IBM’s AI-era warning shows the mainframe moat is not immune to the chip squeeze

Technology reporting

IBM did something companies try hard not to do between earnings dates: it warned investors early. On July 14, the company furnished a letter from CEO Arvind Krishna saying IBM expects second-quarter 2026 revenue of $17.2 billion, up 1 percent from a year earlier, with operating earnings of $2.93 per diluted share. That is not a collapse in the business. It is, however, a sharp miss against the version of IBM investors thought they owned: a steadier enterprise software and hybrid-cloud company with enough mainframe durability to keep the cash engine humming while AI spending moves around the rest of the industry.

The market treated it as more than a quarterly wobble. CNN reported IBM shares were down 24 percent in premarket trading Tuesday, putting the stock on pace for its worst day since Black Monday in 1987. Forbes later reported the shares closed down 25.2 percent, around $217, calling it the largest single-day loss in IBM’s 115-year history. IBM’s preliminary revenue was about $650 million below the $17.85 billion analyst estimate cited by Yahoo Finance, and non-GAAP EPS was $0.09 below the $3.02 estimate.

This is not financial or investment advice. The business question is narrower and more useful: what broke in IBM’s model, and does it say something larger about enterprise technology spending in the AI buildout?

What happened

IBM’s own filing is the cleanest source. In a Form 8-K filed with the Securities and Exchange Commission, IBM said the July 14 investor letter covered “selected preliminary second-quarter 2026 financial results” and cautioned that it was still closing the quarter, so final results “could be slightly different.” The full earnings call is scheduled for July 22.

The preliminary scorecard is mixed. IBM said revenue was $17.2 billion, up 1 percent. Software revenue rose 5 percent. Consulting was flat, or up 1 percent at constant currency. Infrastructure revenue fell 7 percent. GAAP gross margin was 57.7 percent, down 100 basis points. Operating gross margin was 59.4 percent, down 70 basis points. GAAP pre-tax income margin was 14.4 percent, down 90 basis points, while operating pre-tax income margin was 19.2 percent, up 30 basis points. Year-to-date net cash from operating activities was $7.8 billion, and free cash flow was $4.8 billion.

So the headline is not “IBM stopped making money.” It is that IBM’s highest-confidence machinery jammed when investors expected resilience.

Krishna’s explanation was unusually plain. IBM had expected Infrastructure revenue to decline in the low single digits for the year after the launch cycle for its z17 mainframe. Instead, the company said the quarter was “worse than our expectations,” driven by a shortfall in Z performance and related software, primarily transaction processing. In late June, IBM said clients shifted quarterly capital spending toward servers, storage and memory to secure supply-constrained infrastructure ahead of expected price increases. IBM also said clients were distracted by fast-moving cybersecurity concerns.

That distinction matters. IBM is not saying demand for computing disappeared. It is saying customer budgets were re-ranked toward scarce hardware inputs, leaving less room for expected mainframe and associated software deals to close on schedule.

What the numbers actually show

IBM’s preliminary figures show three simultaneous realities.

First, IBM is still a cash-generating enterprise supplier: $17.2 billion in quarterly revenue, operating EPS growth of 5 percent, and $4.8 billion in first-half free cash flow.

Second, the miss is meaningful because it hit the parts of the story that carry strategic weight. Infrastructure was down 7 percent, not the low-single-digit decline management had expected for the year. The Z mainframe cycle matters because mainframes are not nostalgic museum pieces inside IBM; they sit inside banking, payments, insurance and government workloads where reliability, transaction volume and switching costs are core selling points. When Z-related demand slips, the weakness reaches into the software stack attached to it.

Third, the margin picture is not simple. GAAP gross margin declined, but operating pre-tax margin expanded. IBM can still protect adjusted profitability, but margin discipline did not save the revenue narrative. The letter made the growth engine look more exposed to timing, capex competition and client distraction than expected.

One useful comparison: IBM said revenue rose 1 percent to $17.2 billion, implying roughly $17.0 billion in the year-earlier quarter. The issue is not dramatic year-over-year shrinkage. It is underperformance in a market already hypersensitive to whether AI spending is expanding the enterprise software pie or simply crowding out other projects.

