Or maybe its all about compute availability like they say. It could be that they plan to start training a new model on the 22ed, so the amount of compute availa

Or maybe its all about compute availability like they say. It could be that they plan to start training a new model on the 22ed, so the amount of compute available for inference will be greatly reduced.