It's hard not to see it from the perspective of "for the person with a hammer, everything is a nail", Azure and already having the capability to build data cent

It's hard not to see it from the perspective of "for the person with a hammer, everything is a nail", Azure and already having the capability to build data centers is their hammer and the nail is a compute-heavy task. There's an element of speculation if whether it will be a lasting wave of the future people will pay for, but they have the levers to pull in case it is. It wouldn't be the first time they or any other company has made a bet so they don't get left out, time will tell if AI becomes a part of the furniture commodity or doesn't prove broadly valuable and fades away after a change of direction. The interesting thing for me is the push for AI hardware as a baseline in consumer chips, right now I'd assume training and inference is done on their servers, but as it moves from a development/exploration phase to regular operations and scales up they want to offload inference to save costs, and probably provide a better experience running a model locally.