IBM’s results disappoint; CEO Arvind Krishna reveals which part of its software could be replaced by AI

IBM is facing a difficult period: its shares suffered their biggest drop in decades, losing over $70 billion in market capitalization after CEO Arvind Krishna warned investors that the company would miss expectations in its latest earnings report. This has unsettled investors regarding the tech giant’s future in a landscape dominated by increasing AI tools. Nevertheless, Krishna believes AI will not harm the business; on the contrary, he maintains it could act as a growth driver for the company.

At a time when fears exist that AI tools like Claude Cowork might signal the end of Software-as-a-Service (SaaS) companies, Krishna states that only about 2% of IBM’s software could realistically be replaced by AI tools. “The rest of our software helps people prepare for AI: it unlocks real-time data, reduces management costs and complexity, and operates across the hybrid infrastructure used by most of our clients,” he told a famous publication.

Krishna asserted that AI could act as a tailwind propelling IBM’s growth, much like a tailwind accelerates an aircraft. He explained: “And since this would be what we might call infrastructure software—rather than applications—I believe it will be a growth driver for us.”

In its latest earnings report, IBM posted revenue of $17.2 billion and gross profit of $9.9 billion—figures that fell short of Wall Street expectations. To put the concerns of IBM investors regarding the impact of AI on the SaaS sector into context: for every dollar the company earns from its mainframe business, it generates three dollars from software. In other words, if AI were to disrupt the company’s software business, it could suffer significant revenue losses. However, Krishna believes IBM can leverage AI as a growth opportunity.

Some IBM software programs are at risk of being replaced by AI.

This does not mean IBM will remain unaffected by AI. Krishna cited IBM’s Tririga lease management software—used by Starbucks in a deal worth $2 million annually—as an example. He explained: “That is a significant component of the 2 percent I mentioned, and I truly believe software of that type is exposed to risk.”

Krishna and IBM CFO Jim Kavanaugh noted that this weakness reflected a delay in spending rather than a drop in demand. The company argued that its customers were choosing to invest first in servers, storage, memory, and other data centre equipment as costs associated with artificial intelligence rose.

IBM now projects software revenue growth of between 6% and 8% for the year—lower than the double-digit growth rate forecast in January.

A letter Arvind Krishna sent to investors last week warned of revenue weakness in the company’s infrastructure business and margin pressure; shares fell 25%, marking what the company described as its steepest single-day drop. Year-to-date, IBM shares have lost nearly 30% of their value.

However, according to IBM’s CEO, the company is on the verge of closing deals that did not materialize in the previous quarter. “Most of what didn’t happen in the second quarter involved large capital expenditure (capex) contracts with major clients,” he stated. Krishna added that about a third of those deals had already closed in the first three weeks of the new quarter, offering IBM “an indication—though not yet definitive proof, certainly a good indication—that it was a postponement rather than a permanent loss.”

Krishna estimates that nearly 75% of the deals postponed after the second quarter should be finalized before the year ends. He also stated: “We see no signs that customers are abandoning mainframe systems.”

These new concerns arise amidst general scepticism regarding software sector stocks, as investors question whether the more powerful AI tools from companies like Anthropic and OpenAI could disrupt current business models. The iShares Expanded Tech-Software Sector ETF has fallen 17%. In February, IBM shares dropped 13% after Anthropic released information about Claude Code’s ability to modernize code written in COBOL, a language frequently used in mainframe systems.