KEY READ-THROUGHS FROM IBM’S Q3 2026 EARNINGS CALL

IBM’s results indicate that enterprise AI spending is expanding beyond model experimentation into application modernization, data infrastructure, security and production deployment. The distribution of economic benefits remains uneven. OpenShift ARR growth above 30%, strong storage demand and the recovery in mainframe software support infrastructure and mission-critical software franchises. Conversely, Consulting growth of only 2%, despite AI representing approximately 30% of revenue, challenges expectations that AI adoption automatically accelerates services revenue and profitability. The most actionable implications concern competitive redistribution and revenue quality rather than a broad technology spending upgrade: virtualization alternatives are gaining commercial scale, identity security is becoming an essential deployment requirement, and private AI creates opportunities outside hyperscale infrastructure. Magnitudes below describe incremental business implications, not predicted share-price moves. Revenue growth figures are at constant currency unless otherwise stated. Quantum cost comparisons are excluded from directional peer conclusions because the disclosed benchmarks do not establish comparable commercial economics.

IT SERVICES: AI ADOPTION DOES NOT YET SUPPORT A BROAD REVENUE OR MARGIN INFLECTION (READ-THROUGH 1)

Affected companies: Accenture (ACN: Ireland), Cognizant Technology Solutions (CTSH: United States), and Infosys (INFY: India). Directional impact: negative for near-term expectations of an AI-led sector acceleration; moderate for labor-linked delivery economics, but insufficient by itself to justify company-specific earnings reductions.

The critical evidence is the disconnect between AI penetration and financial performance. IBM reported that AI represented approximately 30% of Consulting revenue and 35% of backlog, yet Consulting revenue grew only 2%. The accompanying presentation disclosed signings declining 2%, a trailing-12-month book-to-bill ratio of 1.04, and segment profit declining approximately 5%. Consulting gross margin fell 190 basis points to 27.4%. Management attributed segment margin pressure to reinvestment in digital assets and the transition toward a services-as-software model.

The transmission mechanism is that AI-related work can replace or reshape existing services revenue rather than simply add to it. Application modernization, data preparation and workflow redesign generate demand, but automated development and delivery reduce the labor required for established tasks. Under time-based contracts, fewer delivery hours can reduce revenue. Under fixed-price contracts, productivity can improve margins, but renewals and competitive bidding allow customers to reclaim some of those savings. Simultaneously, providers must invest in reusable software, technical talent and implementation capabilities before achieving sufficient scale.

For Accenture, the positive opportunity in transformation and workflow redesign remains substantial, but IBM’s results argue against treating AI bookings as an immediate proxy for incremental revenue or earnings. For Cognizant and Infosys, the important risk is that improving demand for AI-enabled transformation coexists with pressure on application development, maintenance and other labor-linked revenue pools. IBM does not establish that these companies will suffer identical margin outcomes, but it provides direct evidence that rising AI exposure alone is not an adequate financial thesis.

Near-term trading relevance centers on subsequent peer disclosures of organic growth, booking conversion, utilization, pricing and margins. The longer-duration shift, over approximately 12-36 months, is toward asset-based and outcome-based delivery. The likely differentiation is the ability to retain productivity gains and monetize reusable capabilities, rather than the proportion of engagements labeled AI.

INFRASTRUCTURE SOFTWARE: VIRTUALIZATION ALTERNATIVES ARE BECOMING A MATERIAL COMPETITIVE CHANNEL (READ-THROUGH 2)

Affected companies: Broadcom (AVGO: United States) and Nutanix (NTNX: United States). Directional impact: negative for VMware’s long-term retention and pricing flexibility; moderate at the VMware franchise level and substantially smaller at consolidated Broadcom. For Nutanix, the negative implication is increased competition for migration opportunities, not evidence of declining absolute demand.

IBM disclosed nearly $800 million of virtualization contracts signed since early 2024 and approximately $750 million of pipeline. The cumulative contract figure increased from approximately $680 million disclosed in Q2. OpenShift ARR reached $2.4 billion, with growth accelerating above 30%, while Red Hat bookings growth returned to the high teens.

The important transmission mechanism is the emergence of a credible alternative procurement and deployment path. As customers evaluate virtualization renewals, Red Hat can combine virtual machines, containers, enterprise support and application modernization within a broader platform decision. This creates a more credible alternative than a migration motivated solely by license cost. For VMware, the consequences can include reduced renewal scope, slower expansion, stronger customer negotiating leverage and eventual workload migration.

The figures do not establish that every Red Hat contract displaced VMware, and cumulative contract value is not annual recurring revenue. Nevertheless, the scale of signed business and prospective pipeline makes the competitive channel harder to dismiss as exploratory customer interest. Switching costs remain high, but the call suggests that a commercially meaningful alternative ecosystem is developing.

