In a symmetric independent private value auction model where bidders are con- strained by their privately known budgets, bidders may collude via fund-pooling con- sortiums and a consortium may coordinate actions, in an incentive-feasible manner, to penalize a nonparticipant of the consortium. All the bidders may form a single collusive consortium that buys the good at the minimally admissible price and then randomly selects one of its members as the winner. It is proved that this collusive scheme, de- spite its inefficiency, cannot be preempted by a principal even though the principal can choose any grand mechanism from a class that ranges from standard auctions to those that resemble Che and Kim’s (2006) collusion-proof mechanisms.