The IBM mainframe business decline hit the company at full force in Q2 2026, with hardware revenue cratering 42% and dragging IBM’s stock to its worst single-day loss on record. The shares fell 25% after the company took the extraordinary step of pre-announcing its own miss, with CEO Arvind Krishna publishing a preliminary letter to IBM investors on 14 July 2026. When a chief executive warns you before the bad news lands, you know it’s bad.
The headline numbers were still substantial, in the way that only a 115-year-old franchise can be: $17.2 billion in revenue, $9.9 billion in gross profit, margins of nearly 58%, and $2.2 billion in net earnings for the quarter. But Wall Street had its own numbers in mind, and IBM fell well short of them.
Behind the IBM Mainframe Business Decline
The mechanics of the slump are worth understanding, because this is where the cascading logic gets uncomfortable. CFO Jim Kavanaugh explained on the earnings call that IBM earns roughly $3 in software revenue for every $1 of mainframe hardware it sells. A 42% hardware drop, therefore, is never just a hardware problem. It pulls software and services contracts down with it.
So what happened? According to Krishna, ‘tens’ of customers who were due to upgrade during the quarter made a different call. The AI buildout boom drove component costs up by 15% to 30% across data centre gear and PCs, and enterprise customers facing those price spikes redirected their budgets accordingly. ‘When they were faced with that issue, then they decided to move budget to those areas where they were having that extreme price,’ Krishna said. Dell and HP have flagged similar cost pressures; Apple has said the same.
Crucially, the 42% drop came off a very high base. According to the Q2 2026 earnings call, Kavanaugh noted the decline was measured against a period of 70% growth during the prior year’s mainframe launch quarter. IBM’s own CFO had flagged at the Q4 2025 earnings call in January 2026 that the company expected to lose roughly half a point of margin in 2026 because of the wrapping effect from that unprecedented launch momentum. In other words, a painful fifth-quarter hangover was already baked into the forecast; the AI-driven budget displacement made it worse.
IBM’s infrastructure segment (the category that houses mainframe) was down 7% overall, per the preliminary letter, while software grew 5% and consulting was flat, up 1% at constant currency. The preliminary Q2 2026 results showed GAAP diluted earnings per share from continuing operations of $2.27, down 2%, and Operating (Non-GAAP) diluted EPS of $2.93, up 5%. GAAP gross profit margin came in at 57.7%, down 100 basis points year-over-year.
Krishna’s Case That This Is Not a Structural Break
The CEO and CFO spent most of the call making a single argument: the mainframe is not dying, the customers just paused. Krishna insisted, ‘We see no evidence of clients moving off the mainframe,’ and said some of the delayed buyers had already placed orders in the current quarter. The tech industry has been predicting mainframe obsolescence for decades. That prediction has been wrong every time.
There is also the cash generation story, which tends to get buried in the noise around a big miss. IBM’s preliminary letter disclosed year-to-date net cash from operating activities of $7.8 billion and free cash flow of $4.8 billion for the first half. For context, full-year 2025 delivered $14.7 billion in free cash flow, a 16% increase over 2024, per the Q4 2025 earnings call transcript. The engine still runs; it just had one ugly quarter.
IBM also lowered its full-year revenue growth forecast to 4% to 5%, down from the five-plus percent guided at the start of the year, according to The Next Web. Yet shares rose roughly 3% in extended trading after the full results dropped on Wednesday, a sign that the pre-announcement had already done much of the damage-control work.
The company’s IBM investor relations team now has a simple job: prove, over the next two quarters, that those delayed mainframe orders actually arrive. If they do, this reads as a nasty but temporary blip. If they don’t, the $3-per-dollar software multiplier starts working in the wrong direction for a very long time.
