IBM misses Q2 targets as clients rush to hoard AI infrastructure

On Tuesday, IBM released a letter to investors from CEO Arvind Krishna announcing preliminary second-quarter results a week ahead of its scheduled July 22 earnings report. The numbers disappointed: revenue totaled $17.2 billion（約2.8兆円）, up just 1% year over year and short of the company's own guidance. The market reaction was severe—shares fell about 24% on Tuesday, one of the worst single-day drops in IBM's history, and slid further on Wednesday to a 52-week low, pushing the company's market capitalization below $200 billion（約32兆円）.

The shortfall marked a sharp reversal from the first quarter, when IBM's revenue had risen 9% year over year, driven by infrastructure revenue that jumped 15% as the company rolled out its new z17 mainframe. IBM had expected that mainframe momentum to fade and guided for infrastructure revenue to decline by a low-single-digit rate for the year. Instead, in the second quarter, infrastructure revenue fell 7%, software grew just 5%, and consulting was flat. Earnings per share of $2.27 declined 2% year over year, though adjusted earnings per share rose 5%.

Krishna's explanation of what happened reveals the underlying cause. "In the last few weeks of June, we saw clients shift their quarterly capex spend toward servers, storage, and memory purchases to secure supply constrained infrastructure ahead of expected price increases," he wrote. He acknowledged that IBM "did not anticipate the magnitude of the capex reprioritization" and noted that numerous large deals failed to close on the timelines the company expected, accounting for the majority of the shortfall. He also cited widespread cybersecurity concerns that distracted clients during the quarter. In essence, customers redirected their technology budgets toward hardware before prices rose, and other purchases got pushed out.

The irony is visible in IBM's own numbers. The company's distributed infrastructure business, which includes Power servers and storage hardware, grew 37% year over year—its best performance in the company's reported history—and exited the quarter with a backlog of about $500 million（約800億円）. At a much larger scale, the memory-chip market showed the same pattern: Micron Technology reported revenue of $41.5 billion（約6.6兆円） in its most recent quarter, up 346% year over year, with DRAM selling prices more than doubling. Micron attributed the surge to AI-driven demand for memory and storage accelerating at a rate greater than the industry's ability to increase supply.

The broader implication is clear. The AI infrastructure buildout has moved beyond the giant cloud companies deploying capital in their own data centers. Ordinary enterprises are now competing for the same servers, storage, and memory, pulling money from the rest of their technology budgets to secure supplies. That creates tailwinds for memory and AI-infrastructure suppliers while introducing new risks for vendors—like IBM—whose revenue depends on large deals closing on schedule. For IBM specifically, the July 22 earnings call will determine whether the slipped deals are merely delayed or permanently lost. Management will discuss full-year expectations on the call; the company had guided for constant-currency revenue growth of more than 5% in 2026, and if that outlook survives, most of this quarter's damage was a timing issue. If it comes down, the problem may run deeper than one quarter.

IBM's pre-announcement, released a week ahead of its scheduled July 22 earnings date, marks a significant moment in how enterprise technology spending is reshaping around AI. The company's revenue growth of just 1% year over year reversed the strong 9% growth IBM posted in the first quarter, when its new z17 mainframe drove infrastructure revenue up 15%. Krishna had already expected mainframe momentum to fade, but the company did not anticipate the magnitude or the specific direction of the shift: clients were not pulling back on technology spending in general—they were redirecting it urgently toward hardware.

The irony is that IBM's own business lines reveal exactly where that redirected money landed. While the company's software business grew only 5% and consulting was flat, distributed infrastructure surged 37% year over year. This pattern repeats at much larger scale across the memory-chip market: Micron Technology, one of the world's largest memory-chip makers, reported revenue up 346% year over year with DRAM selling prices more than doubling, and the company attributed the acceleration to AI-driven demand for memory and storage outpacing supply.

What this signals is a phase shift in the AI buildout itself. The cycle is no longer confined to hyperscalers (large cloud providers) making massive capital bets in their own data centers. Mainstream enterprises—the customers IBM depends on for consulting and software deals—are now competing for the same constrained hardware supplies and willing to defer other technology projects to lock in AI infrastructure before prices rise. For vendors like IBM whose revenue depends on large deals closing on predictable timelines, this creates a new risk: customer budgets are being commandeered by urgent hardware procurement, pushing other purchases into the future or off the table entirely.