With the rapid evolution of virtual reality technologies such as digital twins, industrial Internet of Things, edge intelligence, and metaverse, virtual‒reality integration has become a core driving force for the construction of an intelligent society and the reshaping of industrial systems. Computing power, as the underlying supporting element of virtual‒reality integration, is rapidly evolving from a single centralized computing resource to a complex system characterized by multi-layered collaboration, intelligent scheduling, security, and trustworthiness. This study reviews the current development status and key characteristics of new computing power systems, pointing out that current computing power systems are showing a trend of cloud‒edge‒device integration, intelligent computing power is becoming the core engine for upgrading computing power structures, regional computing power layouts are gradually forming a pattern that emphasizes both differentiation and collaboration, and computing power application models driven by virtual‒reality integration are also showing diversified, ubiquitous, and autonomous characteristics. In the context of virtual‒reality integration, this study further analyzes the key technologies supporting the construction of new computing power systems, including the architecture design of computing power systems driven by virtual‒reality integration and the key technical elements for virtual‒reality integration scenarios, revealing the logical basis for matching computing power supply and demand from both system architecture and computing power orchestration perspectives. Through research on hybrid computing architectures, this study focuses on discussing their practical bottlenecks in areas such as heterogeneous collaboration, low-latency and high-bandwidth assurance, multi-source data security, and privacy protection. To address the aforementioned bottlenecks, we propose key development directions for new computing power systems, including constructing ubiquitous intelligent computing networks, developing trusted computing power systems, breaking heterogeneous collaboration barriers, improving security governance mechanisms, and cultivating a virtual‒reality computing power ecosystem. These will provide theoretical references and strategic support for the construction of future virtual‒reality computing power systems, optimization of industrial ecosystems, and allocation of computing power resources.