{"text":[[{"start":6.25,"text":"Spending on data centres has been one of the main things powering a historic boom in IT spending. Now, as the scramble for increasingly scarce resources gets intense, it is becoming the only thing."}],[{"start":18.1,"text":"That is one reading of an earnings warning on Tuesday that sent shares in IBM into a tailspin. If chief executive Arvind Krishna is right to pin the blame on wider market forces, rather than merely mistakes by his own company, then it could point to a new phase in an AI infrastructure boom that rests on increasingly narrow foundations."}],[{"start":37,"text":"IBM stepped in ahead of the second-quarter tech earnings season to reveal that its revenue grew only 1 per cent in the latest period, lower than the 5 per cent analysts had been expecting. According to Big Blue, customers last month began diverting a significant part of their regular IT spending towards the sort of server, storage and memory equipment that has been in high demand in the AI boom, hitting sales of its own mainframe computers and software."}],[{"start":64.05,"text":"The cyber security panic caused by Anthropic’s Mythos model may also have thrown IT spending plans for a loop: As companies deal with the threat, according to IBM, other priorities have taken a back seat."}],[{"start":76.2,"text":"The 25 per cent drop in IBM’s stock price on a relatively modest earnings warning attests to investors’ extreme nervousness. But it also points to a fear that the pattern of IT demand is shifting more significantly."}],[{"start":89.55,"text":"Until now, there hasn’t been much evidence to suggest that AI has been displacing other forms of IT spending. The devastation to software stocks known as the SaaSpocalypse mainly reflects a fear that AI will eventually reduce the need for much of today’s software, rather than any evidence that it is having an impact already."}],[{"start":109,"text":"After the first quarter of this year, IT research firm Gartner reacted to soaring data-centre spending by simply raising its IT spending forecasts for 2026. There was a general expectation that AI would add to overall tech demand, rather than causing cutbacks in other forms of IT spending."}],[{"start":127.75,"text":"IBM’s warning suggests things may be changing. One big question that will be on Wall Street’s mind as the latest earnings season gets under way is how far the infrastructure boom has already gone in crowding out regular IT spending elsewhere. The other is whether spiralling infrastructure costs will raise the price of — and limit demand for — new AI services."}],[{"start":149.85,"text":"Looking for other potential victims, investors reacted to the IBM news by taking another bite out of the shares of software companies including ServiceNow and Adobe. Makers of electronic devices may also start to suffer, as higher component prices for things like PCs, games consoles and smartphones make their products less attractive."}],[{"start":170.6,"text":"For the companies that have been riding the generative AI wave, meanwhile, IBM’s warning highlights new dangers as the competition for resources intensifies. Krishna said his company’s customers were racing to divert more of their budgets into AI equipment before price rises on things such as memory chips took effect. That highlights two risks."}],[{"start":190.45,"text":"One is that timing issues could introduce an extra level of unpredictability into AI infrastructure spending. If, to beat price increases, companies rush to buy equipment before the new data centres to house it have been completed, they may slow spending in future periods as they digest what they have already bought. For a market on high alert for any sign that AI capital spending is weakening, timing considerations like this could introduce a new level of volatility."}],[{"start":219.35,"text":"The other risk is that rising prices will consume an ever larger share of capex budgets. To the extent that more of the billions being poured into data centres are going into the pockets of memory chip investors rather than extra computing capacity, it will limit the supply of AI."}],[{"start":235.9,"text":"Nvidia’s gross profit margin of 75 per cent is evidence of the gains for chipmakers. Likewise the 83 per cent margin that Micron reported in the latest quarter, more than double its level of a year before."}],[{"start":249.05,"text":"Improvements in the performance of the latest generation of AI accelerators and high-bandwidth memory chips will offset some of the effects of rising prices, and the cost of producing tokens — the basic unit of output from large language models — may continue to fall. But the competition for computing resources will keep prices for AI services higher than they otherwise would be, at just the moment that demand is starting to take off."}],[{"start":281.6,"text":""}]],"url":"https://audio.ftcn.net.cn/album/a_1784295086_8604.mp3"}