IBM Q3 Earnings Rise as Krishna Calls AI a Tailwind

IBM reported a 5% increase in quarterly revenue, giving CEO Arvind Krishna support for his argument that artificial intelligence is helping the business. IBM Q3 earnings included $17.1 billion in sales and adjusted earnings of $2.95 per share. Both figures exceeded analyst expectations, according to Bloomberg. The question is how much growth reflects expanding AI demand versus the recovery of previously delayed purchases.

That distinction matters because AI spending can work in opposite directions for the same supplier. It creates demand for software and implementation services, while competing for customers’ technology budgets. IBM’s latest results suggest those forces are producing an uneven recovery rather than a uniform lift across its businesses.

For enterprise buyers, developers, and competitors such as Accenture, the report offers a useful view of where spending is landing. Software improved, infrastructure recovered, and consulting remained comparatively subdued. The central story is the relationship between IBM’s AI ambitions and customers’ willingness to fund deployment alongside their other obligations.

IBM Q3 Earnings Show Where Growth Returned

IBM’s quarter establishes a sales recovery, but the headline figures do not isolate AI’s contribution.

The company announced its results on October 21 for the quarter ended September 30, 2026. Its quarterly financial results show software revenue rising 8%, infrastructure increasing 4%, and consulting remaining flat as reported. That distribution matters more than treating every dollar of company growth as evidence for the same investment theme.

The earnings figure also needs its proper label. IBM’s $2.95 per share represents adjusted operating earnings, which exclude specified items from standard accounting results. Earnings under generally accepted accounting principles, the standard reporting framework, were $2.31 per diluted share. Readers comparing the announcement with forecasts should keep those measures separate.

Currency adds another distinction. IBM reported 6% revenue growth at constant currency, a calculation that removes exchange-rate changes. That measure helps assess underlying business activity across countries. It does not identify which products customers bought because of AI, or establish whether those purchases represent new budgets.

For anyone evaluating IBM software growth, the immediate takeaway is that software led the reported expansion. However, a broad business segment contains different products, purchasing cycles, and reasons for renewal. Its performance provides evidence of commercial momentum without supplying a complete explanation for that momentum.

An earnings beat answers a narrower question: did the company perform better than analysts expected? It does not automatically answer whether demand has changed permanently. Expectations can fall after a disappointing period, and delayed transactions can improve a later quarter when they finally close.

The useful interpretation therefore separates execution from demand creation. Closing contracts demonstrates that a supplier can move business through its sales process. Generating additional demand means customers are expanding their commitments beyond purchases they already intended to make. Those outcomes can occur together, but they should not be assumed equivalent.

This distinction is especially relevant when management attributes progress to a broad technology shift. AI can influence a sale without appearing as a separately identified product purchase. A customer might upgrade supporting software, reorganize data access, or commission integration work. Financial reporting needs more detail before readers can assign a precise contribution.

IBM’s results give its management a stronger starting point for that discussion. They also give customers a reason to examine the business beneath the headline. The next question is why the recovery arrived now, particularly after earlier spending decisions had worked against IBM.

Delayed Deals Explain Part of the Rebound

Recovering postponed purchases strengthens IBM’s execution story without establishing an entirely new demand cycle.

Bloomberg’s account of the infrastructure deal recovery adds an important explanation. Chief Financial Officer Jim Kavanaugh said roughly two-thirds of previously delayed mainframe and associated software deals had since been signed. Mainframes are enterprise computers designed to handle demanding transaction workloads.

That statement helps connect the latest quarter with IBM’s earlier difficulties. Bloomberg reported that customers had shifted spending toward chips and servers amid AI-related shortages. Those priorities had hurt IBM’s mainframe and software businesses. The same technology trend supporting management’s growth narrative had also complicated the timing of customer purchases.

The company’s second-quarter operating update had already described revenue headwinds and changes to its sales approach. IBM said it was expanding customer coverage and investing in specialized technical talent. Management presented those steps as ways to capture demand as AI moved into broader enterprise deployment.

Taken together, the reports support a specific interpretation. Some customers still wanted IBM products but changed when they purchased them. Recovering those contracts reduces concern that the business disappeared altogether. It does not reveal how much additional business will follow after that delayed demand clears.

Consider the difference between a postponed equipment purchase and a newly approved software project. The first can improve current revenue while leaving the customer’s original spending plan largely intact. The second expands the supplier’s opportunity. An earnings report can contain both, although the headline growth rate does not distinguish them.

For IBM’s sales organization, this creates pressure to show that execution improvements extend beyond recovering missed transactions. A completed contract is useful evidence. A repeatable ability to attract and expand customer commitments would support a broader growth argument. The latest quarter should be assessed as progress along that path.

For buyers, the timing issue is equally practical. A technology roadmap can remain unchanged while procurement schedules move because another purchase becomes urgent. Suppliers then face volatile quarterly results even when customers maintain their longer-term intentions. That makes contract timing important without making it a complete explanation for business performance.

