IBM mainframe AI demand: How AI hardware costs are pausing mainframe upgrades, and why IBM expects recovery

IBM’s Q1 2026 results delivered a stark lesson in how AI capital expenditure cycles can disrupt even the most entrenched legacy hardware. Mainframe revenue dropped 42%, the company’s stock fell 25%, and full-year guidance was cut. But the company insists the decline is temporary, arguing that AI’s cost pressures on data-center gear pushed customers to delay mainframe purchases, not abandon them. For AI builders and enterprise operators, the episode highlights the volatility hardware procurement faces when AI infrastructure demand inflates component prices, and why software renewals and hybrid cloud strategies remain the safer bet.

What happened

IBM reported revenue of $17.2 billion, gross profit of $9.9 billion, and net earnings of $2.2 billion for the quarter, missing Wall Street expectations. The company had warned investors ahead of time, triggering a 25% intraday stock drop. The primary culprit was the mainframe business, which declined 42%.

That decline is especially damaging because IBM earns about $3 in software revenue for every $1 of mainframe hardware. CFO Jim Kavanaugh explained that “tens” of customers who were due to buy new mainframes opted not to, instead redirecting budgets to other hardware. CEO Arvind Krishna cited 15-30% cost increases in data-center gear and PCs, driven by the AI build-out. Similar pressures have been flagged by Dell, HP, and Apple.

Why AI builders should care

This episode shows how AI-driven hardware price surges can cascade into legacy infrastructure budgets. For teams building AI products or managing enterprise deployments, it reinforces the importance of monitoring hardware cost trends and planning for unpredictable procurement cycles. The mainframe’s decline here is not a market shift away from the platform but a temporary budget reallocation toward AI-capable hardware. That means AI builders should expect continued software renewals and AI platform adoption (via Red Hat and watsonx) even as hardware sales pause.

Practical implications

For AI product teams and operators, the practical takeaway is threefold. First, budget planning for on-prem hardware should build in potential 15-30% cost increases and consider delaying large capital purchases until component prices stabilize. Second, IBM’s software-heavy business model means that even a hardware pause doesn’t kill the overall revenue stream; AI builders using IBM’s software stack (including Red Hat OpenShift and watsonx) should expect continued support and updates. Third, the episode validates hybrid cloud strategies that decouple software workloads from specific hardware cycles, a pattern that enterprises may accelerate.

Caveats

IBM’s framing of a temporary dip is self-serving. The company has a long history of defending the mainframe, and the industry has predicted its death for decades. The evidence comes from a single quarter, and the recovery depends on macroeconomic conditions and AI infrastructure spending stabilizing. Additionally, the details on which customers delayed and the exact mix of AI versus non-AI hardware spending are not publicly available. The claims about Red Hat and watsonx as part of IBM’s AI strategy come from analyst commentary rather than IBM’s official earnings materials. Readers should treat the recovery timeline as uncertain.