Hard to move
Generative AI on Vultr: what it becomes and what it costs.
On Vultr, Generative AI becomes A model API vendor, or vLLM on a GPU server at Vultr. Hosted models are a vendor decision; self-hosting open-weight models needs GPU servers.
Effort grade
Hard
No equivalent: redesign it, buy a SaaS, or leave it where it is
Cost after the move, before engineer time
100% of today
No equivalent at the destination; priced as a rebuild or a vendor service at close to today's cost until someone decides.
Engineer time
16 h a month
To run the replacement: patching, backups, on-call. Plus about 15 engineer days for the move itself.
Your own Generative AI line, from your export
Drop one month of your Oracle Cloud cost export and the report prices every service, Generative AI included, on Vultr list prices of 18 Oct 2026, net of the engineer time to run it. In your browser; nothing is uploaded.