The headline the tape missed
IBM reported second quarter 2026 results on July 22 with revenue of $17.2 billion, up 1% year over year and below the $17.58 billion analysts expected. Adjusted earnings came in at $2.93 a share against a $2.97 estimate, and management pointed to deal timing and weaker infrastructure sales for the shortfall. The stock reaction focused on the miss. For technology leaders evaluating where to place AI budgets, the more useful number sits lower in the release: a generative AI book of business now above $12.5 billion, compounding since 2023. That figure measures committed enterprise demand, and it grew again in a quarter the market read as soft.
CEO Arvind Krishna framed the quarter as continuity, saying IBM is confident in its strategy and portfolio and its ability to capture growth opportunities ahead. The company held its full year view and guided constant currency revenue growth to a four to five percent range, with free cash flow rising about $1 billion. The signal for buyers is durability rather than acceleration. IBM is converting AI interest into signed, multi year commitments across software and services, and it is doing so while the headline growth rate stays modest. That combination rewards patient reading of the segments underneath the top line.
Consulting is the AI distribution channel
Consulting revenue was flat at $5.33 billion, a number that looks unremarkable until you read the signings. Bookings grew 6% to $5.0 billion, a second consecutive quarter of growth, and generative AI accounted for roughly half of signings and more than 30% of backlog. Of the $12.5 billion AI book, about four-fifths runs through Consulting. IBM is selling AI the way large enterprises actually buy it, wrapped in scoping, integration, and change management rather than as a product download. For CIOs, that mix confirms a pattern visible across the market: the hard part of enterprise AI is delivery, and the money follows the people who own delivery.
This has a direct implication for build versus buy decisions. A model license is cheap relative to the labor of connecting it to messy ERP data, defining guardrails, and retraining staff. IBM's numbers put a price on that labor and show it is where the committed spend concentrates. Enterprises that assumed AI would arrive as shrink wrapped software are discovering a services bill instead. The strategic question for your organization is whether you rent that integration muscle from a systems integrator or build it internally. Either way, budget the implementation as the main line item, because that is what IBM's book of business is measuring.
Software carries the margin
Software revenue rose 5% to $7.8 billion, the segment doing the heavy lifting on profitability. Red Hat grew 11%, the Data business climbed 19%, and software annual recurring revenue reached $24.6 billion, up 8%, with HashiCorp and Confluent adding to the recurring base. IBM has spent years assembling a hybrid cloud and data portfolio precisely so that AI workloads have somewhere governed to run. That thesis is now visible in the growth rates. Data and platform revenue expands because agents and models need trusted, integrated data underneath them, and enterprises are paying for the plumbing before they see the payoff at the application layer.
For technology leaders, the read is that the recurring, higher margin parts of IBM are the ones tied to data governance and hybrid deployment. That is the same infrastructure your own AI program depends on. The Red Hat and Data numbers suggest demand for portable, governed environments remains strong even as one time infrastructure purchases slow. If your roadmap assumes AI value comes from the model, IBM's segment detail argues the durable value accrues to whoever controls the data layer and the runtime. That is a useful correction to price into vendor negotiations and internal platform investments this year.
The infrastructure cycle turns down
The soft spot was Infrastructure, where mainframe purchases slowed as the current hardware cycle matured. This is a familiar rhythm for IBM watchers, since the Z systems franchise sells in waves tied to product refreshes, and revenue dips between them. The weakness dragged on the consolidated top line and contributed to the revenue miss. It carries little strategic meaning about AI demand, yet it matters for how CIOs interpret the quarter. A single number labeled revenue growth blends a services business accelerating on AI signings with a hardware business in a cyclical trough, and averaging the two hides the story that matters.
The lesson for anyone benchmarking vendors is to read segments rather than headlines. IBM's consolidated 1% growth understates what is happening in Consulting and Software and overstates the trouble in the core business. Enterprises evaluating IBM as an AI partner should weight the signings, backlog, and ARR figures, which point up, over the infrastructure line, which reflects a cycle rather than a trend. The same discipline applies internally. When you report your own AI progress to a board, separate the durable recurring signals from the lumpy one time effects, or you will get the strategy debate you deserve.
What the book of business actually signals
A $12.5 billion generative AI book is worth interrogating rather than celebrating. It represents cumulative signings since 2023, so it measures pipeline conversion over years, and the four-fifths weighting toward Consulting tells you these are engagements with long delivery tails. That is a healthier signal than a burst of software licenses, because it implies customers are committing to multi quarter programs with IBM inside them. For CIOs, the takeaway is that peers are treating enterprise AI as a sustained transformation program with external delivery partners, and they are signing paper that commits budget well into future years rather than running perpetual pilots.
The risk sits in the same place as the opportunity. Services heavy AI books depend on delivery capacity and on projects actually reaching production, and Gartner has warned that a large share of agentic projects will stall on cost and governance. IBM's backlog growth suggests it is converting, but a book of business is a promise until the work ships. Technology leaders should ask their own integrators the same question the market should ask IBM: what percentage of signed AI work is live in production and generating measured value. The answer separates a real transformation program from an expensive backlog.
The read for your roadmap
IBM's quarter is a data point about where enterprise AI money lands, and it lands on integration. The clearest planning input is the ratio: for every dollar of AI software, roughly four dollars of consulting and delivery. If your 2026 budget assumes the reverse, revisit it. The organizations moving fastest are the ones that funded the delivery layer, the data plumbing, and the governance scaffolding before they expected application level returns. IBM has monetized exactly that sequence, and its segment detail is a free benchmark for how your own program should be resourced and sequenced over the next several quarters.
The second input is patience with modest headline growth. IBM is compounding a durable AI book while consolidated revenue grows in the low single digits, and that is a realistic picture of enterprise transformation at scale. Boards conditioned to expect AI to spike revenue should recalibrate toward steady conversion of signed commitments into deployed capability. For CIOs, the credible story to tell upward is the one IBM is telling: growing signings, expanding recurring software, and a disciplined path from committed pipeline to production. Sell the compounding, resource the delivery, and measure what actually reaches users.
