Tang Daosheng, Senior Executive Vice President of Tencent Group and CEO of the Cloud and Smart Industries Business Group
On September 16, the 2025 Tencent Global Digital Ecosystem Conference was held. Tang Daosheng, Senior Executive Vice President of Tencent Group and CEO of the Cloud and Smart Industries Business Group, stated that "seeking industrial efficiency from intelligence and revenue scale from globalization" has become two core drivers for enterprise growth. Tencent will build two efficiency engines - "intelligentization" and "globalization" - to help enterprises achieve steady and sustainable growth.
In terms of intelligentization, Tencent Cloud officially released the Tencent Cloud Intelligent Agent Strategic Map, fully opening up AI capabilities, C-end and B-end advantage scenarios. Through three major upgrades - intelligent agent solutions, SaaS+AI, and large model technology - it will stimulate corporate innovation potential. "Tencent will be people-centered, building 'user-friendly AI', allowing AI to serve people in scenarios, meet people's needs, improve work efficiency, optimize interaction experience, and even provide emotional value," said Tang Daosheng.
According to the introduction, AI has become Tencent's "new business gene". Tencent Yuan has been online for more than a year and has become one of the top three AI-native applications in terms of daily active users (DAU) domestically. Now, the number of questions users ask Tencent Yuan each day has reached the total volume of the entire January. The number of files in IMA knowledge base has exceeded 100 million; the monthly active users of QQ Browser have increased by 17.8 times compared to April. At the same time, AI has also helped Tencent's advertising and gaming businesses achieve double-digit growth.
In terms of globalization, Tencent Cloud will assist enterprises in establishing themselves locally and expanding globally from three directions: infrastructure, technical products, and service capabilities. On one hand, accelerate infrastructure construction, building the foundation for global development with a "global network". On the other hand, through industry-leading product toolboxes, help enterprises use the cloud well, serve users well, and reduce operating costs. At the same time, create agile local service teams, becoming enterprises' "global digital partners" with professional knowledge and personalized service.
Tang Daosheng introduced that this year, Tencent Cloud has continued to increase investment in infrastructure construction, and the speed of overseas area expansion ranks among the top domestic cloud vendors. At the same time, we have completed multiple product and function upgrades, fully compatible with the global technology ecosystem. Over the past three years, Tencent Cloud's international business has continuously grown at high double-digit rates, doubling the number of overseas customers compared to the previous year. More than 90% of Chinese internet companies and over 95% of top game companies have chosen Tencent Cloud for their overseas expansion. For example, GoTo Group in Indonesia, CP Group in Thailand, and Chinese industry leaders such as Meituan Keeta, Midea, and Pop Mart.
The full text of the speech is as follows:
Welcome to the Tencent Global Digital Ecosystem Conference. It's a pleasure to reunite with all of you in Shenzhen Bao'an.
Recently, I've had many exchanges with business leaders, and they generally feel a great competitive pressure. Therefore, "seeking industrial efficiency from intelligence and revenue scale from globalization" has become two core drivers for enterprise growth, which is also key to influencing future industry patterns.
This is exactly the topic we are discussing today - how Tencent Cloud will assist all industries in building "intelligentization" and "globalization" engines, helping enterprises achieve steady and sustainable growth.
Tencent is built on "people-centered", building "user-friendly AI". Let AI serve people in scenarios, meet people's needs, improve work efficiency, optimize interaction experiences, and even provide emotional value.
In the C-end, Tencent Yuan has become one of the top three AI-native applications in terms of DAU within a year of its launch. Now, the number of questions users ask Tencent Yuan every day has reached the total amount of the whole January.
This year, Yuan's analytical capabilities and interactive experience have been continuously enhanced. We have integrated Yuan with more than 10 applications such as WeChat, Tencent Meetings, Tencent Docs, Tencent Video, and QQ Music, further enhancing users' experience and efficiency in various scenarios such as life and work. For example, in the comment sections of official accounts and video numbers, users can directly @Yuan to summarize content, ask for knowledge; in Tencent Meetings, Yuan can attend meetings for you, automatically record meeting information, and generate meeting minutes.
Additionally, the number of files in our IMA knowledge base has exceeded 100 million, meeting people's demand for deep knowledge sharing and acquisition. The monthly active users of QQ Browser have increased by 17.8 times compared to April.
In our internal business processes, AI is widely applied and has achieved significant results. We have deeply integrated AI capabilities into advertising, gaming, and social businesses, becoming "new business genes" for Tencent.
In the advertising business, we optimize advertising creation, delivery, recommendation, and effectiveness through large models, significantly improving ad click-through rates and conversion rates, contributing to a 20% growth in marketing services revenue in the second quarter of this year; in the gaming business, we use AI to accelerate game content production, introducing virtual teammates and NPCs with higher simulation to attract new players and enhance engagement, helping game revenue achieve double-digit growth.
Looking ahead, we will comprehensively open up AI implementation capabilities, connect C-end and B-end technologies and scenarios. Through three major upgrades - intelligent agent solutions, "SaaS+AI", and large model technology - we will help customers and partners build "user-friendly AI", stimulate corporate innovation potential, and help society realize "people-centered" AI development.
