IBM stock suffers worst crash since 1987 after AI spending shift blindsides Q2 results

IBM shares crashed as much as 24.8% on July 14, marking the stock's worst single-day decline since the October 1987 Black Monday crash, after the company issued a preliminary Q2 2026 earnings warning that fell well short of Wall Street expectations. The drop wiped out an estimated $70 billion in market value in a single trading session.

The scale of the reaction matters here as much as the miss itself. IBM has spent recent years rebuilding investor confidence around a hybrid cloud and AI strategy, and a crash of this size — on preliminary numbers, not even the final earnings release — signals that markets interpreted this as a structural problem with IBM's AI positioning, not a one-quarter blip.

What actually went wrong

IBM's preliminary Q2 2026 revenue came in at roughly $17.2 billion, missing analyst estimates, with CEO Arvind Krishna writing in a letter to investors that the company had underestimated "the magnitude" of a shift in how clients are spending their technology budgets. According to Krishna, customers diverted capital toward servers, storage, and memory — the physical infrastructure needed to run AI workloads — while several large software and services deals that IBM had expected to close during the quarter slipped into later periods.

In effect, IBM's own customers got spooked by the same AI infrastructure demand crunch that has driven up prices and lead times for GPUs, high-bandwidth memory, and datacenter components industry-wide. Rather than proceeding with planned IBM software and consulting purchases, those clients redirected spending to stockpile the hardware layer first, leaving IBM's higher-margin software and services lines short of their targets for the quarter. IBM's own software revenue still grew — up roughly 5%, with products like Lightwell (built on Red Hat) cited as a bright spot — but it wasn't enough to offset the broader shortfall.

Why this crash is different from a typical earnings miss

Public companies miss earnings estimates regularly without triggering a 58-year record crash. What made this one land so hard is the framing: IBM is telling investors that enterprise AI infrastructure demand is now large enough to distort spending patterns at one of the world's largest enterprise software and services companies. If IBM's own customer base — largely regulated, risk-averse enterprises in banking, government, and healthcare — is reallocating budget away from software modernization toward raw AI compute capacity, that's a signal about the current phase of the AI buildout: infrastructure is still being prioritized over the software layer that's supposed to run on top of it.

That reading lines up with other recent data points. TSMC reported a 68% jump in June revenue driven by AI chip demand outstripping supply, and Micron has committed $250 billion to new US chip manufacturing — both signs that the hardware layer of the AI buildout remains supply-constrained and commanding disproportionate capital. IBM's earnings miss is arguably the first major sign of how that hardware-first spending pattern is squeezing revenue at software-and-services incumbents that don't sell the underlying AI infrastructure themselves.

What to watch next

IBM's full Q2 2026 earnings announcement is scheduled for July 22, which will clarify how much of this shortfall is a timing issue — deals delayed, not lost — versus a more durable shift in enterprise IT budget allocation. Krishna's own language, describing a quarter where the company "faltered," suggests IBM sees this as at least partly self-inflicted: a failure to anticipate how quickly the AI infrastructure crunch would reshape client priorities, not merely bad luck. For enterprise software vendors more broadly, IBM's experience this quarter is now the clearest cautionary data point yet that AI infrastructure spending is actively cannibalizing budget from adjacent categories, not simply adding incremental demand on top of existing IT budgets.