IBM’s Worst Day in 40 Years

There is a question every value investor should ask when a quality business drops 25% in a single session: is the market processing new information correctly, or is it extrapolating short-term pain into a permanently impaired business?

On July 14, 2026, IBM fell further in one day than it did during Black Monday in 1987. Salesforce, Workday, Adobe, and ServiceNow were all swept lower in the opening minutes alongside it. The read-through was obvious, and it was also, in several important cases, wrong.

We ran the Cheap Investor framework across the wreckage. What we found was not one uniformly mispriced sector but a wide dispersion of outcomes hiding beneath a single narrative. Some companies deserved the sell-off. One did not. And IBM itself sits in a third, more complicated category that disciplined investors should understand before writing it off entirely.

What Actually Happened

IBM CEO Arvind Krishna filed an 8-K with the SEC before the Q2 earnings call, pre-announcing results that fell roughly $660 million short of consensus. He described what happened in the final weeks of June: enterprise clients redirected quarterly technology budgets toward servers, storage, and memory to front-run expected price increases on supply-constrained AI hardware. Large software and mainframe deals did not close. “We did not anticipate the magnitude of the capex reprioritization,” he wrote.

The supply-side cause is not a one-quarter phenomenon. High-bandwidth memory requires more than three times the wafer capacity per bit compared with conventional DRAM, and Samsung, SK Hynix, and Micron have all converted production capacity toward HBM because it commands several times the revenue per wafer. 2027 DRAM and HBM capacity from all three major producers is reportedly fully allocated, with customers receiving only 60% to 70% of their initial volume requests. Panic buying tightens supply further, which raises prices, which triggers more panic buying. IBM was the first company large enough to put that loop in a regulatory filing.

The broader context is worth holding in mind. Amazon, Alphabet, Microsoft, and Meta combined are on pace to spend $725 billion on AI infrastructure in 2026, up 77% from $410 billion last year. That capital has to come from somewhere. Enterprise budgets are not expanding at the same rate. When they lurch toward one category, they pull away from another. IBM just showed us, in explicit accounting terms, what that looks like for an affected vendor.

Separating Cheap from Broken

This is where the Cheap Investor framework does its most important work. A stock that has fallen sharply is not automatically a bargain. The first question is always the same: is the problem temporary or structural?

For IBM, the answer is genuinely mixed. IBM Z revenues fell 42% in Q2. Transaction Processing declined 9% at constant currency. These are deals that close in the final weeks of a quarter, involve large capital commitments, and compete for CFO attention alongside hardware procurement decisions. When an enterprise decides in late June to redirect $50 million toward server and storage, a mainframe renewal is exactly what slips.

But not everything slipped. Red Hat revenue grew about 8% year-over-year in Q2, even as the mainframe business deteriorated. Distributed infrastructure delivered 37% growth. Software annual recurring revenue stood at $24.6 billion, up 10% year-over-year. The generative AI book of business had crossed $7.5 billion as of the April quarter. On the July 22 earnings call, Krishna told analysts that roughly a third of the slipped Q2 deals had already closed in early Q3, calling it “deferral not destruction.” Full-year free cash flow growth guidance of approximately $1 billion was kept intact. First-half free cash flow came in at $4.8 billion.

That is not the financial profile of a permanently impaired business. It is the financial profile of a business with an uneven product mix, where one segment is structurally challenged and another is genuinely growing. The market, however, priced it as though every segment had collapsed simultaneously.

The more important question for IBM is not whether Q2 was a timing event. Krishna’s own $8 trillion capex argument raises a harder long-term challenge. He estimated on the Decoder podcast that global commitments for AI computing power have reached roughly 100 gigawatts of capacity, translating to something in the range of $8 trillion in capital spending. Running the interest arithmetic, that buildout requires roughly $800 billion in profit just to service the debt. His conclusion was direct: the revenue required to justify the buildout within any reasonable timeframe does not yet exist. A company whose CEO is publicly warning that the AI infrastructure cycle may be overextended is telling you something about the ceiling on the recovery thesis.

IBM at $214, down 25% from its pre-announcement level, is more interesting than IBM at its prior highs. But it is a show-me situation, not a clear-cut bargain. The Q3 report is the next meaningful data point. Until it confirms that deferred deals closed and software growth stabilized, we hold rather than build.

Where the Market Got It Wrong

The sector-wide sell-off on July 14 was an overreaction. Applying IBM’s specific problem to ServiceNow, Salesforce, and Workday required ignoring the mechanism that caused IBM’s miss in the first place.

IBM’s weakness was concentrated in products that compete directly for CFO-level capital budgets alongside hardware: mainframe computers and large transaction-processing infrastructure. ServiceNow does not sell either. Its Now Platform is embedded in enterprise IT operations, HR workflows, and security management in ways that make renewal largely non-discretionary. You do not defer a ServiceNow renewal the same way you defer a mainframe upgrade. The two decisions live in different parts of the budget and involve different approval cycles.

That distinction became visible eight hours after the IBM sell-off began. ServiceNow reported Q2 results the same evening as IBM’s formal earnings call. Subscription revenues came in at $3.877 billion, up 24.5% year-over-year, beating the high end of guidance. Non-GAAP operating margin reached 29.5%, 300 basis points above guidance. Current remaining performance obligations, which tracks contracted future revenue, grew 21.5% year-over-year in constant currency, matching Q1’s growth rate. The pipeline was intact.

