Decentralized water systems (DWS), consisting of localized facilities designed to treat and reuse water at or near the source, offer a promising approach to mitigate global water scarcity. By recycling used freshwater from households, DWS contribute significantly to water conservation. Yet, their widespread adoption and implementation remain limited, especially in high-rise high-density Asian cities such as Hong Kong. This research aims to investigate the preferences for DWS in the residential sector of Hong Kong. Three types of DWS, namely rainwater harvesting (RWH), greywater recycling (GWR), and wastewater reuse (WWR) systems were examined. An online questionnaire survey incorporating discrete choice experiments was conducted in January 2024 in Hong Kong, and validated samples were collected from 1,010 respondents. A latent class choice model (LCCM) was employed to categorize respondents into distinct preference classes. These latent classes were identified as value maximizing, cost-conscious, neutral and nonchalant, practical budget-conscious, and those prioritizing water safety and long-term lifespan in DWS adopters. Leveraging the insights from the LCCM results, we conducted a subsidy analysis to explore the potential adoption of DWS without and with government subsidy support. The findings indicate that in the absence of government subsidies, both potential apartment relocators and new apartment buyers generally show reluctance towards adopting DWS, with a slight preference for RWH if any system is chosen. However, when supported by government subsidies, GWR emerges as the system of choice, followed by WWR and RWH. Remarkably, with a subsidy ranging from around 0.4 to 0.6 HKD/m³, it can already stimulate a drastic increase in the adoption of both GWR and WWR. These results suggest significant strategic implications for building developers, water system professionals, and government bodies tasked with developing sustainable urban freshwater management strategies, not only in Hong Kong but also in other cities with similar built environments.