Nutanix benefits from the same customer willingness to reconsider incumbent infrastructure, but it cannot be assumed to capture that opportunity without competition. Standalone OpenShift Virtualization competes for some of the same migration budgets. Conversely, OpenShift can operate on Nutanix infrastructure, making the relationship partly complementary. The high-conviction negative read-through is therefore to an assumption of uncontested migration share, rather than to Nutanix’s overall growth outlook.

Near-term relevance is primarily competitive sentiment and scrutiny of renewal, expansion and migration disclosures. The fundamental impact should unfold over approximately 6-24 months as enterprise agreements expire and migrations proceed. Broadcom’s diversified semiconductor exposure and pricing actions can offset VMware customer losses, so this is not a standalone consolidated Broadcom short thesis.

MAINFRAME SOFTWARE: INSTALLED-BASE DURABILITY SUPPORTS BROADCOM’S RECURRING CASH FLOWS (READ-THROUGH 3)

Affected company: Broadcom (AVGO: United States). Directional impact: positive; moderate for the durability of the mainframe software franchise, but low for consolidated earnings expectations based on this call alone.

IBM’s Transaction Processing revenue grew 10%, reversing a 9% decline in Q2. Management reported that customers representing approximately 85% of installed mainframe capacity were maintaining or increasing capacity. It also cited approximately 18% growth in installed MIPS during the z17 program and more than $1 billion of cumulative incremental hardware revenue versus the prior program. Krishna characterized the earlier software weakness as “a timing question, not a destruction question.”

Broadcom’s mainframe software portfolio operates within the same underlying computing environment. Continued mainframe utilization supports demand for security, workload automation, systems management and other software required to operate that environment. The primary transmission is preservation of the installed revenue base, continued renewal relevance and potential expansion as workloads grow. It is not an assumption that IBM’s Transaction Processing growth translates directly into equivalent Broadcom growth.

The distinction between hardware shipments and installed workloads matters. New hardware creates an opportunity, but software monetization depends on deployment, usage, contract structures and the products attached to those workloads. IBM’s evidence of maintained or growing installed capacity is consequently more relevant to Broadcom’s recurring software economics than the headline size of the hardware launch.

Management also cautioned against extrapolating Q3’s Transaction Processing growth. Krishna stated, “I don’t expect double digit performance in TP to keep going.” Part of the improvement reflected deferred transactions. The strongest peer implication is reduced risk of accelerated mainframe abandonment, not a forecast of sustained 10% growth across mainframe vendors.

Near-term relevance is favorable sentiment toward the resilience of legacy infrastructure software and subsequent renewal disclosures. The more important effect is longer-duration cash flow durability over approximately 1-3 years. This positive exposure partially offsets Broadcom’s virtualization competitive risks, demonstrating why infrastructure software should not be treated as a single homogeneous business.

CYBERSECURITY: AGENT IDENTITY CREATES A SPENDING REQUIREMENT THAT DOES NOT DEPEND ON EMPLOYEE GROWTH (READ-THROUGH 4)

Affected companies: Palo Alto Networks (PANW: United States) and SailPoint (SAIL: United States). Directional impact: positive; moderate longer-duration opportunity for identity security, with limited immediate earnings implications from IBM’s disclosure alone.

Krishna stated that, as agents multiply, enterprises need to know “who each agent is, what authority it has and what it can access.” IBM described capabilities to register, authorize, audit and revoke agents and secure their credentials through HashiCorp Vault. It also disclosed collaboration involving IBM Agent Identity, HashiCorp Vault and NVIDIA’s Open Agent Safety platform. Separately, cybersecurity revenue within Consulting grew at nearly 10%.

The transmission mechanism is a shift from predominantly human identity management toward governance of software entities that can take actions. An agent interacting with financial systems, customer data or development environments needs authentication, scoped permissions, credential management, monitoring and revocation. Production deployment therefore creates security requirements even when the number of employees is flat or declining.

For Palo Alto Networks, the relevant exposure is its CyberArk identity-security business, acquired in February 2026. Privileged access, machine identities and credentials are directly adjacent to the controls IBM described. The opportunity is additional product adoption and platform attachment as customers move agents from pilots into sensitive operating environments.

For SailPoint, the exposure is identity discovery, governance, policy enforcement and access lifecycle management. Its agentic identity capabilities address the need to identify agents, determine ownership, govern permissions and remove access when circumstances change. The potential expansion is in the number and complexity of governed identities, rather than simply workforce seats.

The competitive qualification is important: IBM is itself a supplier in this market. The call validates the spending requirement, not the market share of Palo Alto Networks or SailPoint. Agent proliferation also does not translate mechanically into proportionate revenue because pricing may be platform-based rather than charged individually per agent.

