IBM's shares slumped 25% on July 15, 2026, the steepest one‑day loss recorded since 1968. The decline followed a warning that, in late June, customers were buying servers and memory instead of renewing software and consulting contracts, according to IBM's Q2 2026 earnings release filed with the SEC and the Financial Post's preliminary coverage.

IBM lost 25% of its market value in a single trading day on July 15, 2026, its worst since at least 1968. The company had told investors that clients spent the final weeks of June buying servers and memory instead of renewing software and consulting contracts. That account came from IBM's Q2 2026 earnings release filed with the SEC and Financial Post's report on the preliminary miss. The company's preliminary revenue came in at $17.2 billion against a Wall Street estimate of $17.9 billion, a shortfall Financial Post attributed to customers shifting capital spending to chips and servers ahead of anticipated price increases. This was not an IBM problem. Software stocks from Salesforce to Workday to SAP fell the same day, while cybersecurity vendors CrowdStrike, Zscaler, Palo Alto Networks and Fortinet rose, according to Yahoo Finance's chart-of-the-day analysis. The mechanism, per Marketplace's reporting, is a finite enterprise IT budget: AI infrastructure demand is pulling dollars toward chips and servers and away from license renewals.

Strategic implications for boards and allocators:

  • Software renewal cycles are now exposed to hardware pricing shocks. A memory or GPU shortage anywhere in the AI supply chain can show up as a revenue miss in enterprise software three quarters later. - Security spend appears ring-fenced even when other software budgets are raided — a signal worth testing against portfolio company renewal data, not assuming. - Capital-intensive turnarounds carry different risk than they did pre-AI. Intel posted a -23.1% GAAP operating margin in Q1 2026 even as revenue grew 7%, according to Intel's Q1 2026 earnings release — growth funded partly by government and strategic capital, not customer cash flow. - "AI book of business" figures are bookings, not revenue. IBM's own $6 billion figure, cited in its Q1 2025 release, predates the crash by fifteen months.

The relevant question for a board reviewing next year's renewal book is what share of it depends on customers who are simultaneously budgeting for hardware price increases.

The mechanism behind IBM's crash is not exotic. When DRAM and HBM tighten, capex gets triaged, and software licenses lose that fight because they are the easiest line item to defer. IBM's own explanation was blunt: customers redirected spending toward servers, storage, and memory in the final weeks of June to lock in prices ahead of expected increases. That pulled roughly $660 million out of the quarter versus the $17.9 billion Wall Street had modeled (IBM's Q2 2026 earnings release). That is a scheduling problem, not a demand-generation one, and the two are easy to conflate from outside the company.

For architects sitting on renewal negotiations, hardware refresh cycles that would normally be smoothed across a fiscal year are compressing into panic-buy windows. Software vendors sit downstream of that compression with no leverage to reschedule it. IBM's Distributed Infrastructure backlog, Power and Storage orders, was growing "at a record pace" even as the company flagged the software-side miss, the same dynamic showing up on both sides of one balance sheet (IBM's Q2 2026 earnings release).

The reaction across the peer set is a decent proxy for which categories of spend are considered discretionary in a memory shortage. Workday fell 9.7%, ServiceNow 8%, Atlassian 8.3%. Salesforce dropped 6%, all in collaboration, HR, and workflow software (Financial Post's coverage of the software selloff).

Security did not fall the same day; CrowdStrike, Zscaler, Palo Alto, and Fortinet rose (Yahoo Finance's Chart of the Day). The market priced security spending as fixed cost rather than the kind of line item a finance team can push to next quarter.

The build-versus-buy calculus this exposes is narrower than the headline suggests. AI is not eating software budgets in general. AI-driven memory scarcity is forcing a specific, front-loaded hardware purchase that happens to land in the same quarter as software renewal cycles. Vendors whose contracts renew on fixed annual terms will keep absorbing that collision until DRAM and HBM supply catches up with hyperscaler demand. Nothing in IBM's own guidance suggests that supply curve bends soon: the company still raised full-year constant-currency revenue growth guidance to four-to-five percent one week after the preliminary miss, betting the reallocation is transient rather than structural (IBM's Q2 2026 earnings release).

