On Tuesday IBM lost 25 percent of its value in a single session, its largest one-day fall in a 115-year history that includes the crash of 1987. About $67 billion in market capitalisation vanished, taking a storied name in computing down to just under $205 billion. The cause was not a scandal or a failed product. It was, in a roundabout way, the AI boom eating one of its own.
Chief executive Arvind Krishna laid it out in an unusually candid letter to investors, calling the quarter "worse than our expectations" and admitting "we did not adapt and move quickly enough." Two forces did the damage. First, in the final weeks of June, IBM's customers abruptly redirected their budgets, shifting spending away from software and toward buying servers, storage and memory before prices climbed further. Demand from AI data centers has made that hardware scarce, and clients chose to hoard supply rather than sign the software deals IBM was counting on. Second, Krishna said the release of Anthropic's Mythos model, which the company has said can surface software vulnerabilities before defenders find them, stalled several large deals as customers froze to reassess their cybersecurity exposure.
Set IBM beside ASML the same week and the picture sharpens into a story about who the supply shock rewards. The Dutch company, the only maker of the EUV lithography machines that print the most advanced chips, raised its 2026 revenue outlook to as much as €45 billion, up from a prior ceiling of €40 billion, and said it would expand production capacity by 30 percent a year. Its shares rose. The same demand wave that is starving IBM's customers of budget and components is handing ASML years of visibility.
The shock does not stop at the enterprise. The scramble for memory is now visible in the consumer aisle, where each quarter of the crunch pushes phone prices up. Reporting on Google's coming Pixel 11 notes that the baseline model could rise $100 to $899, and that the Pro models may ship with less RAM than last year, dropping from 16GB to 12GB, specifically because of the global memory shortage. Apple and Samsung face the same squeeze. The chips that data centers are buying by the pallet are the same chips that go into handsets, and there are not enough to go around.
What ties these three stories together is that they are not separate. They are one reallocation of the world's chip supply toward AI infrastructure, seen from three vantage points: the equipment maker at the top of the chain enjoying record orders, the enterprise vendor in the middle watching its customers divert cash to hoard hardware, and the consumer at the end paying more for less memory. IBM reports full results on 22 July, and Krishna will get a chance to explain the recovery plan. But the quarter's real lesson is broader than one company. When capital and components pour into a single boom this fast, the boom's shadow falls on balance sheets that never mentioned AI at all.