Our Technology Technology is key to our success in achieving efficiency, improving user experience, and enabling innovation. Our world-class proprietary technology supports peak order volumes of up to hundreds of thousands per second, delivers tens of billions of online marketing impressions per day, and enables millions of merchants, brands and other businesses to conduct their operations efficiently and effectively. The uniqueness of our technology lies in the unparalleled large-scale application environment due to the scale of our businesses as well as our diverse range of product and service offerings. By continually applying our technology across our businesses, we generate knowledge and innovations that drive improvements and further technological development. Members of our research and development team play key roles in various international standardization organizations in areas such as security, and actively participate in international open source foundations focusing on areas such as software engineering, cloud-native applications and databases. This fiscal year, Alibaba established “User first, AI driven” as our strategy. Therefore, we have increased our research and development investment in AI model and infrastructure technology. Key components of our technology include those described below: Technology Infrastructure Our data centers employ leading technologies in distributed fault-tolerant architecture, advanced in-house power and cooling equipment, AI-driven intelligent monitoring and operational technology, to support AI computation businesses with high resilience and high-density capabilities. Combined with a high proportion of clean energy use, our data centers provide a more stable, secure, efficient, agile, and greener infrastructure service. Cloud Operating System Apsara (our proprietary general-purpose distributed computing operating system), ShenlongCompute (our hardware virtualization architecture), Pangu (our distributed cloud storage system) and Luoshen (our cloud network structure), together, provide Cloud Intelligence Group’s customers and our core businesses with enhanced computing power and storage capabilities to support their and our business growth in the new technology era. Database We have comprehensively upgraded the serverless capabilities of our cloud-native transactional database PolarDB, which helps our customers to meet their real-time storage and transaction processing requirements with elasticity and scalability. We also developed AnalyticDB, a cloud-native distributed analytical database that can be integrated with LLMs, to support real-time interaction and complex analysis of massive data. We have also developed a multimodel database Lindorm, which leverages its efficient storage and integrated analysis capabilities to achieve intelligent analysis and processing of unstructured data. Big Data Analytics Platform We have developed a large-scale distributed data analytics platform that can efficiently handle the complex computing tasks of hundreds of petabytes of data per day, such as batch computing, realtime computing, interactive analysis and data governance. We also launched a new generation of streaming lakehouse solutions, Streaming Lakehouse (Flink+Hologres), with a real-time dedicated data lake format Paimon, which innovatively combines lake format with real-time streaming updates into the lake architecture, among others. To meet the needs for multi-modal large model data processing, we launched MaxFrame, a data science computing framework compatible with the Python development ecosystem. Artificial Intelligence Our proprietary distributed deep learning platform, PAI, and our Lingjun intelligent computing service employ leading hardware and software optimization technologies that feature large-scale cluster scalability and boast the speed and stability of the training of LLMs. PAI also supports fine-tuning and deployment of mainstream open-source large models with serverless service modes available. Our proprietary LLM, Tongyi Qianwen, is committed to achieving artificial general intelligence (AGI). In fiscal year 2024, eight industry-oriented models based on Tongyi Qianwen were launched. In May 2024, Tongyi Qianwen was upgraded to version 2.5 and we opensourced Qwen-110B, the 110B parameter version, allowing AI community and customers to customize and innovate based on their own needs. Business Overview 40 Alibaba Group Holding Limited
RkJQdWJsaXNoZXIy MjA2Njc2MQ==