In the summer of 2026, one of the technology industry's oldest names suffered one of its worst days on the market. IBM shares fell roughly 25% in a single session on July 14, 2026, dropping from a close near $290.23 the day before to about $217.07. ³ Zacks, an equity research firm, described the 25.2% plunge as one of the company's steepest single-day declines in decades. ¹ The fall came not on soft demand for artificial intelligence, but during its opposite: money was pouring into AI at unprecedented scale.

That is the puzzle at the heart of this report. IBM was not a struggling company by its own numbers. Its full-year 2025 revenue grew 7.6% to $67.5 billion, its software business grew more than 10%, and its generative AI book of business - the cumulative value of AI deals signed - passed $12.5 billion. ² Yet its stock collapsed just as market researchers were reporting that corporate AI investment had more than doubled year over year. ¹

The immediate trigger was a set of preliminary second-quarter results IBM released on July 14, 2026, showing revenue of $17.2 billion against a Wall Street expectation near $17.9 billion. ¹ What that miss means - a timing accident, or evidence that IBM was being left behind by rivals - is the question the rest of this report examines.

The phrase '25% crash' can mislead, because IBM's stock moved in several distinct episodes across 2026, and only one of them is the crash. The confirmed event is narrow and specific: a single trading session between the July 13, 2026 close of $290.23 and the July 14 close of about $217.07. ³ That is a decline of roughly one quarter of the stock's value in one day. The arithmetic on the confirmed prices yields a 25.21% fall, matching Zacks' independently reported 25.2% figure. ⁴

The strength of this finding comes from agreement across sources that gathered their data separately. Three financial-data platforms - StockInvest.us, Yahoo Finance and Investing.com - report the same two closing prices, the same roughly 25% magnitude, and trading volume of 66.7 to 67.4 million shares. ³ Zacks, working from its own analysis, arrived at the same 25.2% figure. ¹ When four independent origins converge on the same numbers, the event itself is not in doubt.

Other price figures circulating around the crash are less reliable and should not be confused with it. LiteFinance, a forecasting broker, reported a separate 28.4% decline from a January 2026 price of $296.66 to $212.30 by early May, and separately claimed a 30% one-week drop to $223.78 as of August 2 - a figure graded contested because independent data place prices and timing differently, showing an August 3 close of $226.31 and an August 4 close of $235.15. ² These are different measurements of different windows. The all-time high is also mildly disputed: LiteFinance reports $332.14 on June 2, 2026, while Yahoo Finance and Investing.com both report a 52-week high of $332.46. ³ The gap is 32 cents, most likely the difference between an intraday high and a closing price, or ordinary vendor variance - not a genuine conflict about how high the stock reached. ²

One widely quoted characterization deserves a caution. CNBC, citing commentator Michael Khouw, described the drop as matching a scale of destruction not seen since January 3, 1968. ¹ The drop itself is well corroborated, but that specific historical comparison is one analyst's framing from a single outlet. ¹

Open: The specific closing prices for early May, August 2, August 3 and August 4, 2026, and the precise all-time-high figure, remain either contested or single-source; exchange records would settle them.

If the crash reflected a collapsing business, IBM's financial disclosures do not show it - and this is what makes the July session so striking. Through 2025 and into 2026, the company reported growth on most headline measures. Full-year 2025 revenue reached $67.5 billion, up 7.6% as reported and 6% at constant currency, which strips out the distorting effect of exchange-rate swings. ¹ The recomputation on IBM's own figures confirms the 7.6% growth rate. ¹ Software, the segment IBM has staked its future on, grew 10.6% to nearly $30 billion; consulting was essentially flat, up 1.8% as reported. ¹

The AI story inside those numbers looked healthy too. IBM's generative AI book of business - a cumulative tally of AI-related bookings since inception - passed $12.5 billion by the end of 2025. ² That figure had been building steadily: it stood above $6 billion after the first quarter of 2025 and above $7.5 billion after the second. ³ These are cumulative totals at different dates rather than quarterly revenue, so the growth represents genuine accumulation of new business through the year. ⁴

The quarterly record was more mixed, and this is where warning signs appear in hindsight. In the second quarter of 2025, IBM posted $17.0 billion in revenue, up 8% year over year, and raised its full-year free cash flow guidance above $13.5 billion. ² Yet that same quarter, software revenue of $7.39 billion narrowly missed the $7.43 billion analysts expected, and shares fell as much as 6% in after-hours trading. ¹ First-quarter 2025 revenue grew just 1%, with the infrastructure segment down 6%. ¹ The first quarter of 2026 then beat expectations, at $15.9 billion in revenue and $1.91 adjusted earnings per share, with two sources agreeing on those figures. ¹

These earnings numbers are not all measured the same way, which matters for anyone comparing them. Operating, GAAP and adjusted earnings-per-share figures use different definitions and cannot be lined up directly without reconciliation. ⁴ One reported figure - full-year 2025 EPS of $11.36, up 73.7% from $6.54 the prior year, from FullRatio - rests on a single source and lacks independent confirmation, though the growth rate itself checks out arithmetically. ¹

Open: IBM's final Q2 2026 segment-level revenue breakdown, and its reported GAAP diluted EPS for full-year 2025, are not in the evidence set.