Near-term trading relevance is corroboration of agent-security demand ahead of vendor results and customer adoption disclosures. The fundamental opportunity is more likely to emerge over approximately 12-36 months through larger deployments, broader product attachment and renewal expansion. Identity security appears better positioned than discretionary AI experimentation because production access controls become necessary once agents receive meaningful authority.

ENTERPRISE AI INFRASTRUCTURE: PRIVATE INFERENCE BROADENS THE CUSTOMER BASE FOR INTEGRATED SYSTEMS (READ-THROUGH 5)

Affected companies: Dell Technologies (DELL: United States), Hewlett Packard Enterprise (HPE: United States), and NVIDIA (NVDA: United States). Directional impact: positive; moderate for the enterprise AI systems opportunity at Dell and HPE, and a smaller incremental confirmation of NVIDIA’s broader platform demand.

The call linked rising interest in sovereign AI to control over proprietary data, code, models and logs. Krishna stated that sovereign AI was becoming “a much larger conversation” and described enterprises evaluating private deployment rather than relying exclusively on external frontier-model services. He also argued that some large-enterprise deployments could begin with approximately 2-3 racks rather than hyperscale-sized installations.

The direct transmission to Dell and HPE is demand for integrated systems that reduce the complexity of operating private inference. The required purchase extends beyond GPUs to servers, networking, storage, deployment software, security and support. Enterprises without specialized infrastructure teams have a reason to pay for validated configurations and implementation services rather than assemble each component independently.

This exposure is supported by existing product relationships. Dell offers Red Hat OpenShift on its AI Factory with NVIDIA, while HPE’s Private Cloud AI and AI Factory offerings incorporate NVIDIA technologies and enterprise deployment capabilities. These are established routes through which the architecture described by IBM can become hardware and services revenue. IBM did not, however, disclose orders awarded to either company.

NVIDIA benefits because open-weight models and private inference can run on its hardware and software ecosystem. Model openness is not synonymous with open-source replacement of accelerators. The relevant opportunity includes GPU adoption, networking and enterprise software attachment across a broader set of customers.

The call’s inference-efficiency claims should not be interpreted as an equivalent reduction in aggregate accelerator demand. Lower cost per token can reduce hardware required for a fixed workload, but can also make additional workloads economical. The high-conviction conclusion is that private AI expands deployment options and the addressable customer base; the net demand elasticity cannot be quantified from this call.

Near-term relevance is supportive for enterprise AI order commentary and pipeline conversion over the next several quarters. The fundamental shift spans approximately 1-3 years as inference becomes more distributed. The overseas banking example supports demand from regulated and non-US enterprises, but does not establish a specific country-level revenue opportunity. Hardware gross margins and services attachment remain essential because AI systems revenue is not interchangeable with attractive profit growth.

MEMORY AND STORAGE: THE CLEANER READ-THROUGH IS TO ENTERPRISE FLASH DEMAND, NOT UNIVERSAL STORAGE OEM SHARE GAINS (READ-THROUGH 6)

Affected companies: Kioxia Holdings (285A: Japan) and Micron Technology (MU: United States). Directional impact: positive; low-to-moderate confirmation for Kioxia’s enterprise flash exposure and a smaller consolidated read-through for Micron, whose results also depend heavily on DRAM and HBM.

IBM reported 32% growth in Distributed Infrastructure and record storage revenue growth. Management described demand across Flash, Fusion, tape and IBM Z storage, stating that “AI continues to increase demand for storage.” A sizable distributed infrastructure backlog supported management’s confidence in continued Q4 momentum.

The transmission mechanism is the storage required to retain, access and repeatedly use enterprise data in production AI systems. Enterprise inference and agentic applications can increase requirements for persistent datasets, retrieval infrastructure, checkpoints, logs and selected caching layers. Flash can benefit where latency, throughput, density and endurance requirements justify SSDs over slower storage tiers.

Kioxia and Micron both supply enterprise SSD technologies addressing these requirements. Kioxia’s enterprise SSD portfolio includes products positioned for AI environments and context caching. Micron similarly has a direct data-center SSD exposure. The read-through is to the category of demand served by these products, not to a verified IBM supply contract. No incremental supplier award or procurement allocation was disclosed on the call.

This distinction makes the upstream interpretation cleaner than an indiscriminate upgrade of competing storage-system vendors. IBM simultaneously claimed market-share gains. Its strong storage performance can therefore reflect both expanding end demand and competitive displacement. A rival storage OEM may benefit from the category while losing share within it.

Several limitations constrain the magnitude. Distributed Infrastructure includes more than storage, and storage includes technologies other than NAND flash. Revenue growth also does not establish equivalent bit growth because pricing and product mix matter. IBM’s result is consequently corroborating evidence for enterprise flash demand, not a basis for a numerical NAND pricing or supplier earnings revision.

