IBM’s 25% Plunge Puts Its AI Strategy on Trial | Magica
IBM’s 25% Plunge Tests Whether Its AI Miss Was Only a Delay
Editorial Team
••📖9 min read
IBM says customers diverted late-quarter budgets to scarce AI hardware. Its own product results show why investors now need proof that delayed mainframe and software deals were not lost priorities.
Preliminary second-quarter revenue of $17.2 billion was about $660 million below the analyst consensus compiled by LSEG; Software grew 5% and Infrastructure fell 7%.
IBM says customers redirected late-June capital spending to scarce servers, storage and memory, but management also says its own failure to adapt caused most of the shortfall.
Strong Distributed Infrastructure growth and IBM’s cumulative z17 indicators cut against a simple demand-collapse story. The decisive evidence will be whether slipped mainframe and software transactions close and whether full-year targets survive.
IBM’s record 25% one-day share-price fall turned a quarterly warning into a test of the company’s strategy. The issue is no longer whether the AI buildout can disrupt enterprise technology budgets; it is whether IBM’s software and mainframe products can keep their place in those budgets when hardware becomes scarce.
A modest profit miss exposed a much larger growth problem
In a preliminary regulatory filing, IBM said second-quarter revenue was expected to be $17.2 billion, up 1% from a year earlier. Adjusted earnings per share were expected to rise 5% to $2.93. The LSEG consensus, as reported before the full results, was $17.86 billion and $3.02 a share.
The headline gap was therefore roughly $660 million of revenue and 9 cents of adjusted earnings per share. The segment results show why the warning carried more weight than those differences alone.
Anthropic has put Bun’s Rust port into Claude Code ahead of Bun 1.4’s general release, creating a real but tightly controlled proving ground for an AI-led migration whose total cost and broader reliability remain unsettled.
Editorial Team
Up 10% estimate cited by Evercore ISI
Consulting revenue
Flat as reported; up 1% at constant currency
No comparable estimate in the retained sources
Infrastructure revenue
Down 7%
Down 3% estimate cited by Evercore ISI
GAAP EPS
$2.27; down 2%
No comparable estimate in the retained sources
Operating (non-GAAP) EPS
$2.93; up 5%
$3.02 adjusted-EPS LSEG consensus
The Software and Infrastructure estimates came from Evercore ISI and were cited in market reporting; they are not part of the LSEG consensus set. Keeping those bases separate matters: the table locates the miss but does not pretend that every figure came from one standardized forecast.
Nor was this a profit collapse. IBM expects its operating, non-GAAP pretax margin to have increased 30 basis points to 19.2%, helped by productivity savings, while GAAP pretax margin fell 90 basis points to 14.4%. Operating, non-GAAP gross margin fell 70 basis points to 59.4%. Those are company estimates, and the filing says the results remain preliminary until IBM reports on July 22.
Investors nevertheless treated the growth miss as a reset. The shares closed at $217.07, down 25%, erasing nearly $69 billion of market value. Dow Jones Market Data, cited in the market account, called it IBM’s largest one-day decline on record. The magnitude does not measure the economic value of one quarter’s revenue gap; it measures how much investors cut from their expectations for later growth.
TrendForce’s March 2026 projections for quarter-over-quarter conventional DRAM and NAND Flash contract-price increases in the first two quarters of 2026. Source: TrendForce.
Hardware scarcity explains the rush but not the product split
Chief Executive Arvind Krishna said customers changed their buying patterns in the final weeks of June. In a letter issued by IBM, he said they redirected quarterly capital spending toward servers, storage and memory to secure supply before anticipated price increases. He also said industry-wide cybersecurity concerns distracted clients and shifted their priorities.
There is independent evidence for the procurement pressure, but not for its exact effect on IBM. A March pricing survey and forecast projected that conventional DRAM contract prices would rise 58% to 63% quarter over quarter in the second quarter and NAND Flash prices would rise 70% to 75%. It attributed the squeeze to limited supply, suppliers’ shift toward higher-margin HBM and server products, and cloud providers seeking longer-term supply commitments.
Those figures are forecasts, not observed quarter-end prices, and their population is the global contract-memory market—not IBM customers’ budgets. They support an incentive to secure components early. They do not show how much spending moved away from IBM, whether power and data-center capacity also delayed deployments, or whether the money will return on management’s timetable.
