IBM has spent two years telling investors that artificial intelligence would show up in enterprise software before it showed up anywhere else. On Wednesday the company got to point at a number.
Revenue for the third quarter came to $17.1 billion, up 5% from a year earlier and 6% at constant currency, according to results released after the close on Oct. 21. Analysts had expected $16.82 billion. Adjusted earnings of $2.95 a share beat a consensus of $2.86, and net income reached $2.2 billion, or $2.31 a share on a diluted basis.
The quality of the beat mattered as much as its size. Gross margin slipped 50 basis points to 56.8% on a reported basis, but the operating pre-tax margin widened 70 basis points to 19.3%, which is the line that reflects how the company is managing costs rather than how it is pricing hardware. IBM generated $11.8 billion of operating cash in the first nine months and $7.8 billion of free cash flow.
Software did the work. The unit produced $7.8 billion of revenue, up 8% and up 9% at constant currency, with Red Hat and hybrid cloud ahead 11% and the data business ahead 19%. Transaction processing, the mainframe software stack that customers rarely replace, grew 9%. Automation was the exception, declining 3%.
Two acquisitions are compounding inside those figures. Confluent, bought to handle the streaming data that feeds AI models, delivered bookings growth of about 20%, and HashiCorp continued to add bookings in every geography. Red Hat’s OpenShift platform now carries $2.4 billion in annual recurring revenue, a base large enough that its growth rate moves the segment on its own.
Arithmetic explains the attention that mix receives. Software supplied $7.8 billion of the quarter’s $17.1 billion, close to half the total, and grew faster than the company as a whole; consulting contributed $5.3 billion and did not grow at all. Each point of software growth now carries more of the quarter than it once did, which is why the segment’s 8% rate matters more than the headline 5%.
The softer lines explain why the stock did not rally. Consulting revenue was $5.3 billion, flat as reported and up 2% at constant currency, a reading that suggests clients are still buying advice more slowly than they are buying software. Infrastructure rose 4% to $3.7 billion, but IBM Z mainframe revenue fell 20% against a strong comparison from the current product cycle, even as the company reported a record quarter for storage.
Chief Executive Arvind Krishna attributed the growth to clients moving past experiments. “Our clients are now deploying Super Intelligence in ways that transform entire operating models, instead of simply addressing narrow or novel problems,” he said in the release, using the company’s preferred term for AI that runs across a whole business rather than a single task. He also said sovereign AI, the business of building models and data systems inside national borders, is now gaining real traction after two years of talk.
The mainframe question hangs over the quarter. Krishna said deals that slipped in the second quarter closed in the third, which he argued validated the company’s position that the earlier shortfall reflected timing rather than demand. He said the current Z17 generation has generated more than $1 billion of incremental revenue over six quarters compared with prior programs, and that 85% of mainframe customers are holding or expanding capacity.
Guidance was left largely intact, with a raised cash target. IBM still expects constant-currency revenue growth of 4% to 5% for the year and now expects free cash flow to rise by more than $1 billion year over year, a change that speaks to confidence in collections rather than in demand.
Shares slipped in after-hours trading despite the beat, a reaction that analysts linked to the flat consulting line and to expectations built up during IBM’s recovery from a soft second quarter. Investors have grown used to the company clearing the bar it sets for itself, which leaves little room for a beat to move the stock.
The wider question is whether enterprise AI spending is finally turning into software revenue across the industry or only at IBM. The pattern across enterprise technology this year, analysts said, has been firm demand for software and infrastructure running alongside slower services growth, a split that suggests corporations are buying the plumbing for AI while postponing the consulting engagements that would tell them what to do with it.
Krishna’s answer is that application modernization is a multiyear job. Most enterprise data still sits outside public clouds, he said, and every AI project requires rewriting or modernizing an application before it can use that data, which pushes customers toward Red Hat and OpenShift over time. That is a plausible argument, and it is also the argument IBM has made for three years running.
What the quarter does establish is narrower. Software grew faster than the company overall, margins expanded, cash flow improved, and the guidance held. For a company whose critics have spent a decade predicting that its growth would stall at 2%, that is a serviceable result.
If AI demand is real but arrives as software licenses rather than consulting contracts, how long can IBM’s $5.3 billion services business stay flat before the mix stops flattering the margin?
