算力政策利好持续释放 “毫秒用算”建设提速

Core Insights - The construction of computing power infrastructure in China is accelerating, supported by increasing policy initiatives aimed at establishing a millisecond-level computing network by 2027 [1][2] - The "millisecond computing" initiative is seen as a paradigm shift in computing service models, enhancing efficiency across various sectors including public services, healthcare, and daily internet usage [2] - The initiative also aims to facilitate the commercialization of AI applications by reducing latency, which is critical for the performance of large models [2] Group 1: Policy and Infrastructure Development - The Shanghai Municipal Communication Administration and the Shanghai Municipal Economic and Information Commission have issued a notice to develop a millisecond-level computing network, aiming for comprehensive coverage and efficient connectivity [1] - The initiative reflects a transition from macro planning to micro implementation in computing power construction, as noted by experts [1] Group 2: Demand and Application - The demand for computing power is increasing, necessitating a unified national computing power system to address issues like resource misallocation and scheduling inefficiencies [2] - Millisecond-level computing is expected to significantly enhance public service capabilities, such as real-time traffic management and remote medical procedures [2] Group 3: Challenges in Implementation - The construction of the millisecond computing network faces challenges including scheduling difficulties, fragmented computing power trading markets, and environmental concerns related to energy consumption [3] - The energy consumption for training advanced AI models, such as GPT-3, is substantial, highlighting the need for efficient energy management in computing power infrastructure [4] Group 4: Corporate Initiatives - Major telecommunications companies are actively developing the millisecond computing network, with achievements such as a three-tier latency network and high-speed connections to national hubs [5] - Companies are encouraged to transition from traditional telecommunications roles to computing power service providers, promoting differentiated latency guarantees and enhancing computing power utilization [5]