IBM’s own product results redistribute the explanation. Distributed Infrastructure revenue rose 37%, which the company called its best performance, and the business ended the quarter with about $500 million of backlog. Yet total Infrastructure revenue fell 7% because of shortfalls in IBM Z and the related Transaction Processing software business. Customers were still buying infrastructure; IBM did not recognize enough revenue from the products that anchor its mainframe economics.
The company also supplied evidence against a broad mainframe rejection. IBM said overall z17 performance was nearly 130% of z16 at the same point in the prior program cycle, and that clients representing 85% of installed mainframe processing capacity had maintained or increased capacity. Those are cumulative, company-defined program indicators, not quarterly revenue growth: IBM did not disclose the denominator behind the 130% measure, and neither figure can be compared directly with the segment’s 7% decline. Red Hat’s revenue growth rate also accelerated to 11%, while recently acquired HashiCorp and Confluent were described as strong performers.
That combination makes Krishna’s admission the most consequential part of the warning:
“We did not adapt and move quickly enough, and numerous large deals failed to close on the timelines we expected, driving the majority of our shortfall.”
The attribution is IBM’s: Krishna made the statement in the company letter, which IBM also furnished to the SEC. It says most of the miss came from expected deals that did not close on schedule; it does not establish that the deals remain committed or quantify how much revenue may move into a later quarter.
Analysts therefore split on the diagnosis. IDC’s Dave McCarthy described an “infrastructure-first” spending phase, and Futurum’s Neil Osnato said capacity, data and power constraints can force companies to sequence hardware before software. Zeus Kerravala of ZK Research argued that the same conditions should not stop a vendor with a differentiated offer from closing deals and called the episode mainly an IBM execution problem. Their comments were collected in reporting on the infrastructure warning. Barclays analysts took the more forgiving view that the timing shift did not yet change the broader thesis.
The miss lands inside an acquisition-funded software pivot
The comparison with IBM’s own recent performance makes the quarter harder to dismiss. In April, IBM reported first-quarter revenue growth of 6% at constant currency, with Software up 8%, Infrastructure up 12% and IBM Z up 48%. Management’s first-quarter prepared remarks still targeted 5%-plus full-year revenue growth at constant currency and 10%-plus Software growth.
Those targets cover a year, not a quarter, and the preliminary 1% revenue figure is reported rather than constant-currency growth. They are not like-for-like comparisons. They do show that Q2’s 5% Software growth and overall slowdown arrived during a promised acceleration, not after management had guided investors to a weak cycle.
The strategy is also closely tied to Krishna. IBM’s 2026 proxy statement says he joined the company in 1990, became chief executive in April 2020 and was a principal architect of Red Hat, the largest acquisition in IBM’s history. The same filing says Software and Consulting made up more than 75% of IBM’s 2025 business mix. In that year, Software revenue rose 11%, or 9% at constant currency, and IBM spent more than $8 billion on 10 acquisitions. These are IBM’s own annual figures, not independent estimates, but they establish how much capital and strategic credibility now depend on software-led growth.
AI makes that positioning more contested, not automatically more valuable. IBM sits between hyperscale cloud providers and more focused AI vendors, a structural squeeze examined in analysis of Krishna’s strategy. Some companies are also capping AI use or testing lower-cost models, according to a UBS analyst cited in the market reporting. Anthropic has said its Claude Code tool could help modernize COBOL, according to a report on IBM’s legacy exposure. That claim does not show that IBM mainframes are being displaced.
The countercase is that IBM hardware, software and consulting are linked. Several industry analysts argued that postponing a mainframe purchase can also postpone the attached software licenses and transformation work; one said mission-critical systems cannot be deferred indefinitely. That would make Q2 a timing shock whose revenue can reappear. But this is still an analyst hypothesis. It becomes evidence only if IBM shows the transactions converting.
July 22 must turn “delayed” into a measurable bridge
The full report cannot resolve every strategic question, but it can distinguish a temporary procurement shock from a loss of budget priority. Investors need four things from IBM on July 22:
The value and expected timing of the large IBM Z and Transaction Processing transactions that slipped, including how many remain active and how many were lost or resized.
A reconciliation of 37% Distributed Infrastructure growth and roughly $500 million of backlog with the 7% decline in the overall Infrastructure segment.
An update to the full-year targets for 5%-plus revenue growth at constant currency and 10%-plus Software growth, with any currency, acquisition and timing effects separated.
Evidence that Software bookings and Consulting signings are converting to revenue, rather than only another cumulative AI “book of business” figure combining the two.
