On July 14, IBM had its worst trading day since 1987. The stock fell about 25 percent and lost roughly $65 billion in market value after the company warned that preliminary second-quarter revenue had come up short: $17.2 billion against expectations near $17.9 billion.
IBM gave more than one reason for the miss, and both are worth taking seriously. The company pointed partly to execution, large deals that slipped past the close it expected, which CEO Arvind Krishna said drove the majority of the shortfall.
But he named a second cause that deserves every finance leader’s attention. In the final weeks of June, IBM’s customers moved capital toward servers, storage, and memory, racing to secure supply-constrained AI infrastructure before prices rose. IBM’s software revenue grew about 5 percent. Its infrastructure revenue fell about 7 percent. The company, Krishna wrote to investors, “anticipated some supply chain related impact” but “did not anticipate the magnitude of the capex reprioritization.”
The real signal is underneath the stock drop
The stock drop is the surface. The capital move underneath it is the real signal. Budgets are shifting into AI fast enough to reshape a quarter before the finance teams involved can even model the impact. They are moving fast enough that IBM, watching it happen in its own numbers, still misjudged the size.
The companies doing the reallocating are not careless. They run disciplined capital allocation. Every major outlay is underwritten against an expected return and reviewed on a regular cadence. A new data center or an acquisition would get due diligence before the money moves and would be measured after it does. AI spend, so far, is the exception. It is being committed at the scale of those decisions, but without the discipline that governs them.
Report
Finance needs to prove AI’s return: CloudZero report
260 senior finance leaders (more than half CFOs) told us why the speed of seeing AI spend, not the size of it, separates who pulls ahead on AI from who gets burned.
AI spend without the diligence
The reason is ordinary. However the spend reaches finance, whether as a cloud bill or an internal chargeback for GPUs the company runs itself, the record shows what was spent. It does not show what the spend produced, or which part of the business it served. The questions a board actually asks, what did we spend, what did it produce, and what was it worth, cannot be read off a cost report. Most finance teams cannot answer them while the quarter is still live.
Even full visibility isn’t enough
And there is a harder problem beneath visibility. Suppose the spend were fully visible, mapped to every product and customer it served. That tells you what the money bought. It still does not tell you whether the money caused the result, or whether the same result would have come anyway. Separating the two is a question of experimental design, not accounting. It is exactly the question a finance leader is trained to ask of any investment. On AI spend, most organizations cannot ask it yet, because the underlying data was never captured to support the question.
This is why “AI cost control” understates the problem. Cost control trims a line you already understand. This is a capital-allocation decision the size of an acquisition, made without the diligence an acquisition receives. The gap is easy to document. Only 14 percent of CFOs report a clear, measurable impact from their AI investments (RGP, December 2025), and 80 percent of enterprises miss their AI infrastructure forecasts by more than 25 percent (Mavvrik / BenchmarkIT, 2025).
IBM’s miss is this week’s headline. The capital shift underneath it is not going away. It is already moving through enterprise budgets that have not yet reached an earnings call or a board meeting. Boards are going to ask finance to show the return on AI, and the reallocation is already underway whether or not the reporting can see it. The only question is whether you will be able to answer, by outcome and in time to act, or whether you will read about your own AI spend the way the market read about IBM’s, once the quarter is already gone.
