IBM’s Mainframe Revenue Dropped 42%: AI Budget Pressure Is Why

IBM reported second-quarter 2026 earnings on July 22 and the numbers were bad enough that CEO Arvind Krishna had already warned investors in a letter the week before. Revenue came in at $17.2 billion with $9.9 billion in gross profit, but the results still fell well short of analyst forecasts. The culprit was a 42% collapse in mainframe hardware revenue. IBM's stock dropped 25% on the news, the steepest single-day fall in the company's 115-year history, and the company cut its full-year growth outlook.

What happened

Metric	Q2 2026 figure
Total revenue	$17.2 billion
Gross profit	$9.9 billion
Gross margin	~58%
Net earnings	$2.2 billion
Mainframe revenue change	-42%
Stock single-day decline	-25% (largest ever)

IBM’s mainframe segment is the company’s core profit engine. CFO Jim Kavanaugh explained on the investor call that IBM collects roughly $3 in software revenue for every $1 of mainframe hardware it sells. A 42% hardware shortfall therefore ripples far beyond the hardware line itself.

The immediate cause, according to Krishna, was “tens” of enterprise customers who were scheduled to buy new mainframes in the quarter but chose to delay. Mainframes carry price tags ranging from hundreds of thousands to millions of dollars per unit, plus ongoing maintenance and software contracts worth many millions more. A relatively small number of delayed deals can move the revenue needle sharply.

Why did customers delay their mainframe orders?

The AI hardware boom is the short answer. Data center gear and PC prices have climbed 15% to 30%, driven by surging demand for AI infrastructure components, particularly memory. Enterprise hardware makers including Dell and HP have already flagged those cost increases publicly, and Apple has made similar warnings.

When customers faced those sharp price increases elsewhere in their IT budgets, many reallocated spending to cover the gap and pushed their mainframe refresh to a later quarter. Krishna was direct about the chain of events: “When they were faced with that issue, then they decided to move budget to those areas where they were having that extreme price.”

This is the same AI investment wave that had been boosting IBM’s stock for much of Krishna’s six-year tenure. The AI boom that had lifted IBM’s valuation was also the force that diverted customer spending away from its biggest product.

Is the mainframe actually dying?

Krishna and Kavanaugh both argued the answer is no, and that this quarter’s miss is a timing problem rather than a structural one. Krishna stated there is “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 the death of the mainframe for decades without that prediction coming true. Large banks, insurers, and government agencies continue to run core transaction workloads on IBM Z systems because the platforms offer reliability and throughput characteristics that commodity servers do not easily replicate.

That said, IBM also lowered its full-year growth forecasts on Wednesday, confirming the Q2 shortfall will affect the rest of 2026, regardless of how quickly delayed orders come in.

Why it matters

This story is a concrete illustration of how AI capital spending is reshuffling enterprise IT budgets in ways that hurt even profitable incumbents. If your business sells or resells enterprise hardware, or if you depend on a vendor whose revenue is tied to large infrastructure cycles, watch for similar budget displacement effects.

For digital-first businesses, the takeaway is more indirect. AI-driven component price inflation is real and is already visible in quotes for servers, memory, and related gear. Any project that relies on on-premises hardware could face cost surprises this year.

If you are thinking about integrating AI tools into your own operations, this is a good reminder to pressure-test vendor pricing assumptions before committing to multi-year contracts.

Our take

IBM’s numbers are genuinely bad, but Krishna’s explanation is plausible. Enterprise procurement cycles are lumpy by nature, and a handful of delayed mainframe deals can produce exactly this kind of cliff-edge miss. The 42% drop looks catastrophic, but if half of those delayed orders land in Q3, the recovery story writes itself.

What is harder to dismiss is the structural tension. Every dollar a large enterprise spends on GPU clusters or AI cloud services is a dollar it is not spending on a mainframe refresh. IBM has been repositioning itself around AI consulting and hybrid cloud for years, but the mainframe is still where the cash comes from. If the AI buildout drags on and enterprise budgets stay stretched, “temporary blip” could quietly become “new normal.”

The 25% stock drop in a single session is also worth noting as a data point. Markets were not surprised by bad numbers (Krishna pre-warned everyone), but the size of the reaction suggests investors are less certain than management that this is a one-quarter anomaly. Worth watching the Q3 mainframe line closely.

For broader context on how AI infrastructure investment is reshaping enterprise spending patterns, see our ongoing AI news coverage.

What to do about it

1. Audit any IT hardware quotes you have received in the past 60 days. Component price inflation of 15% to 30% is now a documented market condition, so build a buffer into your budget.
2. If you are mid-way through a vendor selection for infrastructure, ask suppliers explicitly whether component costs are locked or subject to revision before delivery.
3. Watch IBM’s Q3 earnings for the mainframe line. If delayed orders do not return, the “temporary blip” narrative breaks down and the stock could reprice again.
4. If your business runs IBM mainframe workloads, factor in the possibility that IBM’s software pricing strategy could shift if hardware volume stays soft.

The bottom line: AI spending is compressing enterprise budgets in unexpected places, and even a company generating $9.9 billion in quarterly gross profit is not immune.

Source: TechCrunch · AI