficiary of enterprise AI, mainframe renewal, Red Hat, and quantum. The May–June rally capitalized long-duration announcements before new operating evidence. After the Q2 preannouncement, belief shifted to timing versus structural slowdown. The rebound from the July trough and flat total return since August 7 indicate that investors still credit the free-cash-flow guide, but the stock below its 50- and 200-day averages indicates that the prior compounder narrative has not returned.

Strongest constructive interpretation

Q2 was an unusual budget-allocation event. Clients prioritized supply-constrained hardware and deferred, rather than cancelled, high-margin software transactions. ARR accelerated to 8%, Red Hat grew 11%, distributed infrastructure grew 37%, Consulting signings improved, and pretax productivity held. A large share of delayed deals closed early in Q3. The recurring pool returns to near 10% growth, Transaction Processing normalizes, and second-half collections deliver the cash guide.

Under this interpretation, the 7.1% guided cash yield is sufficient. Investors do not need frontier-model dominance or immediate quantum revenue; mid-single-digit enterprise growth, flat-to-modest dilution, and stable margins can compound per-share cash. Together AI and OpenAI validate IBM as the trusted integration layer, while the captive mainframe estate funds the transition.

Strongest skeptical interpretation

The delayed-deal explanation masks a mature organic business. Software was flat organically during an AI spending boom; Data growth was acquired; Transaction Processing declined; Consulting barely grew; and clean H1 cash was flat. IBM spent over $10.5B of enterprise cash on Confluent, allocated most of the price to goodwill and intangibles, and then raised more debt. Shares and stock compensation rise without repurchases. Quantum and AI announcements keep the narrative fresh while return disclosure remains absent.

Under this interpretation, IBM is a serial acquirer using extraordinary mainframe cash to rent reported growth at average returns. The equity does not collapse, but per-share cash grows slowly and the sales multiple converges toward a mixed services/infrastructure business.

The differentiated view

The market may underestimate how durable the captive core remains and overestimate how proprietary IBM’s AI growth is. Those statements can both be true. Mainframe modernization tools are more likely to extend mixed estates than eliminate them quickly, because code translation does not remove data, testing, interfaces, regulation, or accountability. Yet the external-model partnerships also show that IBM is a distributor and integrator within AI, not the owner of the scarce model or compute layer.

The key variant is capital allocation: a stable captive core can generate satisfactory returns at today’s lower valuation if management requires high returns from every incremental dollar. The same core can become a value trap if it subsidizes repeated portfolio assembly. The evidence that distinguishes those outcomes is acquisition-cohort ROIC and per-share cash, not market share in a $1.7T TAM.

Falsification

The constructive case is falsified if organic Software remains below 5% for another quarter, ARR decelerates, the full-year cash guide is reduced, or delayed transactions require discounting and push into 2027. The skeptical case is falsified by two consecutive quarters of organic Software growth above 5%, rising operating cash excluding Financing-receivable changes, and disclosed after-tax returns above 9% on recent acquisition capital. Both sides should be required to name these tests before interpreting the next headline.

12. Fact vs. Interpretation

The table below separates filing facts, management-defined measures, claims, analytical interpretations, and assumptions. That distinction is especially important when no new operating filing has followed the Q2 break.

Facts can coexist with opposing interpretations. Flat organic Software is a fact; whether it reflects timing or demand is unresolved. High goodwill is a fact; whether it represents valuable acquired relationships or future impairment depends on cash return. The discipline is to preserve the distinction until an observable test resolves it.

