By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
IBM's Q2 miss and subsequent stock plunge highlight concerns about its ability to transition to high-margin software while being tied to a legacy infrastructure cycle. The key question is whether the margin compression from mix-shift is temporary or permanent.
Risk: Permanent cannibalization of high-margin software demand by AI infrastructure
Opportunity: Converting AI infrastructure wins into recurring, margin-rich software revenue
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
IBM (IBM) shares lost roughly one-quarter of their value on July 14 after the tech giant dropped disappointing preliminary second-quarter results.
The collapse was historic, but the severity of the earnings gap alone doesn't explain the market reaction.
IBM could be in an artificial intelligence expenditure trap.
Corporate IT budgets are still expanding, but buyers are shifting money to servers, storage, and memory before tight supplies push prices higher. That's good news for hardware makers and bad news for the software and consulting deals IBM is relying on more and more for profitable growth.
IBM is on both sides of that market. It provides mainframes, storage, and enterprise software and consults corporations on technological projects.
But such enterprises don't have equal economics.
IBM's software division had a gross margin of 82.8% in the first quarter. Infrastructure accounted for 56.9% and consultancy for 27.5%. A dollar deferred in software might thus cost the organization more than a dollar earned elsewhere benefits it.
The concern for investors is no longer just whether IBM missed one quarter.
Is the AI infrastructure boom just a temporary reshuffling of enterprise budgets, or is it permanently draining dollars from the products that make up the core of IBM's valuation?
"These conditions require our teams to execute perfectly, and this quarter we faltered. We did not adapt and move quickly enough," IBM CEO Arvind Krishna said.
IBM's AI strategy depends on software winning the budget fight
IBM came into the second quarter with a lot of enthusiasm.
Revenue grew 9% to $15.9 billion in the first quarter. Software sales grew 11%, infrastructure jumped 15%, and consultancy grew 4%. The company produced $2.2 billion of free cash flow and reaffirmed its outlook for more than 5% constant currency revenue growth in 2026.
That achievement helped IBM's transition from a slowly expanding legacy technology company to one more and more valued for software, hybrid cloud, and artificial intelligence.
Software drives that change.
IBM's biggest business, the segment generated $7.1 billion of first-quarter revenue. Red Hat, automation, data products, and transaction-processing software drove growth, according to IBM's first-quarter statistics.
IBM has also been investing extensively in watsonx, its suite of technologies for designing, regulating, and deploying artificial intelligence applications.
The company's objective is not only to compete in training the biggest model or building the fastest AI accelerator. It seeks to assist organizations in linking artificial intelligence to their data, existing apps, and regulated business processes.
In theory, the method should benefit IBM as firms move beyond experimentation with AI to deploying it across their operations.
The problem is enterprise technology budgets have their limits.
A corporation might need to shore up its processors, memory, networking equipment, and storage before it can ship more software. But scarcity and anticipated price hikes might make those purchases urgent, while a software agreement could wait until next quarter.
IBM's next mainframe was designed to bolster both sides of the portfolio.
The IBM z17 includes more than 250 potential AI use cases, with built-in artificial intelligence capabilities. Strong mainframe demand also tends to produce related transaction-processing software revenue as customers need software to run the systems.
That linkage is what makes the second-quarter miss even more troubling.
Hardware revenue wasn't the only weak spot in IBM Z. It also cut demand for the related software stack, converting one delayed infrastructure acquisition into a broader earnings problem.
IBM's preliminary results reveal the AI spending trap
IBM's preliminary second-quarter revenue increased 1% to $17.2 billion.
Software sales grew 5%, consulting was flat, and infrastructure decreased 7%. Operating, non-GAAP diluted earnings rose 5% to $2.93 per share. Operating gross margin, however, declined 70 basis points to 59.4%. Year-to-date free cash flow was $4.8 billion.
Those findings were not disastrous by themselves.
Revenue and adjusted earnings were still up, and IBM was still profitable. The market response was a measure of how far the figures fell short of the growth narrative investors were looking for following the first quarter.
Krishna said IBM anticipated some fall after the strong z17 launch but misjudged how steep the decline would be.
The corporation cited lower revenue from related transaction-processing software and inadequate IBM Z performance. IBM said numerous large deals failed to close on the timelines management expected, accounting for most of the shortfall. Recent acquisitions HashiCorp and Confluent, however, performed well.
In the last weeks of June, customers moved capital investment to servers, storage, and memory. They were trying to buy equipment that suppliers had constrained before expected price increases.
IBM didn't totally avoid the spending.
Distributed infrastructure revenue was up 37%, its best performance ever, as IBM said its Power and storage products were up. The unit left the quarter with around $500 million in backlog. Revenue growth for Red Hat was also up 11%.
And that is what creates the trap.
IBM can get a piece of the hardware spending boom, but the revenue mix may be less profitable than the software transactions clients are delaying. Thus, the company may capitalize on demand for AI infrastructure, even while it underperforms for investors.
IBM claimed industrywide security concerns were rapidly evolving and distracting customers. The company's warning follows Anthropic's revelation that their Claude Mythos Preview model was able to discover and exploit previously unknown software vulnerabilities under testing conditions.
