“We did not adapt and move quickly enough”: What IBM’s earnings miss says about enterprise AI spending
The New Stack Amanda Caswell
IBM's stock tanked after it warned Q2 revenue will miss forecasts, blaming clients who diverted budgets into AI hardware. The CEO admitted IBM itself was too slow to react.
Based on reporting by The New Stack, Amanda Caswell — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
IBM dropped a preliminary earnings warning on Tuesday, more than a week ahead of its scheduled July 22 report, and Wall Street did not like what it saw. The company now expects $17.2 billion in second-quarter revenue, versus the $17.86 billion analysts at FactSet had modeled, with non-GAAP EPS of $2.93 against a forecast $3.01. Modest growth on paper, but well below expectations, and the stock fell sharply on the news.
CEO Arvind Krishna's explanation is the part worth sitting with. In the final weeks of June, he says, big enterprise clients yanked their quarterly capex away from software and toward servers, storage, and memory — racing to lock in AI infrastructure before prices climb further amid supply constraints. That's a demand-side story about the AI buildout eating budgets that used to go to companies like IBM. But Krishna didn't stop there. He admitted IBM failed to see how big this shift would get, and that internal teams moved too slowly, causing a string of large deals to slip past their expected close dates. Both things happened at once: clients reprioritized, and IBM fumbled the response.
The fallout lands hardest on the software side of IBM's business — middleware, security tooling, data platforms, the high-margin stuff that keeps enterprise systems talking to each other. When companies freeze spending there to fund GPU clusters and storage arrays instead, the integration work doesn't disappear. It just moves in-house. Platform teams end up building their own golden paths and internal developer portals, cobbling together open-source tools like Kafka or Envoy to do what a paid vendor product used to handle. Connecting an old mainframe database to a new vector store for retrieval-augmented generation becomes a hand-built ETL pipeline instead of a licensed integration.
There's a reasonable read here that this is just a phase. Enterprises are still front-loading raw infrastructure — chips, racks, storage — because that's what an AI rollout requires first. The application layer, the agentic workflows, the systems that actually turn compute into business value, come later, and that's usually where software spending picks back up. Whether IBM's software revenue snaps back once that infrastructure phase settles is genuinely unclear, and Krishna's own admission that IBM didn't move fast enough suggests the company isn't confident it can just wait this out.
For now, developers are stuck holding two jobs at once: wiring new AI systems into old infrastructure while doing it with fewer vendor tools and tighter software budgets than they had a year ago. That tension isn't unique to IBM's customers, but IBM's earnings miss is the clearest number yet showing how real it's become.
My take — AI-written commentary, not fact-checked reporting
IBM getting caught flat-footed by its own customers chasing GPUs is almost poetic — a legacy enterprise vendor watching capex flee toward Nvidia and friends while its high-margin middleware business sits idle. This is what happens when infrastructure hype outpaces the boring plumbing work nobody wants to pay for anymore; open-source tools like Kafka become the default not because they're better, but because procurement froze. I'd bet software budgets come back once someone asks what all that hardware actually produced, and IBM's next few quarters will tell us how much patience enterprises really have.
Read more about this at: The New Stack
