Liu Chuanlin, Head of the Wuwenxinqiong Solution: Thoughts on the Implementation of China's Computing Power Industry in the New AI Era | 36Kr 2025 AI Partner Conference for All Industries
On August 27th, the 2025 AI Partner Conference for All Industries, jointly hosted by 36Kr and the China Europe International Business School, grandly kicked off at the Zhongguancun Software Park in Beijing. The conference, themed "Chinese Solutions", is divided into two major chapters: "Chinese Solutions" and "Who Will Define the Next AI Era". It focuses on four major topics: "The Golden Age of Chinese Innovation", "Can Super Intelligent Agents Become the Core Form of Next-Generation AI?", "Chinese Solutions Reshape the Global Technological Competition Landscape", and "The Prosperous Scenario of AI Integration and Innovation across All Industries", comprehensively presenting the latest breakthroughs and ecological systems of Chinese AI, sharing the growth path and future prospects of Chinese-style AI, and exploring the innovative models of Chinese solutions.
On the same day, Liu Chuanlin, the person in charge of the solutions of Wuwen Xinqiong, delivered a keynote speech titled "Thoughts on the Implementation of China's Computing Power Industry in the New AI Era".
The following is the content of the speech, edited by 36Kr:
Dear friends:
Good afternoon! I'm very honored to communicate with you here today. This sharing will revolve around our thoughts and implementation practices in the domestic AI industry. The core content is divided into three parts: first, an introduction to the company's development overview; second, our in-depth insights into the AI industry; third, a report on specific implementation results.
Our company was established in May 2023. In more than two years, relying on the core technological advantages of multi - heterogeneous and software - hardware collaboration, we have achieved remarkable results in various aspects such as frontier research, product implementation, customer service, and ecological expansion in the past year and grown rapidly. So far, we have joined hands with nearly a hundred partners covering large models, AI chips, and industry customers to continuously build an open and win - win industrial ecosystem.
Since its establishment, Wuwen Xinqiong has been committed to achieving inclusive AI through technological innovation. We believe that water and electricity are the cornerstones of the industrial era, and their scale and efficiency determine the height of a city's industrial productivity. Today, computing power is the soil of the intelligent era, and its scale and efficiency define the boundaries of the digital future. From the development process of AI, the industry has gradually moved from the 1.0 era and 2.0 era to a stage approaching AGI (Artificial General Intelligence).
However, in the process of approaching the AGI critical point, we also face a core contradiction: to achieve the AGI goal, we need to continuously expand computing power resources, but the limited resources are difficult to meet the infinite demand. In the future, resource bottlenecks such as data and computing power may hinder the further development of AGI. In response to this contradiction, we propose a "two - way approach" solution:
1. Accelerate the realization process of AGI by using resources in a more efficient and intensive way, so that the "critical point" of achieving the AGI goal can come faster;
2. Expand the scale of industrial computing power, maximize the value of limited resources, and lay the foundation for the implementation of the AGI era.
The scale of computing resources, intelligent efficiency, and scenarios together constitute the value space for the implementation of artificial intelligence applications. We hope to improve efficiency in intelligent scenarios of any scale. Wherever there is computing, there will be "boundless" intelligence emerging. That is, to meet infinite demands with limited resources.
Based on this idea, we have launched a product system of "large, medium, and small boxes":
Large Box: Wuqiong AI Cloud: For large - scale computing power demands of tens of thousands or even hundreds of thousands of cards, it revitalizes the national computing power resources through multi - heterogeneous mixed training and remote multi - owned resource scheduling, and provides AI empowerment for ultra - large - scale business growth;
Medium Box: Wujie Intelligent Computing Platform: Focusing on activating domestic computing power resources, it specifically solves the technical and application problems in the process of domestic substitution and provides customized intelligent computing services for specific scenarios;
Small Box: Edge - side Computing Power Solution: For single - card, 10 - card level devices and terminals such as mobile phones and PCs, it breaks through the limit of terminal computing power utilization efficiency through software - hardware collaborative optimization and supports the implementation of AI applications at the terminal layer.
