证券代码：600289 证券简称：ST信通 公告编号：2025-141 亿阳信通股份有限公司 登录新浪财经APP 搜索【信披】查看更多考评等级 关于上海证券交易所对公司2025年三季度报告 信息披露监管问询函的回复公告 本公司董事会及全体董事保证本公告内容不存在任何虚假记载、误导性陈述或者重大遗漏，并对其内容 的真实性、准确性和完整性承担法律责任。 亿阳信通股份有限公司（以下简称"公司"）于近日收到上海证券交易所（以下简称"上交所"）《关于亿 阳信通股份有限公司2025年三季度报告的信息披露监管问询函》（上证公函【2025】3908号，以下简称 《问询函》）。公司对《问询函》内容高度重视，积极组织相关各方对《问询函》中涉及的内容进行逐 项落实。现就《问询函》中相关问题回复如下： 一、关于经营业绩。2025年三季报显示，前三季度公司实现营业收入1.93亿元，同比增长0.03%；其中 第三季度确认收入1.27亿元，同比增长88.22%，主要系公司开展硬件定制业务所致。前三季度公司实现 归母净利润-1.37亿元，其中第三季度实现归母净利润-0.42亿元，均同比由盈转亏。请公司： （1）分季度列示各业务板块的收入、毛 ...

Core Insights - The 25th China International Industrial Expo showcased several manufacturers presenting industrial edge AI computing products, highlighting a shift towards edge AI computing that integrates AI algorithms for real-time data processing and analysis near data sources [1] Group 1: Transition from Cloud to Edge Computing - Edge computing is defined as processing data near its source rather than relying on centralized cloud computing, addressing issues of latency, privacy, and bandwidth [2] - The first step in digital transformation for industrial enterprises is equipment networking, which leads to the need for edge computing to gather and process data locally before sending it to the cloud [2] Group 2: Benefits of Edge AI Computing - Edge AI computing devices are typically compact and energy-efficient, reducing data transmission distances and frequencies, which lowers energy consumption [3] - An example provided indicates that a small edge gateway can cover an entire charging station, performing tasks such as sensor data collection and image recognition [3] Group 3: Industrial Large Model Deployment - The demand for inference in industrial applications is increasing, leading to the development of edge AI that allows for real-time inference at the edge [4] - Large models require significant computational resources, but they can be more cost-effective in development compared to small models, which need extensive data for training [5] Group 4: Hardware and Chip Development - The integration of powerful AI chips like GPUs, ASICs, and NPUs into edge devices is essential for supporting large models, with companies like Dawning Network and Advantech leading in this area [5][6] - The demand for low-latency, high-efficiency custom chips is rising, with new products being developed to support industrial-grade applications [6]