The rebound also changes how readers should interpret Krishna’s AI assessment. AI demand can support IBM over a longer period while temporarily redirecting money away from particular offerings. A favorable overall trend does not guarantee that every product benefits during every reporting period.

That is the tension at the center of the quarter. IBM has evidence that postponed business is returning. The stronger claim, that AI is generating sustained expansion across its portfolio, requires a clearer view of what customers fund after those transactions close.

AI Demand Does Not Lift Every Business Equally

The report places pressure on IBM to translate software momentum into broader customer spending.

Software and consulting occupy different positions in a customer’s purchasing process. Software provides a product or capability that an organization can license and deploy. Consulting provides expertise and implementation work around business or technology changes. Demand for one does not require spending on the other to increase simultaneously.

A customer can expand a software agreement while limiting outside implementation work. Another can hire advisers before approving a broader deployment. A third can use its existing technical staff to introduce new capabilities. These are possible purchasing paths, rather than explanations IBM has established for every account.

The distinction helps explain why IBM AI strategy deserves scrutiny at the business level. Management’s argument concerns enterprise adoption across its portfolio. The financial evidence arrives through segments that include work extending beyond AI. Connecting the narrative to results requires attention to what each business actually delivers.

Consulting is particularly relevant because enterprise deployment often involves organizational decisions alongside technical ones. Someone must determine which process changes, who owns the system, and how employees handle exceptions. Those tasks create potential demand for services, but their existence does not guarantee that customers will purchase them from IBM.

Accenture provides a useful comparison. Its fiscal fourth-quarter results reported 6% consulting revenue growth in U.S. dollars for the period ended August 31. IBM’s reporting period ended a month later. The companies also have different business mixes, so the comparison does not establish a direct market-share change.

It does show why broad statements about enterprise services demand need qualification. Buyers are not responding identically across suppliers and categories. For IBM, stronger software performance does not remove the need to explain the services opportunity. For rivals, IBM’s recovery signals that existing enterprise relationships remain commercially relevant.

The pressure falls on suppliers to make their proposals specific. A buyer should be able to identify the workflow being changed, the supporting technology, and the work required. A broad promise about AI adoption offers less purchasing guidance than a defined project with responsibilities and acceptance criteria.

Developers face a related issue. A model, the software component that generates predictions or content, is only part of a deployed application. The application also needs access controls, dependable data, evaluation, and ongoing maintenance. Those requirements help explain the commercial opportunity around AI without proving which supplier will capture it.

For knowledge workers, the same distinction appears at a smaller scale. An organization can buy an assistant without resolving how employees find reliable information. Effective knowledge blending depends on connecting relevant material to the work being performed. Buying a capability and incorporating it into daily decisions are separate accomplishments.

The implication is a narrower, more useful reading of IBM software growth. It indicates that customers are spending within a key business. Whether that spending expands into implementation, wider usage, and durable customer commitments remains a question for subsequent disclosures.

Customer Deployments Make the AI Strategy Concrete

A deployed workflow offers stronger evidence of practical adoption than a general statement about AI opportunity.

IBM’s banking assistant deployment at NatWest illustrates what enterprise adoption can involve. According to the vendor’s case study, NatWest developed Cora+ using IBM watsonx Assistant. IBM says the project progressed from an initial demonstration into production, with work spanning technology and business teams.

The system uses retrieval-augmented generation, a method that supplies relevant source material to an AI model before it produces an answer. IBM describes Cora+ as answering customer questions within the conversation instead of simply directing people toward existing pages. The case study also discusses controls intended to support accuracy and trust.

This is vendor-published evidence, and its scope matters. It describes a particular implementation rather than establishing typical outcomes for every customer. It also does not disclose a contribution to the latest quarter’s revenue. Its value here is showing the kind of operational work behind IBM’s enterprise AI argument.

The example makes the commercial mechanism easier to understand. An organization needs more than generated text when an answer concerns a customer’s financial affairs. It needs approved information, appropriate access, and a process for dealing with uncertainty. Those needs create work around the assistant itself.

A useful analysis separates the interface from the underlying service. The interface is what a customer sees when asking a question. The underlying service determines which information the system can access, which actions it can take, and when a person should intervene. Improving one layer does not automatically resolve the other.

That distinction also creates a more demanding definition of adoption. Launching an experiment demonstrates that a team can assemble a working system. Operating it in a customer setting requires ongoing ownership. Expanding it requires evidence that the system remains useful when requests become less predictable.

For IBM AI strategy, the opportunity lies partly in connecting those requirements. Software, technical integration, and operational expertise can support the same project. However, a supplier must still show why a customer should purchase each component from it. Existing relationships help frame that conversation without deciding its outcome.

Buyers should read such case studies for their implementation details rather than assuming the featured customer represents a universal result. The most useful details concern the original problem, the information involved, the deployment process, and the limits. Marketing claims become more informative when tied to those concrete features.

The broader lesson is that enterprise AI spending can extend well beyond a visible assistant. That makes IBM’s business argument plausible. It also raises the standard of evidence: practical deployments must be assessed individually before their outcomes support conclusions about company-wide growth.

