Hard to move
Vertex AI on UpCloud: what it becomes and what it costs.
On UpCloud, Vertex AI becomes GPU dedicated servers + MLflow / Kubeflow. Training and hosting pipelines are GCP-specific; Gemini pass-through is priced at 100% (same models from Google AI Studio), training and hosting at about 50%.
Effort grade
Hard
No equivalent: redesign it, buy a SaaS, or leave it where it is
Cost after the move, before engineer time
50% of today
A mix of model pass-through (100%) and training / endpoint hours (GPU dedicated servers at list, about 50%); 50% overall.
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 Vertex AI line, from your export
Drop one month of your Google Cloud cost export and the report prices every service, Vertex AI included, on UpCloud list prices of 18 Oct 2026, net of the engineer time to run it. In your browser; nothing is uploaded.