Near-term relevance is positive context for the next round of memory and enterprise storage results. The fundamental implication is broader storage intensity as AI reaches production over approximately 6-24 months. The call provides no sufficient basis to extrapolate the same demand strength to consumer NAND or to infer a specific benefit for HBM.

DATA PLATFORMS: STREAMING CAN EXPAND ANALYTICAL COMPUTE WITHOUT REQUIRING WAREHOUSE DISPLACEMENT (READ-THROUGH 7)

Affected company: Snowflake (SNOW: United States). Directional impact: positive for platform relevance and potential analytical consumption; modest near-term significance and moderate longer-duration opportunity.

IBM reported 20% Data revenue growth and approximately 20% Confluent bookings growth. More importantly for the competitive interpretation, Krishna explicitly said that Streamhouse “complements warehouses and lake houses.” The call also described lakehouse growth and emphasized the need to make current enterprise data available to AI applications without creating unnecessary copies.

The transmission mechanism is that streaming infrastructure can make additional operational datasets available for analytical processing. Data that was previously difficult or expensive to move from operational systems can become accessible to governance, analytics and AI workflows. Where Snowflake performs that processing, broader data availability can increase query activity and compute consumption.

This is not merely a theoretical compatibility argument. Snowflake documents a Confluent Tableflow integration in which Kafka data is materialized into Apache Iceberg tables and exposed to Snowflake. That provides a concrete technical route from greater adoption of streaming infrastructure to additional workloads accessible through Snowflake.

The non-consensus distinction is between the storage layer and the processing opportunity. Open formats and reduced copying can weaken dependence on proprietary data storage without eliminating demand for analytical compute. Snowflake’s opportunity is therefore to monetize useful processing over accessible data, not necessarily to maximize the number of proprietary copies held within its platform.

The positive interpretation remains bounded. IBM’s 20% Data growth includes acquisition effects and is not a clean measure of organic industry demand. Confluent bookings do not automatically create Snowflake revenue, and some processing can remain in streaming engines or move to competing platforms. The call nevertheless provides direct evidence against a simplistic assumption that IBM’s streaming strategy requires warehouse replacement.

Near-term trading relevance is primarily architectural reassurance rather than evidence of a specific quarterly consumption beat. The fundamental opportunity develops over approximately 12-24 months as customers connect operational data to production analytics and AI workflows. Incremental compute consumption, rather than headline data volume alone, is the relevant confirmation.

ENTERPRISE APPLICATIONS: AI REINFORCES THE VALUE OF SYSTEMS OF RECORD MORE THAN IT VALIDATES THEIR REPLACEMENT (READ-THROUGH 8)

Affected companies: SAP (SAP: Germany) and Oracle (ORCL: United States). Directional impact: positive for installed-base durability and the persistence of core application economics; low immediate earnings impact and moderate longer-duration defensive significance.

Krishna’s statement that “you’re not likely going to throw out your ERP system” was central to the discussion of enterprise AI adoption. He argued that customers need workflow redesign and integration with existing systems rather than isolated experiments. The mainframe discussion reinforced the same distinction: mission-critical transactions require deterministic, repeatable and reliable execution even as AI becomes more capable.

The transmission mechanism is that an AI interface does not replace the authoritative business processes beneath it. Finance, procurement, inventory, order management and other core workflows still require validated permissions, transaction integrity, auditability and consistent business logic. An agent can initiate or assist a process, but the underlying system must execute and record the result correctly.

For SAP, this supports the durability of its ERP installed base and the opportunity to embed AI into existing business processes. Customers may change user interfaces or automate tasks without replacing the core application environment. That reduces the credibility of disruption scenarios that assume an AI front end removes the economic relevance of ERP.

For Oracle, the positive exposure spans enterprise applications and the database infrastructure supporting transactional systems. AI adoption can create additional integration and data-access requirements while preserving the need for reliable underlying processing. This read-through is separate from Oracle’s cloud infrastructure capital cycle and does not resolve the economics of that investment program.

The protection is stronger for core business logic and transaction infrastructure than for every application seat. Automation can still reduce human interaction with software, pressure seat-based pricing and shift spending toward usage-based contracts. Incumbent durability therefore does not guarantee unchanged monetization or accelerated revenue growth.

Near-term relevance is reduced support for an indiscriminate AI-displacement trade against enterprise applications. The longer-duration implication, over approximately 2-5 years, is that AI is more likely to change how systems of record are accessed and monetized than to eliminate their underlying function. The most defensible exposures are the platforms controlling authoritative data, permissions and transactions, rather than software whose value resides primarily in a replaceable user interface.