The beauty of React is that it's a small abstraction layer that sits on top of a rendering target. So far, the target has been the browser, but it can also be the server. The render target can be anything, just as long as the correct translation calls are implemented behind the scenes.
In the case of rendering on the server, components are rendered to strings. The server can't actually display rendered HTML; all it can do is send the rendered markup to the browser. The idea is shown in the following diagram:
It's possible to render a React component on the server and send the rendered output to the browser. The question is, why would you want to do this on the server instead of in the browser?
