A record no vendor wants
On July 14, 2026, IBM pre-announced a second quarter that fell short and watched its stock drop about 25%, the steepest single-day decline in the company's history. The move erased roughly $70 billion in market value and cut deeper than Big Blue's 23.7% fall on Black Monday in October 1987. IBM now expects Q2 revenue near $17.2 billion against a consensus of about $17.86 billion, with adjusted earnings of $2.93 a share versus the $3.02 analysts modeled. Revenue growth of roughly 1% would be the weakest in more than a year, and full results land on July 22.
We read a warning like this as a data point about the whole enterprise stack, not one vendor's stumble. IBM sells across infrastructure, software, and consulting, which makes it a rough proxy for where corporate technology money flows in a given quarter. When a company with that spread says it faltered against a spending shift, the message is that budgets moved, and moved fast. Chris Beauchamp, chief market analyst at IG Group, called it 'an ugly moment for IBM and software stocks' and framed the open question as how long the rotation toward infrastructure and cybersecurity lasts. That question now belongs to every buyer.
Where the money actually went
Arvind Krishna, IBM's chief executive, was specific about the cause. In the last few weeks of June, he said, clients shifted quarterly capital spending toward servers, storage, and memory to secure supply-constrained infrastructure ahead of expected price increases. That reprioritization pushed numerous large deals out of the quarter. Krishna also pointed to rapidly evolving, industry-wide cybersecurity concerns drawing more attention and dollars. In plain terms, procurement teams chose to buy the racks now and defer the software and services engagements that usually ride alongside them, because hardware supply looked scarce and prices looked headed up.
This is the detail that should shape your own planning conversations. AI infrastructure and enterprise software increasingly draw from the same capital pool, and in a supply-constrained moment hardware wins the timing contest. Within IBM's own portfolio the split was visible: Red Hat revenue growth accelerated to about 11% and recent acquisitions such as HashiCorp and Confluent performed well, while infrastructure transaction-processing software and consulting lagged. The lesson for a CIO is that a board eager to fund GPUs and memory may quietly starve the application roadmap in the same cycle. We would name that trade-off explicitly before it happens by accident.
The contagion moved through software and services
Markets treated IBM's note as sector news within minutes. Accenture fell 2.86% to $134.56 and DXC Technology dropped 5.66% to $9.16 as investors marked down consulting demand. ServiceNow slid about 5.8% to $104.85, erasing roughly $6.7 billion in value, while Salesforce fell 3.0% and Workday dropped 6.5%. The reaction was indiscriminate, hitting workflow, HR, CRM, and services names alike on the theory that a budget rotation toward hardware squeezes all of them. For platforms priced on durable subscription growth, even a hint of stalled deals is enough to reset expectations sharply.
For buyers, this repricing is leverage worth understanding. Vendors whose stocks just fell on demand fears tend to negotiate harder to protect renewal and expansion numbers. If your enterprise agreements with any of these platforms come up in the second half, the macro backdrop favors the customer who can wait. We would use the moment to revisit consumption commitments, seat true-ups, and multi-year escalators rather than accept a standard renewal. The same pressure that spooked investors gives disciplined procurement teams room to consolidate vendors and trade volume for better terms.
Infrastructure and security got the upside
The mirror image of the software selloff was a rotation into the layers clients were funding. Cybersecurity names CrowdStrike and Okta each rose more than 10%, and AI chip suppliers Micron and Sandisk climbed around 5% on the same day. Investors read Krishna's comments as confirmation that spending is concentrating in silicon, storage, and the security controls that wrap AI systems. That is a coherent story about where enterprises feel they cannot wait: capacity to run models, and protection for the data those models touch. Both are being treated as non-deferrable, and the tape rewarded them accordingly.
We would take that signal literally when sequencing our own investments. If the market believes compute and security are the must-fund layers, the safer path is to lock capacity and governance first, then layer applications on top as budgets recover. That ordering also matches how agentic systems actually fail, which is usually at the data-access and control layer rather than the model. Funding the foundation before the features protects you from the exact whipsaw IBM just described, where a sudden hardware bill crowds out everything downstream and leaves half-built programs stranded.
The budget question you now own
The uncomfortable takeaway is that AI capex has become a line item that competes with your existing software estate inside one fiscal envelope. Boards that approved GPU clusters and memory buys in June are the same boards asking why the ERP modernization or the CRM consolidation slipped. IBM just made that tension public. For technology leaders, the job is to stop treating infrastructure and applications as separate approval tracks and to show the board the whole picture, including what gets deferred when hardware jumps the queue. Silence on the trade-off is how roadmaps quietly lose a quarter.
Practically, we would rebuild the 2026 plan around scenarios rather than a single number. Model what happens to your application and services commitments if 15% or 25% of discretionary spend rotates to infrastructure and security on short notice. Identify which programs are genuinely non-deferrable and which can flex. Then renegotiate contracts so a sudden capex demand does not force a penalty-laden breach of a software agreement. The firms that handle this cycle well will be the ones that planned for the rotation, not the ones surprised when the invoice for compute arrives.
July earnings are the real test
IBM's warning turned the coming earnings season into a referendum on whether software growth is durable. The metric to watch is current remaining performance obligations, the booked-but-unrecognized revenue that signals future demand. ServiceNow reports on July 22 with cRPO of about $12.64 billion, and its last quarter showed subscription revenue up 22% and deals over $5 million rising nearly 80%. Now Assist customers spending more than $1 million grew over 130%. If those agentic-AI numbers hold, they argue that monetized agents can offset the budget rotation IBM described. UBS still lifted its price target to $115 from $100.
We are watching whether the agent story translates into bookings that survive a hardware-first quarter. Strong cRPO growth would suggest enterprises are funding agents and infrastructure together rather than choosing between them. Weak guidance would confirm that AI capex is cannibalizing the application layer, which changes how aggressively any of us should commit to multi-year software expansion right now. Either way, the decision facing every buyer is the same: fund the foundation, keep application commitments flexible, and make the board see the trade-off in one view. IBM paid for that lesson in public.
