This makes sense for HPC and ML workloads. Big batch jobs where you are pushing the hardware and having everything local is a clear advantage. Also this company sells hardware so it makes sense for them to have hardware experience.
Still think that for the majority on here, needing to make a physical phone call to their data center team (!!) to rack a server is a nutty proposition.
You think the AWS api is slow? Trying calling Steve.
If you have fixed compute costs after a year, sure, look at pulling some stuff on prem.
No comments yet.