First, we will drive continuous upgrades of intelligent agent solutions, accelerating the deep implementation and application innovation of AI in industry scenarios.
Just as apps were the main carriers of the mobile internet era, intelligent agents will become the main application carriers of the AI era. Today, we officially released the Tencent Cloud Intelligent Agent Strategic Map, fully opening up development platforms, application scenarios, infrastructure, model capabilities, and partner ecosystems, helping to build intelligent agents that understand customers, make decisions, execute, and are highly reliable.
In terms of technological innovation, the Tencent Cloud Intelligent Agent Development Platform ADP provides enterprises with a development framework combining large language models + search enhancement, workflows, and multi-intelligent agent collaboration, helping enterprises efficiently build customized intelligent agents. Today, we are launching the ADP3.0 version globally. The new version supports integration with third-party databases, further enhancing RAG capabilities, improving question-answering effects, and enabling precise recognition and intelligent association of images, texts, and tables in complex documents. By weighting the timeliness and authority of search results, it significantly reduces hallucinations, making intelligent agents a trusted "smart hub" for enterprises.
We have added an Agent node in the workflow, enabling autonomous execution of processes and orderly task execution, significantly improving the accuracy of intelligent agent task execution; we have also added an application evaluation tool, testing before application launch to ensure Agent performance reliability; the plugin ecosystem has also expanded significantly, now supporting 140+ MCP plugins.
To help intelligent agent development, we have also built a complete set of Agent infra, providing secure sandbox environments, short-term and long-term memory, tool call management, and full-chain security and observability capabilities.
The unique advantages of Tencent's intelligent agents also lie in their wide range of scenario adaptability. Relying on Tencent Cloud, we cover the entire industrial chain scenarios including research, production, supply chain, sales, and service, accumulating standardized industry intelligent agent templates, accelerating the intelligent upgrade of industries such as retail, healthcare, finance, education, and manufacturing.
Recently, Tencent Qidian Marketing Cloud collaborated with Juewei Food for a full-cycle AI marketing practice. Based on the Tencent Intelligent Agent Development Platform combined with Juewei Food's business processes and data accumulation, we jointly developed multiple intelligent agents including audience targeting, product recommendations, dynamic benefits, and content generation. Throughout the project, multiple intelligent agents "cooperated" in combat, from user insight to content delivery, benefit distribution, and finally to transactions, achieving full-process intelligent operations.
The practical results are also very impressive - the AI marketing efficiency reached 2-3 times that of manual groups. Among millions of test users, the AI group's content click rate was 1.8 times that of the manual group, the payment conversion rate was 2.4 times, and the transaction amount reached 3.1 times. These tangible "efficiency improvements" have received high recognition from clients and provided an innovative model for AI marketing for more enterprises.
The prosperity of intelligent agents not only relies on technology and scenarios but also on open cooperation and industry co-construction. Today, we officially launched the "AI Application Prosperity Plan", initiated the AI Co-Creation Camp, and widely collaborated with developers and ecological partners through technology openness, resource sharing, and joint incubation to promote more vertical intelligent agent solutions. At the same time, through activities like "AI Hundred Schools Tour", we support AI talent development.
Secondly, we will comprehensively upgrade the "SaaS+AI" enterprise application scenarios, seamlessly integrating AI into everyone's work environment, helping individuals improve efficiency and enhance organizational efficiency.
For example, we will fully integrate AI capabilities into core SaaS applications such as code development, office collaboration, and knowledge management. In the development scenario, Tencent Cloud CodeBuddy helps millions of developers with scientific and technological innovation, significantly improving R&D efficiency and quality. It broke the waterfall development where upstream and downstream were constrained, transforming it into a parallel model centered around AI. Previously, developing an e-commerce activity page required the product manager to describe the requirements first, then the designer adjusted the style, and the developer wrote the code, which took 2 days. Now, with AI assistance, it can be completed in as little as 2 hours. Currently, more than 40% of the new code added in many products within Tencent is generated by AI, and 35% of tasks are reviewed by AI. The average monthly delivery count of programmers has increased by 34%, and the delivery cycle has been shortened by 10%.
In the field of office collaboration, applications such as Tencent Meetings, Tencent Docs, and Enterprise WeChat have also been fully AI-enabled. For example, the AI minutes in Tencent Meetings can capture participants' emotions and attitudes through intelligent reasoning, understanding intentions, helping managers better understand team intentions and optimize decision-making. In the past year, the user volume of AI features in Tencent Meetings has increased by over 150%, helping more and more people improve efficiency. I often use Tencent Meetings for interviews, and the new AI minutes feature can also analyze the abilities and attitudes reflected in the interviewee's answers in real-time, facilitating comprehensive assessment.
In the knowledge management scenario, Tencent Leshang has been fully upgraded to Leshang Knowledge Base, supporting storage, management, and Q&A of 102 types of multimodal content. By adding deep processing of knowledge and thinking capabilities, the accuracy of AI Q&A reaches 92%, significantly higher than that of ordinary employees.
