It's probably better to use XPath to query such documents. XPath can often work in a streaming mode which doesn't require the whole document to be loaded into a DOM tree.
See Select-Xml
:
The following would count all elements in an XML file:
Select-Xml -Path C:\fso\products.xml -Xpath "count(//*)"
This way you're able to fetch small snippets of the XML you're after or do computations on them.
See: http://technet.microsoft.com/en-us/library/hh849968.aspx