How do I retrieve address balances to build a Merkle sum tree?

Curious about building my own proof of solvency program. I understand how the Merkle sum tree works in theory but I am having issues understanding the practical ways of getting the information I need.

Let's say that my client, for example Newton, has 400,000 addresses for each of their 400,000 clients.

1. How do I get the balance for those 400,000 addresses?

2. How do I create the tree in such way where a customer is able to go check the balance of their address?