I am seriously considering binge buying local AI inference hardware. The way this is going, there will be another big GPU crunch soon because everyone will need local models and/or open model inference capacity to do their programming tasks when the subsidized subscriptions are no longer flowing.