I understand apple's push for US manufacturing in general but what do they mean by AI servers? I thought apple's current AI strategy is using other AI models?
I believe they're choosing to run Google models on their hardware.
They are using M workstation class chips for inference on their own blades since Google's models are meant run on TPU's it would not have been difficult to port it.
They also use Anthropic internally (code/marketing/sales) which runs their models on Cerebras so they also seem to be agnostic so runs on the same Apple hardware.
Private Cloud Compute uses their own hardware: https://security.apple.com/blog/private-cloud-compute/