Why Alibaba’s Top AI Engineer’s Sudden Exit Shook the Global AI Landscape

Background of Lin Junyang

Born in 1993, Lin Junyang completed a computer‑science bachelor’s degree at Peking University and then earned a master’s in linguistics and applied linguistics. The combination of solid low‑level coding skills and deep insight into natural language formed the foundation for his later work on large language models.

Career at Alibaba and Contributions

After graduating in 2019, Lin joined Alibaba DAMO Academy’s Intelligent Computing Lab as a senior algorithm engineer. He contributed heavily to the multimodal pre‑training model M6 and, following the 2022 merger of DAMO teams into Alibaba Cloud’s Tongyi Lab, was appointed technical lead of the Tongyi Qianwen series at the age of 29. Under his leadership the team released models of various scales (7B, 14B, 72B) and opened the Qwen2‑72B model, which achieved top‑ranked evaluations at institutions such as Stanford. In 2025 the team launched the trillion‑parameter flagship Qwen3‑Max, trained on 36 TB of data and surpassing contemporary international models on many benchmarks. By April 2025, the Tongyi open‑source portfolio exceeded 200 models, amassed over 300 million downloads, and spawned more than 100 000 derivative projects, earning praise as the world’s largest open‑source model family.

Rapid Expansion and Organizational Tensions

On March 2 2026 Alibaba announced a unification of its AI branding under the name “Qianwen,” consolidating both foundational and industry‑specific models. The consumer‑facing app was renamed Qianwen App, while the Tongyi Lab retained its role as the internal R&D organization. A promotional “red‑packet” campaign during the Chinese New Year drove nearly 200 million orders, pushing daily active users of the app to 73.5 million—a 940 % increase—and propelling it to the top of free‑app charts. The same day the lab released four compact Qwen3.5 models (0.8B, 2B, 4B, 9B) that run on consumer GPUs and even mobile devices, preserving multimodal capabilities and supporting commercial use under an open‑source licence. These rapid product launches highlighted a clash between the open‑source ethos and aggressive commercialisation, as resources were stretched to meet both development and market‑facing goals.

The 48‑Hour Personnel Shock

On the afternoon of March 3, an internal meeting was held to discuss team splitting, responsibility adjustments, and the introduction of new leadership. The discussion escalated, and Lin abruptly left the room, submitting his resignation. Hours later the news spread within the team, prompting emotional reactions and tears from long‑time colleagues. By early March 4, Lin posted a brief farewell on social media, which garnered 13 000 likes and nearly 6 million views, while other core contributors—post‑training lead Yu Bowen and multimodal/code lead Li Kaixin—also announced their departures. An emergency all‑hands meeting on March 4 was attended by CEO Wu Yongming, Chief Talent Officer Jiang Fang, and Cloud CTO Zhou Jingren. Jiang acknowledged communication gaps in the restructuring, while Wu admitted he should have detected the issues earlier. The company formally approved Lin’s resignation on March 5, confirming that Zhou Jingren would continue to lead the lab and that a new “foundation‑model support group” would coordinate compute, funding, and talent resources, reaffirming Alibaba’s commitment to an open‑source strategy.

Industry Reaction and Talent Implications

The sudden exit triggered intense speculation among investors and media. Analysts estimated Lin’s market value at over $100 million and warned of potential ripple effects if other senior technical leaders followed suit. Observers noted that China’s AI firms face a structural compute imbalance compared with overseas counterparts, forcing domestic teams to optimise engineering and infrastructure under tight resource constraints. The open‑source‑versus‑commercial tension was highlighted as a core challenge: while open contributions build reputation and ecosystem growth, they do not translate quickly into clear profit, conflicting with short‑term business metrics such as app DAU and enterprise revenue. The episode underscored how talent retention now hinges on granting technical leaders autonomy, sufficient compute budgets, and a long‑term open‑source vision.

Lessons for Large‑Scale AI Enterprises

The case illustrates a universal dilemma for AI giants transitioning from founder‑driven, high‑autonomy phases to corporate‑level, standardized operations. Centralised resource scheduling and uniform management structures are necessary for scaling, yet they risk alienating top engineers who value independence and open‑source principles. Over‑aggressive team fragmentation can be perceived as power stripping, prompting talent loss. Providing competitive compensation while preserving genuine technical leadership and long‑term research freedom has become a decisive factor in the global talent war. Alibaba’s experience suggests that maintaining a balance between unified corporate strategy and the autonomy of elite AI teams is critical for sustaining innovation and market leadership.