Subject: Gemini API incorrectly rejects supported India server location

Gemini API
1

We run a production Node.js backend on Akamai Connected Cloud/Linode. Gemini Developer API calls are failing with:

HTTP 400 FAILED_PRECONDITION
User location is not supported for the API use.

Environment:

  • API: Gemini Developer API (generativelanguage.googleapis.com/v1beta)
  • Model: gemini-2.5-flash-lite
  • Server IPv4: 172.235.29.182
  • Provider/ASN: AS63949 Akamai Connected Cloud
  • Reported location: Chennai, Tamil Nadu, India
  • GCP project: invoice-management-app-462215
  • India is listed as a supported Gemini API region.

We reproduced the problem using a minimal text-only Node.js request directly from the server, outside PM2 and without any invoice/document upload:

400
{
  "error": {
    "code": 400,
    "message": "User location is not supported for the API use.",
    "status": "FAILED_PRECONDITION"
  }
}

curl -4 https://ipinfo.io/json reports the server in Chennai, India. The IPv6 test did not return a usable IPv6 address. Therefore, this is not caused by our application code, document format, payload size, PM2 configuration, or invoice-processing workflow.

Business impact:

  • Invoice document analysis is unavailable.
  • Vendor registration document extraction is affected.
  • Other Gemini-backed administrative features are also unavailable.

Could you please:

  1. Confirm how Google currently geolocates 172.235.29.182.
  2. Check whether this IP or AS63949 range is incorrectly classified or blocked.
  3. Confirm whether the API key/project has any regional eligibility restriction.
  4. Correct or allow the IP classification if this is a geolocation issue.
  5. Confirm whether moving production workloads to Vertex AI is now the recommended permanent solution for non-GCP cloud servers.

We are evaluating migration to Vertex AI, but would like confirmation of the root cause because the server is located in an officially supported country.

2

Hello @Nandakumar_Srinivasa ,

Could you please check if you can reproduce this issue directly in AI Studio? Additionally, try testing with other newer models to see if the error persists.