Then you would likely need to force the page to pause so that the speech is rendered as a wav file or whatever and then you'll have to transmit this sound file to the client side so that the user can hear it. Plus the user won't have any control about what is read (ie with those reading software, it will read for example a button label when it is hovered etc...). Oh and the speech SDK  needs of course to be installed on the web server side.

In short it will require much more work for a worst result. Is this really what you want?

Before proceeding further I would really suggest to give a try at a screen reader. For example Windows Narrator comes with Windows. Give it a try before deciding what to do.

My 2nd choice would be an API such as https://www.microsoft.com/cognitive-services/en-us/speech-api/documentation/api-reference-rest/bingvoiceoutput

And doing it on the server side would be likely my very last choice. Once again keep in mind that this code runs on the web server (and so it would speak or listen from the server devices not from the devices where the browser is actually used) and so you'll have to handle transmitted back and forth sound files to and from the client side.