This is the uncomfortable question the numbers cannot cleanly answer, and the honest response is that the evidence is thin on both sides. On one reading, IBM was being outrun by a handful of dominant platforms. Multiple market-research firms place Amazon Web Services, Google and Microsoft as the leaders of the AI-as-a-service market, with IBM named only as a 'key player' outside the top trio. ¹ A separate assessment lists Microsoft, Google, AWS, Meta, IBM and NVIDIA together as dominant in the broader enterprise AI market, while a third estimates that NVIDIA, Microsoft, Google, OpenAI and AWS together held around 45-50% of the generative AI market in 2025. ² IBM does not appear in that leading group.

But every one of those market-position claims comes from a single research firm, with no independent corroboration in the evidence set. ³ They also measure different things - AI-as-a-service, enterprise AI, and generative AI are overlapping but distinct market definitions, so IBM's standing shifts depending on which boundary is drawn. ³ None of them tracks IBM's market share over time, which is what would actually show whether the company lost ground.

The evidence pointing the other way is firmer, though also limited. Gartner named IBM a Leader across seven data and AI-related Magic Quadrant reports in 2025 and 2026 - a third-party assessment, confirmed, that sits awkwardly with the 'also-ran' framing. ¹ IBM also reported, via Mordor Intelligence, that its Watson platform processed more than a billion enterprise customer interactions in 2024, up 40% from the prior year, though that too is a single-source figure. ¹ And IBM's growing AI book of business is real revenue, not a projection. ²

What emerges is a company that is neither a clear loser nor a clear winner in the framing the market applied. IBM appears to have a solid position in AI software and consulting - the layer where it wins Gartner designations and books billions in deals - but a limited presence in the AI infrastructure layer where competitors like Microsoft were buying an estimated 485,000 NVIDIA Hopper GPUs in a single year. ¹ The distinction between selling AI services and supplying AI hardware turns out to be the whole story, as the next section shows.

Open: No source in the evidence set tracks IBM's AI market share over time or provides customer win/loss data against named competitors, leaving the competitive-position question unresolved at its core.

When IBM's stock cracked, the explanations from management, analysts and the financial press converged on a single mechanism - and unusually, IBM's own CEO led with it. Arvind Krishna stated that the company failed to predict that customers would shift their spending away from IBM's products toward servers, storage and memory purchases, made to hedge against further price increases, especially for AI storage. ¹ This is a confirmed account, reported by Bloomberg and corroborated in substance by CNBC and Investor's Business Daily. ¹

Zacks framed the same driver from the outside, reporting that the selloff was driven by management warning that customers are redirecting budgets toward AI infrastructure over conventional software and IT. ¹ The two accounts describe the same event from different vantage points and agree on the substance: enterprise buyers were pouring money into AI hardware, and IBM was not the one selling most of it. The mechanical result showed up in the preliminary Q2 2026 figures, where the $700 million revenue shortfall against expectations - a miss of roughly 4% - was driven specifically by a 7% decline in the infrastructure division. ¹

Analysts then amplified the move, though the detail here rests on a single outlet. Investor's Business Daily reported that at least five Wall Street analysts lowered their IBM price targets after the preliminary results, with the average target among 27 analysts falling to $273.40 from $298.30 at the end of June - a decline of about 8%. ¹ The same outlet reported that Oppenheimer downgraded IBM to a neutral 'perform' rating from its prior 'outperform'. ¹ Both figures are single-source and uncorroborated. ¹

Context for the fall matters. According to CNBC, IBM had risen 28% during 2025 as of late July, against roughly 8% for the S&P 500 - meaning the stock had run up sharply before it fell. ¹ That figure, too, comes from a single source. ¹ A stock that has outperformed the market more than threefold has more room to give back when a quarter disappoints.

Open: The rationale in the downgrade notes from the five analyst firms, and independent confirmation of the $273.40 consensus target and the Oppenheimer downgrade, are not available in the evidence set.

The final piece of the puzzle is whether AI spending genuinely surged while IBM's stock fell, and where that money went. The evidence says spending surged, but nearly all of it comes from single research firms relayed by secondary outlets, so the specific totals should be read as estimates rather than settled facts. Global corporate AI investment reached $581 billion in 2025, up 129.9% from $253 billion in 2024, according to figures from the Stanford AI Index and Quid. ¹ AI infrastructure spending alone hit $318 billion in 2025, more than double the prior year, per IDC, which forecast $487 billion for 2026. ¹ Both figures rest on a single relaying source with no independent corroboration. ¹