Specifically, the large - box "Wuqiong Cloud" relies on the national - level computing power infrastructure and combines the core nodes of the "East - to - West Computing" project. It integrates resources from 26 provinces and 53 core data centers, is compatible with 15 mainstream domestic and foreign chips, and has cumulatively empowered computing power of over 25,000P, which can meet the needs of large - scale training and inference and help realize the value of the AI industry. At the same time, the platform not only manages resources but also encapsulates cloud service applications, including the upper - layer PaaS platform and large - model interface call services, providing customers with a one - stop AI cloud solution and reducing the cost of industrial implementation.
In terms of the ecosystem, Wuqiong AI Cloud provides highly standardized and open interfaces, supporting a unique "platform + self - operation" joint operation model: it not only opens the door to independent intelligent computing centers for seamless integration but also supports enterprises to build their own operation platform sites and integrate them into the underlying computing power supply network. This ecological model has been successfully implemented in Beijing, Shanghai, and Hangzhou, effectively promoting the upgrade of our computing power service from a single resource trading platform to an ecological engine driving the collaborative innovation of the entire industrial chain. Like an engine, it continuously drives down the cost of large - model implementation.
From the perspective of the technical architecture, the underlying computing power infrastructure focuses on building a "user - friendly large - model training and inference platform", with high - performance distributed computing power, efficient fault - tolerance, and high - performance inference capabilities; the middle layer encapsulates MaaS (Model as a Service) large - model applications, following the industry trend - from initial model training to current model calls (such as building Agents and developing applications), we transform convenient computing power into model services to support upper - layer applications and help AI applications be implemented quickly.
There are already phased results: taking the Shanghai Model Speed Space as an example, on this single platform, our daily Token call volume has successfully exceeded 10 billion. It has successfully served more than 10 intelligent scenarios and supported more than 100 cutting - edge AI innovation applications, with the monthly active users of a single application exceeding tens of millions.
Our products are currently playing a role in different business scenarios. Here are a few examples:
AIGC: Covering the "AIGC standard process" - from text generation, image generation to video generation, it provides more inclusive computing power and more stable workflow hosting services for AI - native intelligent agents, helping customers achieve product growth from 1 to 100;
AI + Recruitment: For recruitment scenarios (such as person - position matching, intelligent Q&A, and headhunting assistance), it solves the core pain points of enterprises - on the one hand, it provides a convenient and complete one - stop service of "AI fine - tuning + self - deployment" to meet the customized business needs; on the other hand, it optimizes the response speed through end - to - end acceleration (industry data shows that for every 10% increase in the initial delay of AI interaction, the user churn rate exceeds 50%), ensuring the user experience in highly interactive scenarios.
Looking at the "medium box" Wujie Intelligent Computing Platform, on the basis of the capabilities of Wuqiong AI Cloud, it adds professional services that meet the needs of specific scenarios to help partners achieve efficient and stable large - scale training and inference in production - level requirements. It can help universities and research institutions climb to new heights and make innovations in the frontier fields of artificial intelligence (for example, in cooperation with the Shanghai Algorithm Innovation Research Institute, it has successfully supported the institute to stably run a large - model training with tens of billions of parameters on a domestic GPU cluster of 3000 cards from Muxi for 600 consecutive hours, creating a milestone in the application of domestic computing power); it also empowers AIDC intelligent computing centers to build long - term sustainable operation value (such as in cooperation with Yunnan Mobile, it has efficiently utilized 2000 Huawei Ascend 910B intelligent computing acceleration cards to achieve the distributed deployment and large - scale inference of a model with hundreds of billions of parameters). Currently, it has successfully met the key needs of large - scale model training and inference tasks in nearly 100 large - scale R & D scenarios and helped domestic computing clusters achieve real positive returns.
Finally, there is the "small box" Wuyin Terminal Intelligent Solution. Through an intelligent terminal integrated solution of "edge model + edge engine + edge hardware", we have cooperated with excellent upstream and downstream enterprises of intelligent terminals such as Lenovo, H3C, Yige, and Aixin Yuanzhi.
Ultimately, our goal is to empower all industries. Here, we also hope to meet more application - layer partners through this event to jointly promote the implementation of AI technology. We firmly believe that the AI market is huge and has broad prospects, and the industry is still in its early stage; in the future, we look forward to helping more AI application partners penetrate into various industry scenarios and jointly promote the high - quality development of the AI industry. Thank you!