big integer design question

Hi,
I am developing the DB code for what I hope will be
become a rather large/popular web application. The design
uses a unified object model, all object_ids coming from one
master sequence (suffice it to say it must be this way to meet
other design goals).
Being the worrier that I am, I am concerned that we may some
day overflow the integer currently holding the object_id.

So my question is, keeping performance and feasability in mind, would
it be better to just make all the table and cross-type index query
changes to move to a big integer from the get-go,
or would it be better to just wait to see if/when we reach say a billion on
our
ID sequence and worry about it then.

Any and all opinions, insights, and experiences on this topic would
be much appreciated.

Sincerely,
Ara Anjargolian