Take a K3, or a general $n$-dimensional complex torus $M$, $n>1$, without any integer (1,1)-classes, and remove a point. You will obtain a non-compact Kahler manifold without any non-trivial line bundles (because its second cohomology stays the same). For such a manifold it's not hard to show that no positive bundles exist. Indeed, if there is a positive line bundle, it must be trivial. 
Then you would have a positive form $\eta$ which is exact. By 
<a href="http://en.wikipedia.org/wiki/Skoda%E2%80%93El_Mir_theorem">Skoda-El Mir theorem,</a> the trivial extension of $\eta$ to $M$ is closed, positive and
hence exact current. This is impossible, because $M$ is compact and Kahler.