“1x DGX H100 (SXM) with 8x H100 GPUs is $460k including the required support. $100k of the $460k is required support.”
Nvidia charges $100k for mandatory support on a $460k box and still has a waitlist. Azure and GCP are effectively out of H100 capacity for big orders, and AWS is close. Rough demand comes to 432k H100s, about $15 billion of GPUs, and some of it is startups reserving chips they don’t need yet. Nvidia gets to pick who gets allocation, which is a nicer position than any customer has.