文/陈永伟 2025年的帷幕已经落下，这一年中，AI行业无疑走过了极具里程碑意义的一程。从技术范式的革新， 到商业逻辑的重构，从产业应用的落地，到全球规则的博弈，这一年既有突破，也留下诸多思考。 鉴于AI发展错综复杂，这里只能从十个侧面做一个简要回顾。 一、多模融合 过去几年中，AI大模型在文字、推理等方面进展神速，但它们多模态能力的发展却相对迟缓，这在很 大程度上限制了其能力的发挥。比如，在4.0版本之前，GPT虽然已经能写诗、会编程，但既看不见、 也画不出，如果用户想让它分析一张图片讲了什么，或是根据要求生成一张图片，它就显得力不从心。 虽然从2024年开始，AI开发者们就开始大力发展模型的多模态能力，但在相当长一段时间里，这些努 力仍然主要集中在对既有模型进行组合——文本一个系统，图像一个系统，语音再来一个系统，然后用 工程手段把它们拼在一起。这样的模型可以完成一些多模态任务，但由于各系统之间存在协调问题，其 能力局限性一直十分明显。到了2025年，越来越多的开发者不再满足于这种"拼装式"方案，转而开始设 计"原生多模态"模型，从训练之初起，就让模型在同一个体系里同时处理文本、图像、音频、视频等信 息。 ...

Core Viewpoint - The time series database (TSDB) industry is experiencing rapid growth driven by the exponential increase in time series data generated by IoT devices and cloud platforms, with the global market expected to grow from $388 million in 2024 to $776 million by 2031 [1][10]. Group 1: Industry Overview - Time series databases are specialized databases designed for storing and managing time series data, optimizing the ingestion, processing, and storage of timestamped data [2][3]. - The emergence of smart hardware, smart manufacturing, smart cities, and smart healthcare has led to a significant increase in time series data generation [1][9]. - Traditional relational databases and NoSQL databases face challenges in handling the high volume and concurrency of time series data, leading to the development of time series databases [1][10]. Group 2: Market Size and Trends - The global time series database software market is projected to reach $776 million by 2031, growing from $388 million in 2024 [10]. - As of June 2025, there are 41 time series databases globally, a decrease of 14 from the previous year, indicating increased industry concentration [14]. - In China, the number of time series databases is 17, down by 10 from the previous year, reflecting a competitive market landscape [16]. Group 3: Competitive Landscape - The industry features a mix of open-source and commercial models, with foreign markets leaning towards open-source solutions while domestic markets favor commercial offerings [18]. - Major domestic time series databases include Tdengine, KaiwuDB, DolphinDB, and openGemini, which play significant roles in driving industry development [20][21]. Group 4: Development Trends - Future trends indicate a deep integration of time series databases with artificial intelligence, enhancing capabilities for fault prediction and trend analysis [23][29]. - The adoption of cloud-native technologies is expected to grow, allowing for flexible resource management and cost reduction [25][29]. - The deployment of time series databases at the edge will facilitate real-time data processing and decision-making in IoT applications [26][29]. - There is a movement towards multi-model integration, enabling the management of diverse data types within time series databases [27][29].