The development is a big deal for cyber defenders, as similar models might slash the time and cost of finding exploitable holes, Anthropic said. The company's Project Glasswing update warned of how the usual delays between finding, fixing, and releasing updates might grow more problematic.
That development is both a threat and an opportunity to IBM.
Clients may postpone routine software projects while they weigh emerging vulnerabilities. At the same time, there may be more demand for tools that help firms patch and safeguard their software supply chains.
IBM and Red Hat replied with Lightwell, a $5 billion project to fix open-source software vulnerabilities using artificial intelligence and more than 20,000 developers. The commercial edition, which IBM claimed started July 8, features a portfolio of more than 6,500 remediated and verified software dependencies.
IBM's Lightwell demonstrates it's trying to turn AI disruption into a new business opportunity.
Investors still need evidence that Lightwell can generate business quickly enough to offset delayed software transactions.
What IBM investors should watch after the crash
IBM will issue its comprehensive second quarter results and outline its full-year expectations on July 22. The investor-relations schedule on the company's website shows the earnings webcast at 5 p.m. Eastern.
The first is whether management stays on track with its revenue and free-cash-flow guidance.
Earlier, IBM had expected revenue to climb by more than 5% in constant currency and to generate about $1 billion more in free cash flow than it did in 2025. A cut to either target would indicate that the second-quarter downturn isn't just a matter of deal timing.
A pick-up in transaction processing and completion of delayed contracts would confirm IBM's view that consumers essentially shifted purchases from one quarter to the next.
Any such deterioration would suggest a substantial shift in corporate expenditure priorities.
Investors should also monitor gross margin.
If clients keep buying low-margin infrastructure rather than software, IBM could generate more revenue from servers and storage while producing weaker margins if customers continue postponing higher-margin software purchases.
Customers redirected technology budgets toward supply-constrained servers, storage and memory.
Software carries a substantially higher gross margin than IBM's infrastructure and consulting businesses.
Distributed infrastructure rose 37%, showing that IBM captured some of the hardware demand.
Weak IBM Z sales also reduced associated transaction-processing software revenue.
The July 22 earnings call should clarify whether the problem is temporary timing or a structural budget shift.
The final question: Can IBM's recent acquisitions and fresh products get the wind back in its sails?
Krishna stated Red Hat sped up and HashiCorp and Confluent did well. IBM's 2025 annual report outlines a bigger push to focus its portfolio on hybrid cloud, automation, analytics, and artificial intelligence.
This array of assets gives IBM various opportunities to take advantage of enterprise AI adoption.
They don't promise that consumers will buy IBM software before they get the CPUs and memory to run it.
And that difference is why the market has punished IBM so badly.
But the corporation didn't only declare reduced growth. It showed that the AI boom may throw a wrench in the timing and profitability of its own business, even as management keeps calling artificial intelligence a long-term tailwind.
In the best-case scenario, clients proceeded with hardware purchases and will return for deferred software contracts later in 2026.
The most serious interpretation is bearish.
Enterprise AI spending might be shifting to infrastructure faster than IBM's high-margin software offerings can capture. Newer models are also threatening parts of the conventional software and consulting business.
We'll have to wait for IBM's next earnings call to see which one is more accurate.
The stock's unprecedented plunge revealed the risk.
Artificial intelligence does not automatically benefit every technology company. In IBM's case, the buildout encouraged customers to prioritize scarce infrastructure over the higher-margin software purchases IBM expected.
IBM's future returns for shareholders may rest on making sure software doesn't come second.
Four leading AI models discuss this article
"AI infrastructure spending is structurally prioritizing lower-margin hardware over IBM's high-margin software, pressuring profitability beyond one disappointing quarter."
IBM's 25% plunge on July 14 after soft prelim Q2 (revenue +1% to $17.2B, software +5%, infra -7%, EPS +5% to $2.93) highlights real risks from AI-driven budget shifts toward scarce hardware (servers, storage, memory) at the expense of higher-margin software (82.8% gross margin vs 56.9% infra, 27.5% consulting). The z17 mainframe miss cascaded into transaction-processing software weakness, and Lightwell's open-source remediation push is too nascent to offset near-term margin pressure. Guidance reaffirmation on July 22 is critical; any cut would confirm structural reallocation away from IBM's core valuation drivers.
The article glosses over IBM's record distributed infrastructure growth (+37%, $500M backlog) and strong Red Hat/HashiCorp/Confluent performance, suggesting the "trap" may be a one-quarter timing mismatch after z17 pull-forward rather than permanent software displacement; if deferred deals close in H2 and watsonx gains traction in regulated AI deployment, margins could rebound sharply.
"IBM's valuation is currently backed by a software-growth narrative that is being structurally undermined by a hardware-heavy AI spending cycle that compresses gross margins."