In the legal scenario, Tencent Electronic Signature improves contract review efficiency through "AI intelligent review + manual confirmation." Previously, reviewing a complex contract manually required a whole day, but now AI can complete it in as little as 10 minutes, and simple contracts can be completed in as few as 1 minute. It also supports multi-angle reviews, custom screening points, and real-time external access to massive laws and regulations, reducing contract compliance risk by 80%.
The third core upgrade is to continuously iterate on the underlying large models, solidifying the technical foundation of AI innovation, injecting stronger intelligence into diverse applications.
Over the past year, Tencent Hunyuan has released more than 30 models. One important direction is how to achieve stronger model performance with lower deployment and inference costs, thus accelerating the application of models in the industry.
We have launched multiple open-source small models such as 0.5B, 1.8B, 4B, 7B, and A13B, which can run smoothly on mobile phones and computers, truly realizing technological inclusiveness. For example, in Tencent Mobile Security, we used small-sized models to improve the accuracy of spam SMS identification, achieving millisecond-level interception. Recently, the open-sourced Hunyuan Translation Model, a 7B model, won 30 out of 31 categories in the 2025 International Machine Translation Competition.
In the fields of multimodal and 3D world models, we are also continuously exploring. The Hunyuan 3D model has been downloaded over 2.6 million times in the open-source community, ranking first in the industry, and has been commercially used in quality inspection and design fields. Top 3D printing manufacturers such as Toubz Technology and Chuangxiang 3D have fully integrated Tencent Hunyuan 3D. Beginners without any experience can input a picture and quickly generate a 3D model for printing, greatly improving modeling efficiency.
Today, we officially released the Tencent Hunyuan 3D3.0 model, which has three times the modeling accuracy of the previous version, significantly improving the generation effect of characters and complex details. At the same time, we launched a one-stop tool called Hunyuan 3D Studio for 3D professional production pipelines, covering the entire process from concept design to action systems, solving issues such as multiple production stages and high production costs in traditional 3D development pipelines, significantly improving 3D production efficiency and helping industries such as game development, animation production, and industrial design.
Of course, the efficient implementation of AI is a systematic project. In addition to models and platform tools, it also requires solid infrastructure and advanced engineering capabilities. Later, our colleague Qiu Yuepeng will introduce the new progress in AI infra.
If our intelligent engine aims to turn AI from a "trendy concept" into a "real efficiency" that drives business, then our second engine - the "globalization engine" - hopes to help enterprises win new growth in the "age of great voyages."
Currently, the globalization process is facing multiple challenges: How to choose the most suitable market entry? How to flexibly respond to different country's compliance requirements? How to quickly seize the new needs of overseas users? Answering these questions is key to enterprises expanding global markets.
Over the years, we have exported our mature products in the domestic market to overseas. For example, Tencent Cloud's audio-video, edge security acceleration platform, mini program platform, AI digital human, etc., have become "standard" for many enterprises to layout global markets.
At the same time, we have added more available areas globally and completed multiple product and function upgrades, fully adapting to the global technology ecosystem. Over the past three years, Tencent Cloud's international business has continuously grown at high double-digit rates, doubling the number of global customers compared to the previous year. For example, we have deeply served GoTo Group, the largest internet company in Indonesia, CP Group in Thailand, e&UAE in the UAE, Orange in France, Tabelog, the largest restaurant reservation platform in Japan, and a series of Chinese enterprises such as Meituan Keeta, Midea, Pop Mart, Yang Guofu Spicy Hot Pot. At the same time, 90% of internet companies and 95% of top game companies have chosen Tencent Cloud for their overseas expansion.
We will also upgrade Tencent Cloud's international strategy from three directions: infrastructure, technical products, and service capabilities, building a "globalization engine" to help enterprises establish themselves locally and expand globally with consistent product experiences.
Firstly, accelerating infrastructure upgrades, building a "global network" to lay the foundation for global development.
This year, we have continued to invest heavily in infrastructure construction, and the speed of overseas area expansion ranks among the top domestic cloud vendors. Previously, we also announced a $150 million investment in Saudi Arabia to build the Middle East's first available area; we have also built a third available area and established a regional office in Osaka, Japan. As of the first half of this year, Tencent Cloud has established 11 regional offices worldwide and deployed 9 global technical support centers.
完善的基础设施，是可靠服务的保障。我们希望，企业在全球任何一个地方，都能够就近接入腾讯云的资源与服务，缩短用户与服务的 “物理距离”，高效开展本地业务;也能更好地满足数据合规、灾备，以及降本增效的要求。
For example, since 2017, we have cooperated with Guangqi in smart driving, smart cabin, and enterprise digitalization. Facing global development, we have also helped Guangqi rapidly replicate its advantages and services to overseas. In recent years, Midea's European business has developed rapidly, and the traditional IT architecture has faced great challenges. They built a new technical foundation based on Tencent Cloud's data center in Frankfurt, unifying nearly 50 independent business systems into a standardized cloud-native architecture, significantly improving system stability, scalability, and R&D collaboration efficiency, and completing cost optimization goals ahead of schedule.
