Yes, bsxfun is a very useful function, but it has not been implemented for CVX objects. However, that does not mean you have to use a loop, and indeed you certainly never should.
For instance, consider this:
cvx_begin
variable X(n,n)
variable y(n)
and suppose you want to subtract y from each column of X. Then you can do this:
X - y(:,ones(1,n))