Source: Vertex AI — Generative AI Locations and Endpoints

Global, multi-region (us/eu) and regional endpoints for Gemini on Vertex AI, and what each implies for data residency and features.

Contents

Summary

Explains the three endpoint types for generative AI on Vertex AI (now Gemini Enterprise Agent Platform) and the trade-offs between availability and in-jurisdiction processing.

Key claims

  • Regional endpoints (e.g. us-central1) route to one region; multi-region endpoints use location us or eu and hostnames like aiplatform.us.rep.googleapis.com; the global endpoint routes worldwide.
  • The page says endpoints alone don’t guarantee data residency or in-region ML processing; multi-region endpoints are the option for keeping ML processing inside the US or EU.
  • Don’t use global if you have ML-processing location requirements; for partner models it does not support data residency.
  • global gives higher availability and fewer 429s.
  • Not available on global: tuning, batch prediction for Anthropic and open MaaS models, RAG corpora (RAG requests work), and Provisioned Throughput except for listed models.
  • Private Google Access doesn’t work with multi-region endpoints; use Private Service Connect.
  • Python: vertexai.init(location=...) (google-cloud-aiplatform ≥ 1.79.0) or the genai client with enterprise=True.

Feeds

Google Vertex AI

Suggest changes →