News · 2026-07-14

IBM blames a rival AI model for its earnings miss

IBM missed second-quarter earnings expectations and its chief executive publicly attributed part of the shortfall to a competitor's artificial intelligence model. Revenue came in at $17.2 billion, up 1 percent, with adjusted earnings of $2.93 per share against an expected $3.01, and infrastructure revenue down 7 percent. Arvind Krishna told CNBC that customers are freezing cybersecurity deals because they no longer know what security is worth in a world with frontier AI in it.

Key facts

• Revenue $17.2 billion (up 1 percent); adjusted earnings per share $2.93 versus $3.01 expected; infrastructure revenue down 7 percent.
• Krishna to CNBC: "Mythos is making people pause to say, wait, how much do I need to spend on cyber? They're pausing on new deals until they know."
• A second cause: clients shifted capital spending toward servers, storage, and memory ahead of expected price increases.
• Reported in IBM's Q2 2026 letter to shareholders via the IBM newsroom and in Krishna's CNBC interview.

Corporate earnings calls have been full of AI for three years, but almost always in one of two registers: we are spending on it, or we are selling it. Krishna's comment is a third thing. He is saying that the existence of a capable enough model -- Anthropic's Mythos, which is not IBM's product and which IBM does not sell -- changed his customers' buying behavior badly enough to show up in a quarterly number. The uncertainty did the damage, not the competition.

The logic, once you sit with it, is not strange. Cybersecurity budgets are built on assumptions: this many analysts, this many tools, this many hours of human attention against this much threat. A model that can plausibly automate a large share of a security operations center does not just compete with a product -- it invalidates the arithmetic underneath the purchase order. And a buyer who suspects the arithmetic is about to change does not buy a cheaper thing. They buy nothing, and wait. "They're pausing on new deals until they know," as Krishna put it, is a description of a market that has stopped being able to price itself.

Picture a town that hears a bridge might be built across the river. Nobody knows when, or where, or whether. The ferry operator's problem is not that the bridge took their customers -- there is no bridge. It is that everyone stopped signing annual ferry contracts, because who signs a year of anything when the map might change? The ferry still runs. The revenue does not.

The skeptical read deserves a fair hearing, because it is strong. IBM's infrastructure business fell 7 percent in a quarter with genuinely constrained memory and server supply, and the capital-spending explanation -- clients pulling budget forward into hardware before prices rise -- covers the miss on its own without any AI narrative attached. "A competitor's AI paused our deals" is an unusually flattering way to describe losing, and it has the shape of a story told after the fact. This week also produced a candid counterexample: Mark Zuckerberg told Meta staff that its cuts were about capital expenditure rather than AI productivity -- an admission, as one report on the trend put it, that the AI framing is often applied after the fact.

What argues against the cynical read is specificity. Naming a competitor's model on CNBC is not the safe move; the safe move is "macroeconomic headwinds" and a shrug. Krishna described a mechanism precise enough to be wrong, which is more than most executives risk. And it fits a broader pattern of cost pressure that keeps surfacing from unrelated directions: the Palo Alto Networks chief executive arguing AI pricing needs to fall 90 percent, Ramp's chief executive claiming token spending will reach roughly 1 percent of American economic output, and practitioners on public forums saying plainly that they are not using the best models because of what they cost. Something is unsettled in how this technology gets bought, and IBM is the largest company so far to put a number on it.

The honest caveat: this is one quarter, from one company, with an alternative explanation sitting right next to it in the same letter. A single earnings miss is not a market trend, and Krishna has an interest in the version where IBM's execution is fine and the weather is bad. Whether "AI froze our buyers" becomes a recurring line on other companies' calls -- or quietly disappears next quarter -- is the thing worth watching.

What makes it notable regardless is the direction of the causation. This is not AI capital spending crowding out other budgets, the story of the last two years. It is AI capability uncertainty -- not what the technology does, but what it might do -- appearing as a line item in the results of a company that does not sell it.

Primary source, verified: read the paper →

Ground Truth. (2026, July 14). IBM blames a rival AI model for its earnings miss. Ground Truth. https://groundtruth.day/news/ibm-blames-a-rival-ai-model-for-its-earnings-miss.html

@misc{groundtruth:ibm-blames-a-rival-ai-model-for-its-earnings-miss,
  title  = {IBM blames a rival AI model for its earnings miss},
  author = {{Ground Truth}},
  year   = {2026},
  month  = {jul},
  url    = {https://groundtruth.day/news/ibm-blames-a-rival-ai-model-for-its-earnings-miss.html}
}

Topics: industry · earnings · ibm · enterprise · ai-economics