13. Open Questions

1. Of the $400–500M Q2 Software shortfall, how much closed in Q3, at what price, and how much moved beyond 2026?
2. What are organic growth rates for Hybrid Cloud, Automation, and Data after removing every acquired contribution?
3. Can Software ARR remain at least high-single-digit while Transaction Processing and the transactional pool normalize?
4. What collection, working-capital, tax, pension, and transaction assumptions bridge H1 IBM-defined free cash flow of $4.76B to the $15.7B full-year guide?
5. How much of second-half cash growth will come from underlying earnings rather than timing of receivables, payables, taxes, or restructuring?
6. What were Confluent’s stand-alone revenue, gross margin, retention, and cash contribution at acquisition, and what cross-sell assumptions justified approximately $10.62B of invested capital?
7. When will IBM report acquisition-cohort ROIC including goodwill, assumed debt, integration cost, stock compensation, and incremental interest?
8. Why was the relative-ROIC modifier replaced by relative TSR, and why did the former calculation exclude Red Hat goodwill?
9. When will repurchases or reduced issuance offset employee dilution and make enterprise cash growth translate fully into per-share growth?
10. What is the total capital, depreciation, utilization guarantee, and expected gross margin associated with the Together AI cluster?
11. Does the OpenAI arrangement include minimum commitments, preferred economics, implementation exclusivity, or only a joint go-to-market framework?
12. Will the dual-architecture processor expand IBM Z/LinuxONE workloads or make application portability away from IBM easier at the margin?
13. How much of the announced quantum and foundry commitment is incremental to existing R&D and capital expenditure, and what commercial milestones gate each tranche?
14. What price did IBM pay for HRL Laboratories, what liabilities were assumed, and when should its capabilities produce measurable revenue or technical advantage?
15. Can Consulting retain AI productivity in margin, or will competitive bidding pass it to clients through lower rates and fewer billable hours?
16. How much underlying workload growth, rather than pricing or capacity transactions, supports the long-term Transaction Processing pool?
17. What operational remediation and incremental cost follow IBM’s DORA critical-provider designation?
18. Will management disclose net retention, remaining performance obligations, or another contracted measure that makes ARR quality more comparable with software peers?

These questions are ordered by near-term decision value. Q3 conversion and cash matter more than distant quantum milestones; acquisition return matters more than broad market size.

14. What Must Be True — Constructive and Skeptical Cases

Constructive case: required conditions

1. Organic Software exceeds 5% for two consecutive quarters, showing that Q2 was timing rather than an underlying stall.
2. ARR remains at least high-single-digit and delayed transactional revenue closes without material discounts or unusual contract concessions.
3. FY2026 constant-currency revenue remains within 4–5% and IBM produces approximately $15.7B of free cash flow without relying on another large Financing or working-capital release.
4. Red Hat sustains near-double-digit growth while Automation and Data improve organically after stripping out acquisition contribution.
5. Productivity produces durable pretax and free-cash-flow margin expansion after acquisition amortization, integration cost, and stock compensation.
6. Confluent and the recent acquisition cohort generate disclosed after-tax returns above 8–9% on full invested capital.
7. Share growth slows materially or repurchases offset employee issuance, allowing enterprise growth to compound per share.
8. IBM uses external AI models and compute to deepen governance, data, and workflow captivity without accepting hyperscaler-like capital intensity.

Constructive-case falsification: another quarter of organic Software below 5% accompanied by ARR deceleration; a full-year cash-guide reduction; or evidence that delayed deals moved into 2027 or required economic concessions would break the timing explanation. A large new acquisition before Confluent return disclosure would weaken the allocation premise even if revenue growth held.

Skeptical case: required conditions

1. Q2 exposed a 1–3% organic company whose reported growth increasingly depends on acquisitions and hardware cycles.
2. ARR eventually follows transactional weakness, showing that annualization lagged the demand change.
3. H2 cash conversion falls short after adjusting for Financing receivables and working-capital timing.
4. Confluent and other acquired products grow revenue but fail to clear the cost of capital after purchase price, integration, amortization, interest, and dilution.
5. AI reduces Consulting hours faster than IBM can create proprietary reusable assets, passing productivity to clients.
6. Mainframe modernization gradually reduces workload captivity, even if the installed estate remains important for many years.
7. Management continues to issue equity and protect the dividend without repurchases, suppressing per-share compounding.
8. Technology announcements support valuation sentiment but do not create material, high-margin recurring revenue.

