I think when you spin up open AI in azure, that instance is yours, so I don't believe that would be subject to this order.
The plans scale down far enough that they can't possibly cover the cost of a private model-loaded-to-vram instance at the low end.