On every good team I’ve workd on, we’ve (the leads and architects) stayed away from utilizing GUIDs for authentication and userIDs just because they are especially for an e-commerce site very under-performing in terms of performance vs. just a PK Int. At least that was the consensus in addition to a pain to work with.
How can I use Int Ids but for large data sets and still get past the uniqueness issue (running out of unique Ids) without using GUIDs?
Take a look here for your options. Also, you should check if storing all those records inside single table is a good option for you. Perhaps you should think about table partitioning?