The market sold ServiceNow on IBM’s news and then bought it back on ServiceNow’s own results. That sequence created a brief window where perception and reality diverged. Investors who understood why IBM missed, and who recognized that the cause did not apply to ServiceNow’s business model, had an opportunity that disappeared once the earnings numbers were public.

The Cheap Investor Case: IBM

IBM is the more classic value investor question here: is this a bargain, or a value trap dressed up in blue-chip clothing?

The bull case rests on several durable facts. Red Hat is becoming the operating substrate for hybrid AI deployments across regulated industries, and IBM’s close of the Confluent acquisition moves the company deeper into streaming data infrastructure that AI use cases increasingly need. The company has more than $10 billion planned for quantum computing over the next five years, a long-duration bet that does not need to pay off in the next two quarters to be strategically rational. Free cash flow remains substantial. The dividend has not been cut. Software ARR of $24.6 billion provides a recurring revenue floor that limits downside.

The bear case is that the mainframe and transaction processing businesses are in structural decline that no amount of Red Hat growth fully offsets. Software growth guidance was cut from double digits to 6% to 8% for the year. Bank of America concluded that double-digit software growth is out of reach. Management’s own $8 trillion warning implies that the enterprise technology cycle is mid-disruption, and IBM’s hybrid AI positioning, while credible, places it in the distribution layer of a market still sorting out who actually wins.

The honest assessment is that IBM at current levels is a hold for existing shareholders who bought at higher prices, and a watchlist candidate for new buyers who want to see Q3 confirm the deferral argument before committing capital. That is not a criticism. Patience is a feature of this framework, not a failure mode.

The Cheap Investor Scorecard

What the Market Believes vs. What the Evidence Suggests

The market, on July 14, believed that IBM’s quarter was a read-through for the entire enterprise software sector. Enterprise buyers were redirecting budgets to hardware, which meant software renewals were at risk across the board. That belief was logical but imprecise.

The evidence suggests the exposure is highly specific. Companies whose deals are large, discretionary, capital-intensive, and close in the final two weeks of a quarter are genuinely exposed to the dynamic Krishna described. Companies whose revenue comes from embedded platform subscriptions with largely automatic renewal cycles are not exposed in the same way. The IBM sell-off treated these two categories identically. They are not identical.

The disconnect was measurable and short-lived for ServiceNow. The more durable disconnect may still exist in IBM, where the post-announcement price reflects genuine uncertainty about whether the Q2 deferrals actually close in the second half, and whether the mainframe business faces a structural ceiling that no AI repositioning can fully address.

Disciplined investors should hold both possibilities without collapsing them into a single verdict. IBM is worth watching. ServiceNow proved it does not belong in the same conversation.

What to Watch

For IBM, the Q3 earnings report is the thesis test. If Krishna’s “deferral not destruction” framing is accurate, a meaningful portion of the slipped Q2 deals will appear in Q3 revenue. Software growth needs to hold within the revised 6% to 8% range, and Red Hat needs to maintain or accelerate from its 8% Q2 pace. Free cash flow is the cleanest signal: management’s full-year growth target of approximately $1 billion should remain intact if the underlying business is sound. Any Q3 guidance cut to that figure changes the picture materially.

For the broader sector, watch whether the memory and server shortage extends into fall. If enterprise hardware prices remain elevated and another late-quarter budget reallocation cycle repeats in September, the companies most exposed are those with large, discretionary, quarter-end deal structures. The companies most insulated are those, like ServiceNow, whose renewal economics do not require the customer to make an active spending decision in the final days of a quarter.

One additional name worth monitoring is Micron Technology. The supply squeeze that created IBM’s problem is Micron’s structural advantage. Its entire 2026 HBM output is committed under long-term contracts, the HBM total addressable market is now expected to cross $100 billion in 2027, and Strategic Customer Agreements carry more than $22 billion in total commitments including nearly $18 billion in cash deposits. Micron is not a value situation in the traditional sense. It is a supply-constrained beneficiary of the same dynamic that is pressuring software vendors. Worth understanding even if it does not fit the classic Cheap Investor mold.

The Disciplined Assessment

Value investing is not about buying everything that has fallen sharply. It is about distinguishing companies where the market has over-extrapolated short-term weakness from companies where the weakness reflects something real and lasting. IBM’s Q2 quarter gave us a rare case study in both at once.

IBM itself sits in the ambiguous middle: not broken, not clearly cheap, but worth monitoring closely as Q3 data arrives. Its free cash flow durability, Red Hat growth, and $24.6 billion in software ARR argue against writing it off. Its mainframe exposure, revised guidance, and the structural implications of its own CEO’s $8 trillion argument argue against rushing in. Patient investors who bought on the 25% decline may be rewarded. Those who buy now should understand that the recovery timeline is not short.

ServiceNow proved, on the same evening IBM formalized its miss, that the IBM dynamic does not apply universally to enterprise software. Its subscription revenues grew 24.5%, forward commitments held steady, and AI annual contract value crossed $1 billion. It is not a bargain in the traditional sense. A quality business growing at that pace does not trade at a discount. But the July 14 sell-off temporarily made it available at a price that did not reflect its results, and that is the kind of moment the Cheap Investor framework is designed to identify.

The market sometimes misprice quality businesses. Usually the reason is fear, contagion, or the habit of selling first and reading the earnings later. July 14 was all three. The correction came within eight hours. That is not always the case. Sometimes the gap takes quarters to close. The investors who benefit are the ones who understood why the gap opened in the first place.