Alibaba Group Holding unveiled a broad suite of artificial intelligence products — including its latest large language model, cloud infrastructure upgrades and proprietary chips — at the Alibaba Cloud Summit in Hangzhou on Wednesday, 20 May 2026, declaring its ambition to become China's dominant 'AI factory' as autonomous agents reshape enterprise computing.

The 'AI factory' vision

Liu Weiguang, senior vice-president of Alibaba's cloud computing unit, framed artificial intelligence as a new form of industrial manufacturing, one that generates revenue through what he called 'training and inference factories'. Liu described Alibaba as the only AI and cloud company in China operating across all five layers of the full AI stack — spanning chips, agentic cloud, AI models, model service platforms and agentic applications.

'What we're building is China's AI factory,' Liu said, positioning the company's vertically integrated approach as a structural competitive advantage over domestic and international rivals.

Qwen3.7-Max: built for autonomous agents

Alibaba introduced Qwen3.7-Max, its newest proprietary large language model, designed as 'a robust foundation' for AI agents. According to the company, the model is optimised for agentic coding, complex reasoning and what it terms 'long-horizon tasks' — multi-step missions requiring continuous decision-making over extended periods.

The company said Qwen3.7-Max can autonomously operate for up to 35 hours without performance degradation, a claim that, if validated, would mark a significant step forward in AI agent endurance. Zhou Jingren, recently appointed as chief AI architect of the newly formed Alibaba Group Technology Committee — and formerly chief technology officer of Alibaba Cloud — said the model consistently ranked among the top tier on various benchmarks and outperformed all other AI models in China.

Custom silicon enters the picture

Alongside the model launch, Alibaba also announced custom chip developments as part of its bid to reduce dependence on third-party semiconductor suppliers such as Nvidia. The company's in-house chip division, T-Head, is central to this strategy, enabling Alibaba to claim end-to-end control of its AI infrastructure stack.

The move mirrors a broader industry trend in which hyperscalers — including Google and Amazon — have invested heavily in proprietary silicon to optimise cost and performance for their specific AI workloads, particularly as global chip supply chains remain under geopolitical pressure.

Competitive backdrop

Alibaba's announcements come as competition in China's AI landscape intensifies, with domestic rivals and international players such as OpenAI and Anthropic advancing their own agent-capable models. The company's five-layer stack narrative is a direct response to the growing complexity of enterprise AI deployment, where customers increasingly demand integrated solutions rather than point products.

Industry analysts have noted that the race to control full-stack AI infrastructure is accelerating globally, and Alibaba's positioning in Hangzhou signals it intends to be a primary supplier — not merely a participant — in that buildout.

What's next

The durability of Alibaba's 'AI factory' thesis will depend on enterprise adoption of Qwen3.7-Max in production environments and the commercial performance of its custom chips against established alternatives. Investors and developers will be watching whether the 35-hour autonomous operation claim holds up under real-world workloads — and whether Alibaba Cloud's integrated stack can convert the summit narrative into measurable revenue growth in coming quarters.