Platform giants in China have operated with persistently compressed margins in highly concentrated markets for much of the past decade, despite market shares exceeding 60\% in core segments. Standard theory predicts otherwise: either the weaker firm exits, or survivors raise prices to monopoly levels. We argue the puzzle dissolves once firms are viewed as ecosystem optimizers rather than single-market profit maximizers. We develop a dynamic game in which a firm's willingness to subsidize depends on the spillover value its users generate in adjacent markets -- what we call \textit{ecosystem complementarity}. When this complementarity is strong enough, perpetual below-cost pricing emerges as the unique stable equilibrium. The result is not predation in the classical sense; there is no recoupment phase. It is a permanent state of subsidized competition, rational for each firm individually but potentially inefficient in aggregate. We characterize the equilibrium, establish its dynamic stability, and show that welfare losses compound over time as capital flows into subsidy wars rather than innovation. The model's predictions are consistent with observed patterns in Chinese platform markets and suggest that effective antitrust intervention should target cross-market capital flows rather than prices.

翻译：过去十年间，中国平台巨头在高度集中的市场中持续以微薄利润率运营，尽管其核心业务市场份额超过60%。标准理论预测与此相反：要么弱势企业退出市场，要么存活企业将价格提升至垄断水平。我们认为，一旦将企业视为生态系统优化者而非单一市场利润最大化者，这一悖论便迎刃而解。我们构建了一个动态博弈模型，其中企业的补贴意愿取决于其用户在相邻市场产生的溢出价值——我们称之为\textit{生态系统互补性}。当这种互补性足够强时，永久性低于成本定价将成为唯一稳定的均衡状态。该结果并非经典意义上的掠夺性定价；不存在利润回收阶段。这是一种永久性的补贴竞争状态，对每个企业个体而言是理性的，但整体上可能造成效率损失。我们刻画了均衡特征，证明了其动态稳定性，并指出随着资本持续流入补贴战争而非创新领域，福利损失将随时间不断加剧。该模型的预测与中国平台市场的观测模式一致，表明有效的反垄断干预应针对跨市场资本流动而非价格本身。