Today, Ant Group Co., Ltd. released its annual sustainability report for 2025. The report reveals that over the past year, its AI health assistant, "Ant Afu," has served over 84 million users in third-tier cities and below, with more than 30% of these users being middle-aged and elderly. For its AI payment services, users from these regions constitute 40% of the user base. The AI financial assistant, "Ma Xiaocai," has cumulatively served over 100 million users, issuing 120 million rational investment reminders throughout the year. Beyond AI consultation, the Afu service is now integrated with 300,000 certified human doctors, offering online medical consultation services.
The report indicates that the "Touch to Pay" feature has been deployed across 2,200 specific scenarios, including dining and transportation, integrated with over 20 million merchants and serving 300 million users. Furthermore, Alipay's digital employment solutions have cumulatively assisted 160 million users, successfully helping more than 8 million people secure employment. Among these solutions, the employment AI assistant "Xiao Ye" has served 3 million users, and flexible work arrangements have benefited a cumulative total of 160 million individuals.
Research and Development Investment
The report shows that in 2025, Ant Group and its subsidiaries achieved a record high in research and development investment, reaching RMB 35.03 billion. In foundational large model exploration, the company open-sourced Ring-1T, the world's first trillion-parameter inference model, as well as embodied intelligence models like LingBot-World, the world's first open-source world model. In training and inference infrastructure, Ant developed the AI post-training foundational platform ASystem and open-sourced its core components.
Sustainability and Green Computing
According to the report, Ant Group has achieved carbon neutrality for its own operational emissions (Scope 1 and 2) for five consecutive years since 2021. In 2025, absolute Scope 2 emissions decreased by 55.32% compared to the baseline year, and the proportion of clean energy consumption in data centers increased to 65%. The company's self-developed AI infrastructure, codenamed Theta, increased GPU utilization during the model inference phase by 2.3 times, leading to a significant simultaneous reduction in energy consumption and carbon emissions per effective token. Over the year, green computing technologies facilitated a reduction of 139,545.27 metric tons of CO2 equivalent in upstream data center emissions.
Energy and Industry Standards
Leveraging its "Energy TS" time-series large model, Ant provided energy forecasting services for two provincial power grids, achieving an average day-ahead photovoltaic prediction accuracy rate of 92.6%. Its trusted blockchain technology has been applied in the new energy sector, with over 16 million devices having data recorded on-chain. Concurrently, Ant collaborated with institutions like Tsinghua University to lead the development of the world's first "AI Agent Operational Safety Testing Standard" and engaged nearly 20 universities and institutions in co-building large model ethics and governance frameworks.
In addition, Ant launched 45 new open-source projects throughout the year, bringing its total code repositories to 2,467, with over 13,000 developers participating in the community. The report states that core technologies in AI, blockchain, distributed databases, and security have cumulatively helped more than 33,000 enterprises improve their operational efficiency.
The report also notes that by the end of 2025, after nine years of operation, the Ant Forest initiative has cumulatively pledged donations totaling RMB 4.5 billion, resulting in the planting of 619 million trees across an area exceeding 4.33 million hectares. The program has attracted over 700 million users to participate in low-carbon living and has created more than 4.79 million labor income opportunities in various regions.
Comments