The market's visceral reaction to IBM's Q2 miss reflects a fundamental mispricing of the 'AI transition' narrative. While the article frames this as a temporary budget reshuffling, the structural reality is that IBM is attempting to pivot to high-margin software while tethered to a legacy infrastructure cycle that is currently being cannibalized by the GPU-centric hardware arms race. With IBM Z (mainframe) demand faltering and consulting margins under pressure at 27.5%, the 'AI software' thesis is being deferred indefinitely. Investors are right to punish the stock; until IBM can prove that watsonx and Red Hat can decouple from hardware-dependent transaction cycles, the valuation premium for a 'software-first' firm is unsustainable.
If the 'AI infrastructure' spending phase is as short-lived as the historical semiconductor cycle, IBM’s massive installed base of enterprise clients may return to high-margin software upgrades as soon as their data centers reach a temporary capacity equilibrium.
"IBM's Q2 miss is a timing problem masquerading as a structural one until July 22 guidance clarifies whether deferred deals actually close in H2 or whether customers have permanently reallocated budgets away from software."
The article frames IBM's miss as a budget-sequencing problem—customers buying constrained hardware before software—but conflates two distinct risks. The timing thesis (deals slip one quarter) is testable on July 22; the structural thesis (AI infrastructure permanently cannibalizes high-margin software demand) is not yet proven. What's missing: IBM's guidance cut magnitude matters enormously. A 1-2% revenue guide reduction signals timing; a 5%+ cut signals structural. Also: Red Hat +11% and distributed infrastructure +37% show IBM IS capturing some AI hardware tailwinds, yet the article treats this as a trap rather than a partial offset. The real question isn't whether AI helps IBM—it's whether the margin compression from mix-shift is temporary or permanent.
If enterprise budgets are truly constrained and AI infrastructure is genuinely supply-limited, customers may rationally defer software indefinitely rather than pay premium prices for delayed delivery. IBM's historical software stickiness doesn't guarantee it survives a structural repricing of IT priorities.
"IBM's value hinges on converting AI infrastructure wins into durable, high-margin software and services revenue, not merely hardware orders."
IBM’s stock tumble after July 14 is plausible given the cash-flow mix shift toward hardware, but the takeaway that AI spend is a permanent drag misses nuance. The hardware cycle could be a timing issue; IBM is betting on software and services to monetize AI across its z/Power platforms (z17, Red Hat, automation, data products) and on AI-enabled security with Lightwell—the longer a client uses IBM tools, the more revenue on higher-margin software and managed services. Backlog in distributed infrastructure shows sustained demand even as software deals lag. The real test is whether IBM can convert AI infrastructure wins into recurring, margin-rich software revenue rather than case-by-case hardware orders.
The strongest counter is that even with healthy hardware demand, IBM may fail to translate AI infrastructure wins into durable software margin expansion. Customers could migrate to specialized AI platforms, and Lightwell may take years to monetize, leaving earnings at risk if software deals lag.
"z17 miss is cycle-specific timing, not proof of permanent software cannibalization by AI hardware."
Gemini and Claude both treat the GPU arms race as structurally cannibalizing IBM's software margins, yet ignore that z17's -7% infra drag stems from a specific mainframe refresh cycle pull-forward—not permanent displacement. Red Hat's +11% and $500M distributed backlog already demonstrate conversion from hardware tailwinds into software. A modest July 22 guide cut would confirm timing over secular risk.
"IBM's consulting segment is masking structural margin erosion that cannot be fixed by a mainframe refresh cycle."
Grok and Gemini are missing the elephant in the room: the consulting segment. At 27.5% margins, consulting is the anchor dragging down IBM’s enterprise value. Even if software recovers, the 'AI-led' consulting boom is proving to be low-margin labor arbitrage rather than high-value implementation. Unless IBM proves that watsonx deployments are driving software-led recurring revenue rather than just billable hours for consultants, the stock remains a value trap regardless of z17 timing.
"Consulting margin weakness matters only if watsonx fails to establish recurring, non-labor-dependent revenue streams."
Gemini's consulting margin critique is sharp, but conflates two problems. Yes, 27.5% is low—but that's a legacy drag, not an AI-specific failure. The real test: are watsonx deals *replacing* low-margin consulting labor, or *adding* to it? If IBM is bundling AI software into existing consulting engagements without repricing, that's a margin trap. If watsonx becomes a standalone recurring product sold to existing clients, consulting margins stay depressed but overall software mix improves. July 22 guidance will clarify whether consulting is a structural anchor or a temporary headwind masking software strength.
"IBM can turn AI deployments into recurring software-enabled services with higher margins, not just consulting hours."
Gemini pins IBM's 27.5% consulting margins as a permanent drag. My read: that assumes no monetization of watsonx as recurring software-enabled services paired with managed outcomes. If IBM prices AI deployments as annuities (software plus ops), gross margins could stabilize or improve even with lean consulting. The risk to this thesis is execution: stall adoption or mispricing, which would still leave hardware-driven volatility intact and could drive multiple compression.
IBM's Q2 miss and subsequent stock plunge highlight concerns about its ability to transition to high-margin software while being tied to a legacy infrastructure cycle. The key question is whether the margin compression from mix-shift is temporary or permanent.
Converting AI infrastructure wins into recurring, margin-rich software revenue
Permanent cannibalization of high-margin software demand by AI infrastructure