Jack Ma has been making conspicuous public appearances as Alibaba steps up an urgent campaign to catch up in China's intensifying AI wars. On March 3 the founder convened a rare, full-scale meeting of Alibaba and Ant Group's core management at YunGu School to discuss the opportunities and threats posed by generative AI, signalling that the company regards the moment as existential.
Photographs and participant lists from the meeting show an unusually broad cast of senior figures — Joe Tsai, CEO Wu Yongming, Shao Xiaofeng, e-commerce chief Jiang Fan, Ant Group chairman Jing Xiandong and CEO Han Xinyi — underscoring that this was meant to be more than a technical seminar. The gathering read as a deliberate attempt to steadied nerves and reassert top-level coordination amid mounting external pressure and internal churn.
The timing was not accidental. Ma had already surfaced in early February at Alibaba's Qianwen project headquarters in Hangzhou, close to a major push by the group: on Feb 2 Alibaba’s Qianwen unit announced a Spring Festival campaign that would plough more than RMB 3 billion into red packets, sponsorship of four regional gala shows and the public release of the Qwen3.5 model. Those moves were designed to generate scale quickly and to demonstrate that Alibaba would be a heavyweight in the AI consumer layer.
But usage figures show the gap is still wide. Data cited from QuestMobile place ByteDance’s Doubao at the head of the field, with peak daily activity many multiples above Alibaba's Qianwen: peak daily users during the festival reached roughly 145 million for Doubao versus 73.5 million for Qianwen and 40.5 million for Tencent’s Yuanbao. Analysts warn that stimulus-driven growth—red packets and marketing—yields fleeting lift: seven-day retention from cash incentives typically falls below 20 percent and 30-day retention can drop under 5 percent.
The urgency is compounded by a high-profile departure inside Alibaba’s AI ranks. Lin Junyang, the engineer widely credited with leading Qwen3.5’s compact-model series and a visible figure in open-source model communities, announced his resignation on X. Lin had been a global figure in model development and his exit sent immediate ripples through the developer ecosystem; it followed an internal rebranding that unified AI efforts under the Qianwen label and reportedly narrowed his remit while shifting his performance metrics from research excellence to user numbers.
That personnel shift helps explain why Ma’s March meeting convened the full “all-star” management line-up. Stabilising morale, reasserting strategic priorities and repairing organisational incentives are necessary if Alibaba intends to convert marketing heft into durable product advantage. The company’s attempt to centralise coordination around Ma at this juncture suggests leadership believes distributed, business-unit-led campaigns alone will not close the technology gap.
Alibaba’s predicament is not only commercial. Its diminished primacy in the AI era has corollaries in political and civic visibility. Ma’s recent placements at high-profile state and industry events have been more peripheral than before, while a generation of younger, AI-first startups in Hangzhou and beyond — accompanied by government and foreign delegations — are increasingly treated as the city’s new tech face. That shift matters because access, local support and regulatory goodwill still shape commercial outcomes in China.
Looking ahead, Alibaba faces a two-front challenge: to regain model leadership and to keep the talent and organisational incentives aligned for a sustained AI push. Large-scale marketing and ecosystem coordination can buy time and scale, but the firm’s long-term standing will hinge on whether it can retain people able to deliver model-level innovation and convert temporary spikes into habitual usage. Jack Ma’s renewed visibility is a sign that Alibaba recognises the stakes, but the next phase will test whether reasserting a centralised command can translate into technological lead rather than just headline-grabbing campaigns.