IBM's preliminary second-quarter numbers, released July 15, 2026, showed revenue of $17.2 billion against a Wall Street estimate of $17.9 billion, a miss of $660 million. Adjusted earnings per share came in at $2.93 against a consensus of $3.03, according to Financial Post's report on the software selloff. The stock fell as much as 26% that day, its worst single session since at least 1968, per the same report. What makes the move worth dissecting is not its size. It is the explanation IBM's own chief executive gave for it, because that explanation implicates an entire asset class rather than one company's execution.

Arvind Krishna told investors that in the final weeks of June, customers redirected capital spending toward servers, storage, and memory to lock in prices ahead of expected increases, according to Yahoo Finance's account of the crash. That is a company describing its own revenue line as collateral damage in someone else's supply chain. The mechanism, per the same report: rising memory prices, driven by AI data-center demand, have been feeding chip producers' profits for months, and IBM's quarter showed who ends up paying for that.

The peculiar part is timing. Seven days after the crash, IBM's formal Q2 filing raised full-year constant-currency revenue growth guidance to 4-5% and reaffirmed an expected $1 billion year-over-year increase in free cash flow, with Krishna stating IBM is "in the early innings of a structural shift for business," according to IBM's second-quarter 2026 SEC filing. Reconciling a stock-crashing preliminary miss with a guidance raise a week later requires assuming these are two distinct disclosure events: a warning, then a formal release with revised full-year math. The filing does not narrate that bridge, and no source in this reporting does either. The filing itself makes no reference to the July 15 selloff or the customer capex shift Krishna described that week. That gap is worth flagging rather than smoothing over.

What is not in dispute is the breadth of the sympathetic selloff. Accenture fell 7%, ServiceNow 8%, Workday 9.7%, and Salesforce 6%, all on the same day, according to Investing.com's tally of the software rout.

Atlassian fell 8.3%, SAP 5.5%, and Adobe 6.1%, per the same tally. Three in five software stocks fell that Tuesday, per Yahoo Finance's chart-of-the-day analysis, which also noted that cybersecurity names were the exception, rallying even as the broader sector sold off. That split matters analytically: if the market were simply repricing enterprise software as a category, security vendors should have fallen too. They did not, which suggests investors are drawing a line between discretionary software spend and spend treated as insurance.

Vital Knowledge analyst Adam Crisafulli called the update "a devastating blow" to software and services stocks, warning that the capex pivot toward hardware "raises its own set of worries" even for the winners, per Investing.com's report. That second clause is easy to skip past. A shortage that forces enterprises to overpay for capacity is not obviously good news for the chip and memory suppliers benefiting from it either, if the same shortage suppresses the software spending that funds those enterprises' AI ambitions in the first place.

The counterfactual worth taking seriously: this could be cyclical rather than structural. The redirected spending IBM described was explicitly tied to customers front-running expected price increases in a specific quarter, not a permanent reallocation of enterprise IT budgets away from software. If memory supply normalizes, the crowding-out effect could reverse, and the "AI is eating software budgets" framing that Futurum Group's Daniel Newman offered — calling IBM's report "the validation point the market had been waiting for" — would look overstated in hindsight, per Marketplace's report on the sector-wide pinch. Marketplace's reporting also noted that Oracle, Microsoft, and Accenture had already lost between 20% and 50% of their value at points during 2026, before the IBM print, which weakens the case that IBM's quarter was the sole catalyst rather than a confirming data point on a trend already priced in by parts of the market.

None of the sources reviewed establish that Intel is a direct beneficiary of the specific dollars diverted from IBM. Intel's own Q1 2026 filing shows revenue up 7% year-over-year to $13.6 billion, a sixth consecutive quarter of results above expectations per CEO Lip-Bu Tan, alongside a GAAP operating margin of negative 23.1% and a GAAP loss of $0.73 per share, according to Intel's first-quarter 2026 SEC filing. CFO David Zinsner attributed part of that quarter's strength to "unprecedented demand for silicon," which is consistent with a tight memory and compute market, but the filing does not trace that demand to enterprise software budgets specifically. The correlation between IBM's software shortfall and Intel's hardware strength is a plausible mechanism, supported by IBM's own account of where its customers sent their money. It is not, on this evidence, a demonstrated transfer between the two named companies.