Alibaba has handed over its AI business to the post-90s generation.

Those born in the 1990s are no longer "young prodigies" in the venture capital circle, but at large tech companies, they are in the prime age group for generational transition. The 2026 Alibaba Cloud Apsara Conference delivered the clearest signal on generational succession.

On the keynote forum speaker list, after Joseph Tsai and Wu Yongming, there is a name relatively unfamiliar to the public — Liu Daoyiheng, Project Lead of the Qwen large model. In the afternoon, at the Qwen Office Forum, Chen Yusen, the youngest business unit CEO in Alibaba's history, made his first official public appearance.

Inside the exhibition hall, you can see hundreds of junior high school students from Beijing in unified school uniforms, second-generation enterprise successors from various regions in the Yangtze River Delta who came for research tours, and groups of young foreign developers. Three generations gathering at the same venue form a more complete industrial landscape than any product launch event.

After attending this Apsara Conference, I have a distinct feeling that AI has accelerated generational transition, and Alibaba has handed over the position to define the next era to the generation that is most eager to seize AI opportunities and put AI into practical implementation.

Two post-90s Executives in Charge of Models and Products Respectively

At the keynote forum, Joseph Tsai's opening remarks set the tone for the AI industry, emphasizing that the focus should shift from technical imagination to large-scale implementation in real scenarios. Then Wu Yongming took the stage, putting forward the view that "the total amount of machine thinking is less than 3% of that of humans, and it will reach more than 1000 times that of humans in the future", and outlined a roadmap for the global data center market to exceed 20GW by 2032 for the capital market.

When Liu Daoyiheng stepped onto the stage, many people were taken aback. This senior Alibaba AI researcher, born in 1993, wore a yellow sweatshirt in his photo, standing among formally dressed industry leaders including Joseph Tsai, Wu Yongming, Zheng Bo and Li Feifei, and some joked that he looked like "an intern who accidentally walked into a meeting of bigwigs".

He is exactly Alibaba's "four-character brother" in the public's playful description.

But his identity is by no means that of an intern. He is the head of the Token Foundry Qwen LLM project under Alibaba's ATH Business Group. It was not until the day before the opening of the Apsara Conference that the public confirmed the outcome of the rumors that had been circulating since March: the new head of the Tongyi Qwen large language model is this Zigong native born around 1993, who completed all his bachelor's, master's and doctoral degrees at Sichuan University.

Compared with his identity, Liu Daoyiheng's resume goes against the typical "star executive" narrative: no Ivy League education, no overseas experience at top foreign tech companies. He was admitted to the School of Computer Science of Sichuan University in 2012, and was recommended for the 3+2+3 combined bachelor-master-doctor program in 2015 with the top comprehensive score in his major. After that, he continued to make breakthroughs in the academic circle, and in 2020, he became the first top-tier selected candidate in Southwest China for Huawei's "Genius Youth Program".

Shortly after graduating with his doctorate, Liu Daoyiheng joined Qwen, which was just launched in 2021. In the following five years, he only focused on one thing: pre-training. He led the development of the full series of Qwen language models from Qwen1 to Qwen3.5, participated in the open source of more than 300 models in total, published more than 100 papers with over 36,000 citations on Google Scholar, and won the Best Paper Award at NeurIPS 2025 as the corresponding author.

In the foundational model circle where personnel turnover is extremely fierce, Liu Daoyiheng is the only person who has been present throughout the whole process and never left the team.

In the six months after Lin Junyang left, Alibaba first established the ATH Business Group led by Wu Yongming personally at the organizational level, then merged to form the Token Foundry Division directly managed by Wu Yongming, and transferred chief scientist Zhou Jingren to the AI Future Research Institute to focus on long-term research. Against this background, a person who has managed pre-training, taken charge of post-training and coding, and left his mark on every version of Qwen is the most suitable choice. So Alibaba chose him for his certainty and reliability.

Chen Yusen, on the other hand, is a sample from a different direction.

Born in 1992, he was a core member of China's top hacker team Blue Lotus from the Qiushi Science Class of Chu Kochen Honors College of Zhejiang University, and won numerous awards at DEFCON CTF and Pwn2Own. He founded Longting Technology at the age of 22, sold the entire company to Alibaba Cloud at 27, and achieved the financial goals that most people spend their whole lives pursuing in more than four years.

Later, in 2021, he gave up unvested equity and left Alibaba to start a new business, and returned to Alibaba Cloud in 2025 to incubate the AI Agent product MuleRun. Shortly after the product was launched, paying users covered 43 countries with almost zero promotion, and more than 40% of monthly paying users spent over 200 US dollars per month.

The story that followed is relatively familiar to us. In June this year, he succeeded Chen Hang (Wu Zhao) as CEO of DingTalk. In July, Alibaba integrated three scattered AI Agent product lines including QoderWork, Wukong and MuleRun into the Qwen Office Division, which is under his unified management.

Among Alibaba's product portfolio, Qwen Office is probably the most familiar one for C-end users, so the Qwen special session became one of the most popular yet most relaxed forums at this year's Apsara Conference.

Chen Yusen, with curly hair, took out a microphone wrapped with three small green octopuses from behind, and opened his speech by self-deprecating, saying "I hope there can also be proper relaxing moments in work". He rarely looked at the teleprompter, and the title of his speech is the concentration of his product philosophy over the years: Context is All You Need.

