Wang Jian, who founded Alibaba Cloud in 2009 and is often credited as the architect of China’s domestic cloud-computing industry, has spent the past year making a specific, contrarian argument inside a domestic AI race that has largely been framed around who can build the biggest, most capable foundation model. “We really need to fund some people with creativity to write application[s],” Wang told Bloomberg in an interview published July 29, 2025, arguing that foundation models like Alibaba’s own Qwen and rival DeepSeek’s models are already powerful enough — what China’s AI ecosystem actually lacks, in his view, is imaginative application-layer development capable of unlocking their broader commercial and social value.
Skeptical of the AGI Framing
Wang, who now serves as director of Zhejiang Lab, used the same interview to push back on the industry’s common practice of framing AI progress as a linear march from narrow AI toward artificial general intelligence and then artificial superintelligence, calling that categorization “misleading and non-essential.” He described AI development instead as “a marathon rather than a sprint,” expressing confidence that no single lab’s current technical lead — implicitly including US frontier labs like OpenAI — would translate into an insurmountable long-term barrier for Chinese competitors. Wang pointed to China’s enormous domestic market as a practical advantage in that marathon, describing it as “a testbed of our new technology” that allows for faster real-world experimentation and validation than smaller markets can offer.
From Infrastructure to Science
By June 2026, Wang had extended his public commentary from application-layer creativity to AI’s effect on the scientific process itself. Speaking at the 8th Beijing Zhiyuan Conference, he argued that AI has already produced a qualitative transformation in how scientific research gets done, tracing the evolution from early language models that could only process text to current multimodal systems capable of understanding code, data, and natural language together — and, in his framing, capable of meaningfully distinguishing between the two. He described the impact on scientific research as following a logic similar to AI’s earlier disruption of professional programmers: a genuine restructuring of the underlying workflow rather than a simple productivity add-on, while stopping short of predicting AI would replace human researchers outright.
The Infrastructure His Original Vision Set in Motion
Wang’s application-layer argument plays out against the backdrop of an extraordinary capital commitment from the company he founded. Alibaba has pledged to invest at least 380 billion yuan (about $53 billion) over three years into cloud computing and AI infrastructure, and has said it now expects to overshoot that plan. The company’s AI-related revenue has posted eleven consecutive quarters of triple-digit growth, and Alibaba Cloud’s own Qwen model family has become the world’s most-used open-weight model family by download count, having crossed 1 billion downloads with more than 200,000 community-built derivatives. Alibaba has also consolidated its Qwen models, its Wukong enterprise-agent platform, and its consumer AI applications under a single new Alibaba Token Hub Business Group, an organizational move aimed at tightening the link between model development and the kind of real-world application deployment Wang has been publicly advocating for.
A Founder Who Stepped Back From Operations, Not From Influence
Wang left day-to-day operational leadership of Alibaba Cloud years ago, but has remained a visible, publicly quoted voice inside the company’s broader technical direction, including through his current role directing Zhejiang Lab, a state-backed research institute, and through recurring appearances at major domestic AI conferences. That positioning — no longer running the business he founded, but still shaping the intellectual framing around what China’s AI industry should prioritize next — has made him a distinct kind of figure in China’s tech landscape: less a builder of a single current product and more a standing reference point for how the country’s AI strategy should be argued about in public.
What This Means for Philippine Founders
Wang’s core argument — that foundation models have become powerful enough, and the real bottleneck now is creative application-layer development — is directly actionable for Philippine AI founders who will never have the capital to build a competing foundation model but can absolutely build the kind of imaginative, locally-specific application he says the ecosystem needs, whether on top of Qwen’s open-weight models, DeepSeek’s, or any other openly available base. Qwen’s billion-plus downloads and 200,000-plus derivatives are a concrete signal that a genuinely usable, freely licensed alternative to closed US models now exists at scale, lowering the cost of entry for any Philippine startup that wants to fine-tune or build on top of a capable open model rather than pay per-token fees to a closed provider. And Wang’s skepticism of the AGI/ASI framing — treating AI progress as a long, gradual marathon rather than a single dramatic breakthrough — is a useful corrective for founders here who might otherwise feel structurally behind in a race that, on Wang’s own telling, has no clear finish line to have already missed.
Share this article