IBM 25%|AI Capex Shift Breaks a Software Giants Core

The Day IBM Collapsed

IBM stock plunged 25% on July 14, 2026, its single worst trading session in over half a century, after CEO Arvind Krishna released a pre-earnings letter to investors admitting the tech giant had faltered. The company reported preliminary Q2 revenue of $17.2 billion — up just 1% year over year but $600 million short of the $17.8 billion analysts had expected, with EPS of $2.27 against a $2.60 consensus.

The miss had a specific cause: Krishna said IBM's largest corporate customers abruptly shifted spending toward servers, storage, and memory to secure AI infrastructure ahead of expected price increases. This was not a demand collapse — it was a budget redirection. IBM's own software segment still grew 5%, but that was not enough to close the gap the hardware binge opened.

The damage radiated outward. IBM's fall alone subtracted more than 400 points from the Dow Jones Industrial Average. Salesforce, Adobe, Intuit, and ServiceNow each dropped between 2% and 6% as investors asked whether IBM's problem was contagious. The question the market opened — is this IBM or is this all of enterprise software — is the question this video answers.

IBM's Hidden Split: Winners and Losers Inside the Same Company

Here is the fact the sell-off obscures: IBM is not uniformly losing the AI era. Its server and storage business outside of mainframes surged 37% in Q2 as clients snapped up exactly the hardware they were redirecting budgets toward. Red Hat, the open-source software unit, grew 11%. The collapse was concentrated in one place: IBM's flagship mainframe line, whose revenue fell 7%, and the related high-margin software that runs on it.

The paradox tightens when you follow the money: the customers who pulled budgets from IBM's software were in many cases buying IBM's own server hardware to replace it. The capital did not leave IBM's address book — it shifted from the high-margin mainframe and software stack to the lower-margin infrastructure stack. What the market priced as an exit from IBM was, in part, a migration within IBM's own product catalog.

The structural problem is that mainframes carry IBM's best margins. A 7% mainframe decline is not noise — it reflects enterprise CIOs making a generational bet that AI-native infrastructure replaces the batch-processing architecture mainframes were built for. The quarter's second driver compounded the damage: Anthropic's Mythos AI model, released this spring, surfaced thousands of high-severity network vulnerabilities, triggering a corporate cybersecurity scramble that pulled client attention — and budget — away from planned IBM projects. CrowdStrike surged 12% and Okta jumped 11% on the same day IBM fell 25%, mapping exactly where that spending went.

IBM-Specific or Sector-Wide: The Split That Matters

The $400-point Dow drag and the 2% to 6% sympathy falls in Salesforce, Adobe, Intuit, and ServiceNow forced a sector-wide reassessment. The question is whether IBM's miss signals a broad SaaS implosion or a company-specific failure. The pool carries both readings — not as analyst spin, but as directly conflicting source-grounded positions.

Kirk Materne at Evercore ISI argued the miss is largely tied to IBM's mainframe business and that applying it to the broader software sector is "a bit of a stretch." AIP Asset Management's CEO took the opposite view, calling IBM a "canary in the coal mines" and arguing that any enterprise software company not adapting fast enough to AI faces the same fate. Susquehanna's Jamie Friedman held a middle position: IBM-specific, but worth watching Accenture and Cognizant for signs the enterprise spending shift is spreading.

The internal IBM data offers a clue for resolving the disagreement. Red Hat — open-source, cloud-native, AI-adjacent — grew 11%. The mainframe and proprietary software stack fell. The distinction maps cleanly onto adaptability: IBM's AI-era tools are growing; its legacy lock-in products are shrinking. A Bloomberg report circulating before the earnings letter said Starbucks was exploring replacing some IBM software with AI-native alternatives, precisely the client behavior Krishna's letter described. This is not an industry-wide SaaS collapse — it is a legacy-technology purge inside enterprises that have decided AI infrastructure comes first.

The Evercore read holds for the right reason: Salesforce and ServiceNow do not depend on mainframe-era batch-processing margins. Their 2%-to-6% sympathy drops reflect fear contagion, not fundamental exposure to the same budget reallocation IBM hit. If the IBM-specific thesis is correct, those names snapped back as soon as investors processed what the actual IBM segment data showed. The resolution of that question is what the final checkpoint tests.

July 22 and the Two-Outcome Map

IBM's full Q2 earnings call on July 22 is the nearest hard checkpoint. The preliminary results stated the revenue and EPS figures, but they did not give segment guidance or management's read on whether the mainframe compression will continue into Q3. That is what July 22 delivers. The question is not whether IBM missed — it did. The question is whether Krishna describes the mainframe erosion as a one-quarter capex disruption that reverses once clients finish their hardware purchases, or as a structural shift with no recovery path on the mainframe line.

For a holder sitting on a 25% one-day loss, the decision posture is to hold through July 22 only if the thesis is that clients finish their hardware buying surge and resume IBM software contracts in Q3. The signal to watch is not the July 22 EPS print — it is management's language on whether mainframe client pipelines are intact. Pipeline intact means a one-quarter disruption. Pipeline eroding means the Starbucks scenario — AI-native replacements permanently taking IBM's legacy software budget. For a watcher considering an entry after the crash, the entry trigger is mainframe revenue stabilization, not the July 22 date itself. A stabilizing pipeline at $17.2B revenue at a 25%-lower price is a different value proposition than IBM at its June record high. The move becomes an entry setup if management confirms clients are returning to the software pipeline post-hardware surge; it becomes a trap if IBM guides Q3 revenue below $17.2B, signaling the 27% year-to-date decline has not yet found a floor.