If the deals remain intact and guidance holds, the warning will look like a severe but bounded timing failure. If management cannot quantify the bridge, the stronger interpretation is that scarce AI infrastructure did more than move purchases between quarters: it exposed which parts of IBM’s portfolio customers considered deferrable.
AIdeaLab has released two anime-focused Wan 2.2 video-model fine-tunes under Apache 2.0. The weights are commercially usable and locally deployable, but current benchmarks, production economics and training-data detail remain undisclosed.
Zhipu reportedly reached $1 billion in annual recurring revenue in July, roughly four times a March estimate, but the unconfirmed run rate is not annual sales and still sits far ahead of recognized cloud revenue while margins remain thin.
OpenAI strategy executive Dean Ball says open-weight AI will deter investment and end in state provision. Kimi K3 exposes a nearer contest over API prices, scarce compute, hosting revenue and regulatory barriers.
A competitor identified a concrete case in which MEDLEY-BENCH may aggregate confidence labels attached to different propositions. The retained record does not show whether the problem is widespread or changes model rankings, while Kaggle's defense of its human review does not resolve the technical question.
OpenAI's Codex 0.144.6 sets 272,000-token context metadata for GPT-5.6 Sol, Terra, and Luna after one Sol user documented a brief 372,000-token setting, renewing a dispute over how Codex accounts for input, output, and compaction headroom.
Dave Eggers told OpenAI that ChatGPT could deprive students of their writing voices. A national teen survey and a small essay-writing preprint identify real risks without proving generational harm, while schools and unions are already adopting a broader mix of AI products and training.
Anthropic workers are giving to federal candidates in unusual numbers, but the largest AI-politics balance sheets remain concentrated among executives and super PACs. The comparisons show a new donor network—not a unified bloc, proof of company direction or evidence that spending decides elections.
Reddit says upgraded automated systems block 23 million spam views daily and reduce exposure, but its disclosure does not identify how much caught spam is AI-generated or provide the denominators and error rates needed to judge the filter.
Elon Musk says xAI's 2-trillion-parameter Grok 4.6 will finish initial training next week. The claim sets a checkpoint, not a release date, and leaves too little technical or commercial detail for a sound comparison with Kimi K3.
A review classified 88 of 120 repositories linked to AI-flagged Flathub submissions as abandoned, a stark result that helps explain reviewer frustration but does not measure AI’s effect on software maintenance or the ban’s impact.
Cmpunlocker says patched Nvidia open kernel modules can expose 64GB of HBM on an 8GB CMP 170HX or 40GB on a 10GB card while removing an SM throttle. The exploit mechanism is documented, but the retained evidence does not independently validate the tool across full-memory workloads, error rates, power use or A100-class performance.
An unidentified industry source says at least one Nvidia board partner has physical RTX 50 Super cards but cannot sell them, with 3GB GDDR7 quoted at $60 to $70 per chip. The reported memory bill explains a plausible pricing problem, but neither Nvidia nor a second independent source has confirmed the products, the hold or the component terms.
Christian Faith Madison’s estate alleges GPT-4o sustained a months-long narrative of prophecy, sacrifice and resurrection before her death. Public safety disclosures and controlled research make that failure mode plausible, but neither establishes which model behavior she encountered or whether it caused her death.
Shanghai AI Laboratory has released Intern-S2-Preview-397B as downloadable weights and a hosted API, but its own 35B alternative offers the same context window and a far smaller disclosed scale while the larger model still lacks auditable comparative scores, pricing and physical-validation results.
Mixfont's free Decoy Font gives each glyph a sharp decoy and a blurred intended letter, creating a hurdle for pixel-based readers while leaving selectable text, known-image techniques and accessibility costs outside its protection.
Anthropic’s 20-year lease gives TeraWulf a customer for 401 megawatts of future AI infrastructure, but rent starts only after delivery and the proposed utility deal passes power and grid costs to TeraWulf, leaving financing and project margin unresolved.
OpenAI has put conversations and Projects back in its redesigned ChatGPT desktop app and enabled cloud Work threads to move across devices, correcting the launch's biggest usability failures without merging local Work or Codex histories.
Twenty-nine countries signed an agreement creating WAICO as an independent intergovernmental organization, while China paired the launch with capacity-building offers that are not yet confirmed as WAICO programs.
Meta reportedly plans to put departing AWS compute executive Dave Brown to work on its data-center buildout, adding hyperscale operating experience while leaving any customer-facing cloud business conditional and undefined.