The one confirmed spending figure clarifies where the money was going. CNBC reported that Meta, Amazon, Alphabet and Microsoft planned up to $320 billion combined on AI and data centers in 2025, up from $230 billion in 2024 - an increase of roughly 39%. ¹ That spending was concentrated in physical infrastructure. Microsoft alone was estimated to have purchased 485,000 NVIDIA Hopper GPUs in 2024, outpacing Meta's 224,000, according to an Omdia estimate. ¹

These figures measure genuinely different things and cannot be added together. Corporate AI investment, AI infrastructure spending, and total AI spend cover overlapping but distinct scopes; infrastructure is a subset of the broader total. ² Longer-range projections point the same direction: IDC projects global AI spending of $749 billion by 2028, and MarketsandMarkets projects the cloud AI market rising from $80.31 billion in 2024 to $327.16 billion by 2029. ²

The pattern that connects to IBM is the composition of the spending, not just its size. The confirmed and estimated figures alike show the biggest dollars flowing into servers, GPUs and data centers - the hardware layer where hyperscalers and chipmakers dominate. That is precisely the spending shift IBM's CEO said the company failed to anticipate, and it is why an exploding AI market and a crashing IBM stock are not, in fact, a contradiction. ¹

Open: Direct citations to the underlying Stanford AI Index, IDC and Omdia publications - rather than the secondary outlets relaying them - are absent, so the precise spending totals cannot be independently verified here.

Four explanations compete to account for the crash, and they are not mutually exclusive - the strongest reading combines several. The most directly supported is that a specific customer budget shift drove the miss. On this account, enterprise customers redirected IT budgets away from IBM's software and consulting toward AI infrastructure - servers, storage, memory and GPUs - that IBM does not primarily supply. ² This explanation rests on confirmed evidence: the CEO's own statement, corroborating analyst commentary, and the confirmed 7% infrastructure-division decline behind the revenue miss. ³ Nothing in the evidence set contradicts it.

A closely related explanation focuses on the mechanics of the selloff itself: that the preliminary Q2 2026 miss, roughly a 4% revenue shortfall, triggered the single-day crash, with analyst downgrades amplifying it. ² This too is well supported by the confirmed crash and the reported downgrades, though the downgrade detail is single-source. ⁴ These first two explanations are best read as one story told at two scales - what customers did, and how the market reacted.

A third explanation is more sweeping: that IBM has structurally lost the AI platform war to Microsoft, Google, AWS, Meta and NVIDIA, and that the crash was investors finally recognizing its irrelevance. ² This is the reading advanced by the competitive-displacement case. But its footing is weak. Every market-position claim is single-source and REPORTED, and confirmed evidence cuts against it - IBM's growing $12.5 billion AI book, its rising quarterly AI bookings, and its Gartner Leader status across seven reports. ⁴ What would settle it - IBM's AI market share tracked over time - simply does not exist in this evidence set. This hypothesis is plausible but unproven.

A fourth explanation holds that IBM's strong 2025 run had already priced in AI optimism - the stock rose 28% against the market's 8% - so that when the Q2 2026 tailwind stalled, the correction was sharp. ⁴ This is consistent with all the confirmed data and helps explain the violence of a single-day move, though the 28% figure is single-source. The most defensible synthesis: a confirmed customer-spending shift produced a real revenue miss, which struck a stock that had run up substantially, producing an outsized one-day fall that analysts amplified. Whether that shift signals permanent competitive defeat, or a timing mismatch in one quarter, the evidence here cannot resolve.

The evidence forces a narrow but firm conclusion about the event. IBM's stock fell about 25% in a single session on July 14, 2026 - this is confirmed by four independent origins. ⁴ It fell because IBM's preliminary Q2 2026 revenue of $17.2 billion missed a $17.9 billion expectation, on a 7% infrastructure decline, and because customers shifted spending toward AI hardware IBM does not primarily sell, as the CEO himself stated. ³ This chain is confirmed at every link. The apparent paradox in the topic - a crash amid an AI boom - dissolves: the boom was overwhelmingly a hardware and infrastructure boom, and IBM's strength lies in the software and services layer. ⁴

The evidence does not force the stronger conclusion that IBM has lost the AI race. That claim rests on single-source market-research estimates, and it is directly weakened by confirmed facts pointing the other way: IBM's AI bookings grew through 2025, its overall revenue grew 7.6%, and Gartner rated it a Leader across seven data and AI reports. ⁴ On the available evidence, the competitive-displacement reading remains a plausible interpretation rather than a demonstrated fact.

One genuine stalemate stands out. The market treated IBM's Q2 2026 miss as a signal about the company's future, and analysts extrapolated accordingly. ¹ Whether that extrapolation was correct - whether one quarter's spending shift marks a lasting structural problem or a recoverable timing mismatch - cannot be settled from this evidence. Were IBM to report a subsequent quarter showing infrastructure recovery and continued AI-book growth, the timing-mismatch reading would strengthen; were the infrastructure decline to persist across quarters while hyperscaler capex kept rising, the displacement reading would gain force. Both remain open, and the reasoning here rests on the confirmed pattern of one segment declining within an otherwise growing business. ²