Putting the two together, Liu Daoyiheng is in charge of models, which is the source, while Chen Yusen is in charge of products and commercialization, which is the export. One is responsible for creating intelligence, and the other is responsible for commercializing intelligence. So Alibaba has handed over both ends of the entire AI chain to the post-90s generation, which means that youthfulness is no longer a gesture, but a real arrangement placed on the table of core business.

Second-generation Enterprise Successors on Research Tours

Similarly, the flowing crowd in the exhibition hall is also full of young faces, most of whom represent the present of Chinese private manufacturing enterprises.

During the three-day event, I met multiple batches of visiting groups with different badges in the exhibition hall. Most of them are from manufacturing and consumer goods private enterprises in the Yangtze River Delta, who signed up in groups, moved in unison, and are young successors full of curiosity and awe for AI.

Their presence is not difficult to understand. The Yangtze River Delta is the region with the highest density of private enterprises in China, and a large number of manufacturing enterprises founded in the 1980s and 1990s are in the period of generational handover. The older generation completed informatization with ERP and management consulting, while the new generation is faced with the exam of intelligent transformation. For them, the Apsara Conference has changed from a tech exhibition for casual viewing to a frontline reference for their next technology selection.

A second-generation enterprise successor friend who came for the research tour told me that the No.3 Creation Pavilion is the most practical. There is a Skill wall in the pavilion, covering 17 major professional identities and more than 200 reusable Skills, corresponding implementation solutions can be found for scenarios such as finance, law, human resources, and even marketing.

Of course, for her, these are just functions she inspected on the side. The focus of this visit is to find a set of AI systems that can be applied to her own products, and to learn practical experience from other consumer brands.

When asked why she chose to visit in a group instead of visiting alone, this friend admitted that the exhibition hall is huge with a large number of participating enterprises, and it takes time and energy to filter by herself. At the same time, her understanding of enterprises and AI is limited. More importantly, all the people in the visiting group are second-generation enterprise successors, so they can communicate and expand their social network.

Their way of visiting the exhibition is also simple and straightforward, and the only question they keep asking repeatedly is: "How can this product be used in our factory, how much does it cost, and how long will it take to see results". So for this group of second-generation enterprise successors, the significance of AI has been recalibrated: it is a clearly calculable return, for example, more than 300 steps are shortened to a few hours, 50 working days are reduced to 30 days, and a full day of work for a dedicated person is compressed to the minute level.

A more subtle change lies at the organizational level. For the previous generation taking over the business, tangible assets such as factories, machines, channels and customer relationships are all about people and business, and it is fine to follow the established routine. When the next generation comes to the forefront, what they need to manage is far more than people and business, and everything changes and adjusts every few months. Following the old routine will lead to major problems.

School Uniforms and Foreign Faces

In the afternoon of the third day, I met a group of students in white school uniforms in the corridor of the exhibition hall, with the leading teacher holding a sign. The words "The High School Affiliated to Renmin University of China" were printed on their school uniforms.

Later, when talking with people around, I learned that it was a study tour group that came all the way from Beijing. They are ninth-grade students from The High School Affiliated to Renmin University of China, with a four-day trip costing 4,000 yuan per capita, and the Apsara Conference is their third stop. Although they have different preferences for AI, they visited the exhibition carefully driven by their school assignments. In a popular video on Douyin that day, several teenagers in school uniforms were chatting and laughing with a CTO in front of the booth.

This is not the first time that the Apsara Conference has received students. At previous conferences, Wang Jian once had a dialogue with seven children who used the "Youth Cloud Classroom". Last year, teachers and students from a mountain school at an altitude of 4,000 meters in Kangding, Sichuan, danced with humanoid robots in front of the booth. But the difference this year is that students are no longer invited as decorative guests, but the main participants who come here on their own initiative.

In addition to students from The High School Affiliated to Renmin University of China, the Student Union of Zhejiang University organized nearly 100 students to come by chartered bus. The Alibaba Cloud Big Data College of Changzhou University moved the entire top-notch talent class into the Computing Power Pavilion and the Creation Pavilion, calling this visit "on-site teaching outside the classroom".

Another striking scene at the venue is the presence of young foreign faces.

This year's Apsara Conference has more than 2,000 speakers from over 50 countries and regions. Young overseas visitors communicating with booth staff in English can be seen everywhere in the exhibition area. Some flew here specially from Southeast Asia and the Middle East, and sat in front of the Agent development booth for half a day. The data of last year's Apsara Conference shows that the total number of offline participants exceeded 120,000, with guests covering more than 70 countries and regions around the world. It is obvious that the proportion of young people in this event is increasing visibly.

For Alibaba, this is almost an inevitable strategic choice.

Wu Yongming said, "The Industrial Revolution liberated human physical strength, and gave birth to sports and art. Machine intelligence will liberate human mental strength, leaving heavy tasks to machines, and leaving time, creativity and the perception of beautiful things to human beings." This sentence is addressed to the entire audience, and it is the young people under the stage who will personally practice and verify whether this future is really so promising.

This article is from the WeChat Official Account "Dong 40th Street Capital" (ID: DsstCapital), written by Zhang Xue, and published with authorization from 36Kr.