Microsoft has created data gateways as a way that cloud products such as Power BI and Azure Machine Learning can access on-prem data, often called a hybrid model.  This is most useful for companies that do not want to put their data in the cloud, or there is simply too much data to upload to the cloud.

The confusion is that there are a number of different gateways, and some of those have been renamed multiple times.  I’ll try to provide a quick summary to help you understand all the current options:

On-premises data gateway: Formally called the enterprise version.  Multiple users can share and reuse a gateway in this mode.  This gateway can be used by Power BI, PowerApps, Microsoft Flow or Azure Logic Apps.  For Power BI, this includes support for both scheduled refresh and DirectQuery.  To add a data source such as SQL Server that can be used by the gateway, check out Manage your data source – SQL Server.  To connect the gateway to your Power BI, you will sign in to Power BI after you install it (see On-premises data gateway in-depth).

Personal gateway: This is for Power BI only and can be used as an individual without any administrator configuration.  This can only be used for on-demand refresh and scheduled refresh (not DirectQuery).

To install either of these gateways, see On-premises data gateway.  The differences between the two:

Data Management Gateway: This is the old solution for Power BI but still used for Azure Data Factory and Azure Machine Learning.  You can download it here.

Here is the differences between the Data Management Gateway and the On-premises data gateway:

Microsoft has another different gateway that is specific to the Azure Analysis Services (Preview) service, at least for now: On-premises data gateway.  UPDATE 8/19/17: Azure Analysis Services now supports the shared On-Premises Data Gateway – see On-premises data gateway support for Azure Analysis Services.