'Will definately use more resources', as everyone else has said, is just wrong. CPU resources? Disk resources? Bandwidth, memory, or network resources? At design time, at runtime, compared to nothing or compared to what.

I have experimented with computed columns; in write-once, read-many reporting tables on SQL Server 2000 I found a permanent column with a CHECK constraint to enforce the computation's integrity was a superior solution from a performance standpoint, given adequate disk IO.

In one case, the business users concatenated valuable data into a character type field, so SUBSTRING became very important; changing the core columns was a nonstarter, but I could add one.