What the Earnings Cannot Tell Us About AI

IBM Q3 earnings support a recovery narrative more clearly than they establish a separately measurable AI growth engine.

A financial segment groups revenue for reporting purposes. It does not necessarily group revenue by the customer’s motivation for buying. Software purchased to support AI can share a category with software purchased for other operational needs. That limits what readers can infer from segment growth alone.

IBM’s release does not provide a separate AI revenue line that reconciles the quarter’s growth directly to AI purchases. Without that bridge, assigning a precise share of the improvement to AI would go beyond the disclosed evidence. Management’s interpretation should remain attributed as its interpretation.

This is an attribution problem rather than evidence that AI demand is absent. A product can benefit from a technology trend indirectly. Customers might require better access to information before deploying an application, for example. The resulting sale can be relevant to AI while remaining difficult to classify cleanly.

IBM’s software details also resist a uniform reading. The company reported growth in data and transaction processing, while automation revenue declined 3%. Those categories do not map neatly onto a simple AI-versus-traditional-software division. They do show that the portfolio’s performance varied beneath the headline.

Profitability adds another qualification. IBM reported an adjusted gross margin of 58.5%, down 20 basis points, or 0.2 percentage points. Its adjusted pretax margin increased to 19.3%. Different profit measures therefore moved in different directions, which argues against describing the quarter as an improvement across every measure.

These figures do not identify an AI-specific margin effect. Changes in product mix, operating expenses, and other business factors can affect profitability. Readers should avoid treating an improvement in one measure as proof that AI projects have become more profitable, or a decline as proof of the opposite.

The distinction between contract value and recognized revenue is also important when evaluating future disclosures. A signed agreement describes a commercial commitment. Revenue reflects the amount recorded under applicable accounting rules during a period. Neither should be substituted for the other when assessing the pace of growth.

Customer outcomes require another kind of evidence again. A supplier can record revenue from a deployment before a customer has demonstrated sustained operating benefits. Conversely, a successful project can take time to produce expanded purchasing. Financial results and implementation outcomes answer related but different questions.

The strongest skeptical reading therefore stays specific. The report does not justify dismissing IBM’s AI opportunity, and it does not establish that opportunity as the sole growth driver. It supports further examination of spending composition, customer expansion, and profitability. That is a firmer basis for judgment than either enthusiasm or blanket skepticism.

The Next Quarter Needs a Clearer Demand Picture

The most useful follow-up evidence will distinguish continuing customer expansion from the completion of delayed purchases.

First, watch the composition of software demand through year-end updates and the next reporting cycle. Additional detail about new agreements, renewals, and customer expansion would help explain what is driving sales. Continued strength after delayed transactions clear would strengthen the case for durable demand.

A slowdown would require context rather than an automatic negative verdict. Contract timing can still distort a reporting period. The useful question is whether management provides a consistent explanation supported by product and customer detail. Changing explanations without supporting evidence would make the broader growth argument harder to assess.

Second, watch consulting revenue alongside evidence of projects moving into production. Improvement would suggest that software demand is connecting with more implementation activity. Continued softness would leave open whether customers are delaying work, handling it internally, or purchasing services elsewhere. Those possibilities require evidence rather than assumption.

Customer announcements can help, provided they describe what changed. A deployment with a defined operating purpose offers more information than a partnership statement. Details about expansion and ongoing use would strengthen the commercial narrative. Announcements without such details would do less to clarify the financial opportunity.

Third, watch cash generation against the revised full-year outlook. IBM now expects free cash flow, its measure of cash remaining after specified adjustments and investment spending, to increase by more than $1 billion. Management previously expected an increase of about $1 billion. Delivering that improvement would support its execution argument.

Cash performance would still need to be read alongside revenue and profitability. Strong collections do not establish which technology motivated a purchase. They do help assess whether reported business activity is translating into cash. A weaker outcome would warrant examining timing and operating factors before assigning a cause.

These signals follow directly from the quarter’s central uncertainty. Software performance shows commercial momentum, deployment evidence helps explain its purpose, and cash provides another view of execution. Together, they would offer a more complete assessment than repeatedly asking whether AI is simply good or bad for IBM.

For buyers, the practical response is to bring that same specificity into procurement. Ask vendors to connect their proposals to a defined workflow and explain the evidence required for expansion. A supplier’s improving earnings can indicate business health, but they do not replace a customer’s own evaluation of a project.

Use IBM Q3 earnings as a starting point for comparing financial momentum with deployment evidence. Read the segment results alongside examples of customer work, then track whether the next disclosures connect those stories more clearly. If your organization is evaluating enterprise AI, ask each supplier to identify the information it needs, the work it changes, and the responsibilities that remain. Request a way to assess accuracy, adoption, and operating benefits before expanding the commitment. The useful question for the coming quarter is specific: does returning business lead to additional, sustained customer spending? Keep that question attached to the evidence as IBM reports its progress.