Prediction markets have widely emerged, especially in corporate settings. In order to overcome the perpetual problem of illiquidity, many prediction markets apply so-called automated market makers as market mechanism for trading. However, because they usually need a subsidy to work, they only have been used in play-money markets, where no real money is at stake and losses for operators cannot occur. In this paper, we analyze the only two automated market makers with upper-bounded losses and study maximum and expected subsidies. For PM operators, these amounts to subsidize markets can potentially be compared against the costs of running pure playmoney markets.