Why IBM did it

The company framed the warning as transparency before the formal July 22 earnings call. Strategically, it also lets IBM shape the interpretation before analysts and investors fill the gap themselves.

IBM’s letter tries to make three points. The first is external: clients shifted spending toward supply-constrained servers, storage and memory. The second is operational: IBM’s teams did not adapt quickly enough, and large deals missed expected closing timelines. The third is strategic: IBM argues the longer-term portfolio remains strong, citing Red Hat revenue growth accelerating to 11 percent, strong performance from HashiCorp and Confluent acquisitions, Distributed Infrastructure up 37 percent, and z17 still running at nearly 130 percent program-to-program compared with z16.

That is company framing, not independent proof that the issue is solved. But it tells us how IBM wants investors to read the miss: hardware scarcity disrupted customer sequencing faster than IBM’s sales machine could adjust.

The more skeptical read is that AI budgets are not bottomless magic wallets. When companies face price increases, scarce memory, cybersecurity triage and infrastructure deadlines, even mission-critical platforms can be pushed into the next quarter.

Who wins and loses

The immediate loser is IBM’s credibility premium. The company has spent years repositioning around software, Red Hat, automation, consulting and enterprise AI. A preliminary miss tied to mainframe timing and related software does not erase that work, but it dents the idea that IBM’s installed base insulates it from AI supply-chain turbulence.

Customers may gain negotiating leverage if IBM needs to pull delayed deals back into the pipeline quickly. But if scarce AI infrastructure is forcing tradeoffs, enterprises may have to delay other modernization projects or accept higher total costs.

Hardware and infrastructure suppliers are the relative winners in IBM’s own telling. If clients diverted spending to secure servers, storage and memory before expected price increases, that suggests suppliers of constrained AI-era infrastructure have pricing and timing power. IBM’s Distributed Infrastructure business also benefited, with the company saying it had its best reported performance, up 37 percent, and exited the quarter with about $500 million in backlog.

Competitors in enterprise software and cloud services get a complicated opening. Microsoft, Amazon Web Services, Oracle, ServiceNow, SAP and others can point to IBM’s stumble, but they are not immune to the same budget physics. If AI infrastructure spending crowds out other enterprise purchases, every vendor selling transformation projects has to prove priority, not just relevance.

What to watch next

The July 22 earnings call now matters more than the preliminary numbers. Watch full-year guidance first. If management keeps its broader outlook, this may look like a painful timing issue. If guidance comes down, the market will treat it as a demand or execution problem.

Second, watch delayed deal conversion. IBM said numerous large deals failed to close on expected timelines. The key question is whether those deals close in the third quarter or disappear into budget purgatory.

Third, watch Red Hat and software durability. Software was still up 5 percent, and Red Hat was up 11 percent. If those lines stay strong, IBM can argue its strategic center is intact. If software slows further, the mainframe shortfall becomes a wider enterprise-platform problem.

Finally, watch AI capex crowd-out across the sector. IBM’s most consequential disclosure may not be IBM-specific at all. If customers are shifting spending toward servers, storage and memory ahead of price increases, other enterprise vendors should be asked the same question this earnings season: are AI infrastructure needs expanding budgets, or simply taking the first seat at the table?

IBM is still profitable and entrenched. But the AI boom is rearranging customer budgets in real time, and IBM just showed how quickly that can hit even very sticky systems.

Sources

• IBM Form 8-K and July 14, 2026 investor letter filed with the SEC
• IBM Exhibit 99.1: Arvind Krishna’s letter to IBM investors
• IBM investor events and earnings webcast page
• Yahoo Finance: IBM stock crashes after major warning
• CNN Business: IBM is on pace for its worst day ever
• Forbes: Here’s Why IBM Had Its Worst Day Ever

---

Shadowfetch is a technology publication. Explore Shadowfetch Linux — our own Linux build — and the Shadowfetch apps on the App Store.