Xu Li, chairman and CEO of Shanghai-based SenseTime and a member of the Shanghai Municipal CPPCC, has sketched a provocative roadmap for China’s next phase of artificial intelligence development. He argued that while the United States’ model-innovation landscape is shaped by superstar-led organisations, China’s challenge is to cultivate its own “super individuals” and construct teams around them to convert large models into real-world delivery.
Xu frames large, or foundation, models not simply as research artifacts but as emerging end-to-end production tools. He predicts that when these models reach sufficient maturity they will create integrated delivery capabilities — systems that can take workflows from input to commercial output — but still require human–machine collaboration. The crucial inflection, in his view, comes when models empower single individuals to complete delivery tasks, unlocking a sharp productivity surge.
That judgement carries a practical caveat. If businesses treat large models merely as management tools for trimming costs and improving efficiency, commercial adoption will stall because the revenue and value chain will not “close the loop.” By contrast, when models are productised into process-driven delivery systems that let people execute entire tasks end-to-end, the economic impact will be far greater and more sustainable.
Xu’s prescription speaks to deeper institutional questions for China’s innovation ecosystem. The U.S. pattern he points to — high-profile founders, star researchers and venture-backed startups that rapidly scale research into products — emphasises individual leadership and market signaling. China’s strengths lie in industrial deployment and scale, but to translate foundation models into transformative commercial platforms it must combine institutional muscle with mechanisms to identify, reward and amplify exceptional talent.
The implications extend beyond corporate strategy. If large models truly empower individual contributors to own delivery, expect shifts in organisational design, skills demand and the fortunes of small and medium enterprises. Firms that integrate model-driven workflows into product offerings could leapfrog competitors; workers with the ability to harness these tools will see their productivity and market value rise, while routine roles face compression.
On the geopolitical front, the divergence in innovation models matters. Different paths to commercialisation — superstar-led breakthroughs versus ecosystem-driven scaling and talent cultivation — will shape which products and governance norms gain traction globally. For Chinese policymakers and tech leaders, the task is to create incentives, talent pipelines and interfaces between models and industry that convert technical capability into repeatable commercial delivery.