AI capex narrative revives

The market stopped treating artificial intelligence spending like a one-quarter trade and started treating it like a construction boom again. Nvidia said on February 25 that quarterly data center revenue hit $62.3 billion, up 75% from a year earlier, and its total quarterly revenue reached $68.1 billion. (nvidianews.nvidia.com) That shift changes what investors listen for on earnings calls. A company can beat revenue by a little, but if it trims plans for new servers, new buildings, or new electricity contracts, the whole artificial intelligence chain feels it. (nvidianews.nvidia.com) The buyers here are the giant cloud companies. Alphabet said on February 4 that its 2026 capital spending is expected to land between $175 billion and $185 billion, after cloud revenue grew 48% and backlog jumped to $240 billion. (abc.xyz) Amazon gave the same message in a different form. Its February 5 results showed Amazon Web Services revenue up 24% to $35.6 billion and operating income up to $12.5 billion, which is why investors keep reading Amazon’s spending plans as a direct signal for how fast more artificial intelligence capacity is being installed. (ir.aboutamazon.com) Microsoft has been saying the bottleneck is not demand but supply. In its January 28 fiscal second-quarter materials, the company said capital spending would move around with the timing of cloud buildouts, which is another way of saying the pace depends on how fast it can get data centers and equipment online. (microsoft.com) Meta is in the same lane. Meta reported fourth-quarter 2025 revenue of $59.9 billion on January 28, and its investor materials for that quarter kept attention on 2026 infrastructure spending because its artificial intelligence push needs more in-house computing power, not just better ad sales. (investor.atmeta.com) That is why Nvidia sits at the center of the trade. Its chips are the engines inside these new artificial intelligence clusters, and Jensen Huang said customers are “racing to invest in AI compute,” which links Nvidia’s sales directly to cloud companies’ building schedules. (nvidianews.nvidia.com) But the money does not stop at chips. Alphabet said 60% of its 2025 capital spending went to servers and 40% went to data centers and networking, which means every new artificial intelligence buildout also pulls in rack makers, network gear vendors, contractors, and switch suppliers. (abc.xyz) Then comes the part that looks boring until demand spikes: power and cooling. A warehouse full of graphics processors draws far more electricity and throws off far more heat than a normal corporate server room, so utilities, transformers, liquid cooling systems, and backup power equipment become part of the same earnings story. (abc.xyz) The new question for the next round of earnings is not “did revenue beat by 2%.” The new question is whether Microsoft, Amazon, Alphabet, and Meta still sound like companies trying to catch up with demand, because if they do, Nvidia and the rest of the artificial intelligence infrastructure chain keep looking less like a fad and more like a multi-year build cycle. (microsoft.com)