Skeptical-case falsification: two consecutive quarters of organic Software above 5%, consolidated organic growth near 5%, rising operating cash excluding Financing-receivable changes, stable ARR, and acquisition-inclusive ROIC above 9% would invalidate the serial-acquirer thesis. Evidence that modernization expands IBM Z workloads and that AI Consulting margins remain durable would further weaken it.

Decision dashboard

The dashboard prevents narrative drift. A press release can inform a row, but only operating and return data can close it.

15. Source Appendix

Primary filings control financial facts; company presentations and calls supply management-defined measures and forward claims; regulator and competitor releases provide external context; third-party feeds are used only for market and factor data. All links were accessed through August 30, 2026.

IBM filings and governance

• IBM 2025 Form 10-K, filed February 24, 2026 — audited operations, risks, segments, debt and annual financial statements.
• IBM 2025 Annual Report, filed March 10, 2026 — five-year financials, cash reconciliation, R&D, capitalized software, pension and segment detail.
• IBM Q2 2026 Form 10-Q, filed July 23, 2026 — current statements, acquisition accounting, debt, cash, segments and guidance context.
• IBM Q2 2026 earnings exhibit, filed July 22, 2026 — segment KPIs, organic growth, ARR, revised guidance and free-cash-flow reconciliation.
• IBM Q1 2026 earnings exhibit, filed April 22, 2026 — prior-quarter operating baseline and original guidance.
• IBM 2026 proxy statement, filed March 10, 2026 — incentive design, payout results, governance and ownership.
• IBM 2023 Form 10-K, filed February 26, 2024 — historical pension and five-year context.
• Canadian-notes prospectus and underwriting 8-K, filed August 11 and 14, 2026 — principal amounts, coupons, maturities and proceeds.
• IBM filing and ownership index — five-year filing sweep and two-year ownership review.
• Robert Thomas Form 144, Robert Thomas Form 4, and Arvind Krishna Form 4, filed August 26–28, 2026 — post-baseline ownership transactions.

Results, strategy and post-baseline events

• Q2 2026 earnings page and transcript and Q1 2026 earnings page — management explanation, delayed transactions, ARR and product-cycle commentary.
• 2025 Investor Day charts — Red Hat components, long-term architecture and company-defined market opportunity.
• Preliminary Q2 investor letter, published July 14, 2026 — initial shortfall and capital-budget explanation.
• Together AI agreement, published August 11, 2026 — contract amount, architecture and timing.
• OpenAI partnership, published August 13, 2026 — joint practice and product integration.
• Modular cryogenic milestone, dual-architecture processor, and HRL acquisition completion, published August 19–26, 2026 — technical roadmap changes.
• IBM/Red Hat Lightwell commitment, quantum investment roadmap, and proposed U.S. quantum foundry — historical event and capital-commitment context.

Industry, competitors and regulation

• DORA critical-provider list and European supervisory methodology — independent evidence of financial-system embeddedness and oversight.
• European Commission AI Act enforcement notice and European supervisors’ governance statement — current policy context.
• Microsoft FY2026 Q4 metrics, Amazon Q2 2026 results, Oracle FY2026 results, and Accenture Q3 FY2026 results — primary competitor growth, margin and capital-cycle evidence.

Market and risk data

• AZI IBM daily-price series, data through August 28, 2026 — adjusted and unadjusted history, event returns, moving averages and drawdown. Filing share and debt data replace vendor capitalization fields.
• FactorsToday methodology and IBM endpoints — factor loadings, explained variance, related securities and specific volatility. Used only for market-positioning context.
• Reuters report on the February AI-modernization reaction — secondary event attribution, not fundamental evidence.

Prepared as research publication. Outside the clearly labeled Claude’s Take block, this memo contains no investment recommendation or price target. Figures may not sum because of rounding.