According to foreign media reports, ZhiPu AI (Z.ai) has completed the construction of a 1GW-level domestic AI computing data center, with the overall system using domestically produced AI chips. Meanwhile, the company recently completed the acquisition of XCore Sigma, a domestic AI heterogeneous computing software enterprise, further enhancing its AI infrastructure layout from computing resources to foundational software.
It is understood that XCore Sigma originated from the Compiler Laboratory of the Institute of Computing Technology at the Chinese Academy of Sciences. The company has long focused on heterogeneous computing software stacks, compiler optimization, runtime systems, and AI inference infrastructure, and is regarded as one of the leading AI Infra technology teams in China. This acquisition will enhance ZhiPu's capabilities in adapting to domestic chips, computing resource scheduling, and model deployment optimization.
Industry experts believe that ZhiPu's two recent moves correspond to the "computing power supply" and "computing power release" stages of the AI industry chain. Among them, the 1GW-level domestic computing center can provide stable computing resources for large-scale model training; XCore Sigma's software technology helps improve the computing efficiency of different types of AI chips, enhancing hardware utilization and reducing model inference costs through compiler, Runtime, and inference engine optimizations.
As the competition in large models enters a stage where infrastructure and engineering capabilities are being compared, AI companies are shifting from merely pursuing model size to building a complete technical system covering chips, computing power, software stacks, and application deployment. Recently, the market has been closely watching ZhiPu's next-generation foundational models. Some views suggest that with the support of large-scale domestic computing resources, an established AI Infra system, and long-term accumulated post-training capabilities, ZhiPu's future models are expected to continue developing toward larger parameter scales and higher levels of intelligence, while improving inference efficiency and practical deployment capabilities.
This computing center construction and integration of basic software capabilities mark that domestic large model enterprises are accelerating the development of an autonomous and controllable AI infrastructure ecosystem, reflecting that AI industry competition is expanding from model capabilities to full-stack technological system competition.
