NEW YORK – The number IBM’s chief executive Arvind Krishna reached for on Tuesday, after the company’s shares lost roughly 25 percent of their value in what the Wall Street Journal called the worst single-session collapse in IBM’s 115-year history, was not about artificial intelligence or the cloud. It was something older: “We did not adapt and move quickly enough.”
IBM reported second-quarter revenue of $17.2 billion, up one percent from a year earlier but below analyst estimates. The headline number masked a significant divergence across the company’s businesses. Infrastructure revenue fell seven percent, pulled down by IBM’s Z-series mainframe and Transaction Processing software lines, as clients redirected capital spending toward servers and storage purpose-built for artificial intelligence workloads rather than IBM’s proprietary mainframe stack. Software revenue rose five percent. Consulting grew one percent at constant currency. Red Hat, the open-source software business IBM acquired in 2019, accelerated to 11 percent sequential growth. The mainframe unit’s decline was severe enough to drag total revenue below where analysts expected it.
“These conditions require our teams to execute perfectly, and this quarter we faltered,” Krishna said, using a self-assessment that stripped away the usual corporate hedging about macroeconomic conditions and external factors. IBM acknowledged delayed deal closures and said the supply-chain pricing environment had caused some clients to pull forward spending on commodity hardware, accelerating the shift away from IBM’s differentiated mainframe architecture.
What made the selloff most striking was its context. The Nasdaq Composite and the Dow Jones Industrial Average both ended higher on Tuesday. IBM’s collapse was not the product of a broad risk-off session dragging the market down. It was a company-specific verdict delivered while the rest of the technology sector held steady, which underscored that investors were not retreating from technology as a theme but from a specific version of it – one that depends on mainframe infrastructure and traditional Transaction Processing software rather than AI-native architecture.
IBM’s results present a more textured picture of the AI spending shift than the shorthand “AI is killing enterprise software” might suggest. IBM’s software division, which includes hybrid cloud products and the watsonx AI platform, grew five percent. Red Hat, which sits at the heart of IBM’s open-source and cloud-native strategy, is accelerating. Distributed Infrastructure – the part of IBM’s business building out AI-adjacent capacity – grew 37 percent, its best performance in reported history. The problem is concentrated in the Z-series mainframe franchise and the legacy Transaction Processing business, which runs critical workloads for banks, insurance companies, and government agencies that have been IBM customers for decades.
Fox Business described the warning as sending a “shockwave through the tech industry,” a framing that resonated because IBM’s mainframe installed base is enormous and the institutions that run it are among the most conservative buyers in enterprise technology. CNBC reported the stock cratered 23 percent at one point in the session, with some trackers recording intraday declines above 25 percent. The final figure represents the largest single-day drop in IBM’s history regardless of where it closes.
The magnitude distinguishes this quarter from IBM’s long history of difficult transitions. The company survived the shift from punch-cards to mainframes, from mainframes to personal computers, from personal computers to client-server networks, and from on-premise software to cloud services. Each transition produced a version of this argument: that IBM was too wedded to its existing install base to see the next paradigm. Each time, IBM found a way to reposition. Tuesday’s session does not settle whether the AI transition will be different, but it makes the cost of being caught in transition unusually visible.
The data center construction boom that is redirecting IBM’s clients toward AI-specific hardware carries its own constraints. Power demand from new AI facilities has become a bottleneck that governments have begun to address directly: New York imposed a moratorium on new hyperscale data centers this month over grid capacity concerns. Whether those constraints eventually slow the capex shift is a question IBM’s leadership cannot answer yet.
IBM’s non-GAAP diluted earnings per share came in at $2.93 for the quarter, up five percent, and the company generated $4.8 billion in free cash flow in the first half of 2026. Those figures suggest IBM is not in financial distress – it is generating cash even as its flagship mainframe line struggles. Whether the Software and Red Hat growth can eventually compensate for Infrastructure’s decline depends on whether the AI capex cycle, which is currently hurting IBM’s oldest product line, eventually creates demand for IBM’s newer ones. The full Q2 earnings call with analysts will provide Krishna another chance to make that case. Tuesday’s market verdict was that it was not yet convincing.
