Features with limited regional availability
This article lists features with availability differences by region.
Some features are not included in the tables below. These features, known as Designated Services, use Databricks Geos to manage data residency when processing customer content. A Geo is a group of data center regions that Databricks uses to provide predictability and transparency regarding where your data is processed. See Databricks Designated Services and Databricks Geos: Data residency.
The following tables list support for features that are only available on a subset of Databricks on Google Cloud regions. The tables indicate whether or not a region supports each of these features. If a feature is supported in all regions, it is not included.
All Databricks on Google Cloud features are supported in these regions. However, some regions do not support GPU instance types. For a list of regions and zones that support GPU instance types, see the Google Cloud Platform documentation.
Serverless availability
The following table shows supported regions for these serverless compute features that have limited availability:
- Serverless compute for notebooks and jobs
- Serverless Lakeflow Declarative Pipelines
- Serverless SQL warehouses
- Databricks Apps
- Managed connectors in Lakeflow Connect
Region | Location | Serverless compute for notebooks and workflows | Serverless SQL warehouses | Serverless Lakeflow Declarative Pipelines | Databricks Apps | Managed connectors in Lakeflow Connect |
|---|---|---|---|---|---|---|
| Tokyo, Japan | ✓ | ✓ | ✓ | ||
| Mumbai, India | ✓ | ✓ | ✓ | ✓ | ✓ |
| Singapore | ✓ | ✓ | ✓ | ✓ | ✓ |
| Sydney, Australia | ✓ | ✓ | ✓ | ||
| Belgium, Europe | ✓ | ✓ | ✓ | ||
| England, Europe | ✓ | ✓ | ✓ | ✓ | ✓ |
| Frankfurt, Germany | ✓ | ✓ | ✓ | ✓ | |
| Montréal, Canada | ✓ | ✓ | ✓ | ||
| São Paulo, Brazil | |||||
| Iowa, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| South Carolina, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| Virginia, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| Oregon, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| Nevada, US | ✓ | ✓ | ✓ |
AI and machine learning features availability
The following table shows supported regions for these AI and machine learning features that have limited availability:
- Predictive optimization
- Vector Search
- Mosaic AI Agent Framework and Agent Evaluation
- AI Functions using Foundation Model APIs
- Databricks Lakehouse Monitoring
Region | Location | Vector Search | AI Functions optimized for batch inference | Lakehouse Monitoring | Predictive Optimization | Mosaic AI Agent Framework & Evaluation |
|---|---|---|---|---|---|---|
| Tokyo, Japan | |||||
| Mumbai, India | ✓ | ✓† | ✓ | ✓ | ✓ |
| Singapore | ✓ | ✓† | ✓ | ✓ | ✓ |
| Sydney, Australia | |||||
| Belgium, Europe | |||||
| England, Europe | ✓ | ✓ | ✓ | ✓ | |
| Frankfurt, Germany | ✓ | ✓ | ✓ | ||
| Montréal, Canada | |||||
| São Paulo, Brazil | |||||
| Iowa, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| South Carolina, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| Virginia, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| Oregon, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| Nevada, US |
Model serving features availability
The following table shows supported regions for these Mosaic AI Model Serving capabilities and features that have limited availability:
- Core Model Serving
- Foundation Model API pay-per-token
- Foundation Model API provisioned throughput
- External models
Region | Location | Core Model Serving capability (CPU serving) | Core Model Serving capability (GPU serving) | External models | Foundation Model APIs (provisioned throughput)** | Foundation Model APIs (pay-per-token) |
|---|---|---|---|---|---|---|
| Tokyo, Japan | |||||
| Mumbai, India | ✓ | ✓ | ✓ | ✓† | ✓† |
| Singapore | ✓ | ✓ | ✓ | ✓† | ✓† |
| Sydney, Australia | |||||
| Belgium, Europe | |||||
| England, Europe | ✓ | ✓ | ✓ | ✓† | ✓† |
| Frankfurt, Germany | ✓ | ✓ | ✓ | ✓† | ✓† |
| Montréal, Canada | |||||
| São Paulo, Brazil | |||||
| Iowa, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| South Carolina, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| Virginia, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| Oregon, US | ✓ | ✓ | ✓ | ✓ | ✓ |
| Nevada, US |
** Includes GPU support
† Region supported only when cross geography routing is enabled. Foundation Model API workloads that rely on GPU compute availability, such as workloads that use models from the Meta Llama family, require cross geography routing.