At the 2026 Zhongguancun Forum in Beijing, Yang Zhilin, the visionary founder of Moonshot AI (Kimi), signaled a fundamental shift in the global artificial intelligence race. As the industry moves beyond the initial hype of large language models, the focus is pivoting from pure algorithmic breakthroughs to what Yang describes as the 'token factory'—a regime where industrial-scale compute, infrastructure efficiency, and energy costs dictate the winners of the intelligence era.
Yang contends that the debate between open-source and closed-source models is approaching a critical inflection point. While proprietary models currently hold significant market share, he argues that once the capabilities of open-source alternatives reach parity, the collaborative ecosystem of the open-source movement will provide an insurmountable advantage. This shift is expected to commoditize intelligence, allowing developers to focus on scaling token output rather than refining isolated architectures.
China’s strategic position in this new landscape relies on its unique trifecta of advantages: a robust talent pipeline from elite institutions like Tsinghua and Peking University, an expansive energy and infrastructure base, and a deeply ingrained culture of technical openness. According to Yang, as model training transitions toward AI-led R&D—where artificial intelligence itself designs new network architectures and reward functions—the speed of innovation will accelerate beyond human-centric limitations.
The ultimate goal of this industrialization is the deployment of AI Agents capable of autonomous, long-term task execution. Unlike simple chatbots, these agents are designed to integrate into the professional workflow, potentially acting as a force multiplier for global GDP. As these agents transition from niche tools for early adopters to essential assistants for all knowledge workers, the consumption of tokens is projected to grow exponentially, effectively linking a nation’s economic output to its capacity for synthetic intelligence production.