IDC Evaluates China's AI Computing Power Management Platforms: Paradigm Ranks First in Overall Score, Achieves Full Marks in Four Key Dimensions

Global technology market research firm IDC recently released the "China AI Computing Power Management Platform Technical Capability Assessment, 2026." The report shows that Paradigm received full marks in four dimensionscomputing resource management, computing operations and maintenance, technical architecture, and delivery servicesranking first in overall score among participating vendors. This assessment was conducted across six major dimensions: computing resource management, computing service management, computing operations and maintenance, technical architecture, business scenario support, and delivery services. IDC noted that as AI applications enter the stage of large-scale deployment and agent technology continues to develop, AI infrastructure is evolving from a basic computing power platform into core IT infrastructure supporting enterprise intelligent transformation. The report shows that Paradigm performed outstandingly in unified management of heterogeneous computing power, fine-grained GPU virtualization, large-scale cluster operations and maintenance, and hybrid cloud delivery. Its platform is not bound to any chip vendor and has currently adapted to compute cards from more than ten vendors, including Ascend, Hygon DCU, Cambricon, ILUVATAR COREX, MetaX, Moore Threads, Kunlunxin, Enflame, T-Head, and Biren, supporting mixed deployment of training and inference tasks. In terms of product capabilities, the Paradigm platform offers zero-intrusion multi-chip virtualization, exclusive and shared resource pools, intelligent scheduling, AI gateway, dual-track metering of card-hours and Tokens, and agent-oriented computing power servitization capabilities, covering full-chain operations and maintenance from compute nodes, accelerator cards, and tasks to model services. IDC also mentioned in the report that a policy bank achieved unified management of computing power across two locations and three data centers through Paradigm's three-track resource pool architecture, improving computing resource utilization efficiency and supporting smooth business migration. Paradigm founder and CEO Dai Wenyuan stated that as the AI industry moves from "model competition" into the stage of large-scale deployment, the key issue facing enterprises has shifted from using large models to how to continuously transform agent capabilities into business productivity under controllable costs and stable operations. Paradigm will continue to deepen its work in heterogeneous computing power scheduling and agent infrastructure to help enterprises improve the efficiency, economy, and reliability of AI applications.