FEBRUARYASIA BUSINESS OUTLOOK8NEWSROOMZhipu AI CEO Zhang Peng argues that the Western subscription model for AI apps is difficult to replicate in China, forcing Chinese AI companies to choose between building industrial-strength enterprise solutions or becoming obsolete.Founded in 2019 by Tsinghua University researchers, Zhipu AI recently listed in Hong Kong after raising US$558 million in its IPO.Backed by Alibaba, Tencent, and government funds, it reported 312.4 million yuan (US$44.6 million) in 2024 revenue.To survive, Zhipu built its P2P (Paper-to-Product) Engineering System, converting academic research into commercial infrastructure through four stages: Research, Stabilization, Delivery, and Feedback.Zhang stresses that partners want deployable systems, not just papers, describing AI as a loop between science and product.Different payment cultures have pushed Chinese firms away from B2C apps, where price wars and subsidies dominate. "The US has subscription culture; China has freebie wars," Zhang says. He rejects chasing unrelated quick profits, insisting capital must serve core AI research.Open-source competition initially confused clients who equated "open" with "free," but hidden maintenance costs drove them back to Zhipu for services. Zhang defines progress through a five-level AI intelligence hierarchy: Pre-training, Alignment (SFT), Self-learning (RL), Self-cognition, and eventual Consciousness.He says the industry over-invested in pre-training while under-investing in Reinforcement Learning (RL). Competitor DeepSeek's efficiency underscored that the future of AI development lies in better reasoning and systems that recognize their own limits -- not just more data. ZHIPU AI BETS ON ENTERPRISE AI OVER CHINA SUBSCRIPTION MODEL· Zhipu AI shifts from B2C to enterprise AI due to China's pricing culture· Uses P2P system to turn AI research into commercial products· CEO says Reinforcement Learning is the next AI frontierThe Beijing Innovation Center of Humanoid Robotics Co. Ltd. has raised over 700 million yuan (US$100 million) in its first funding round, marking a major step in China's push toward industrial-scale humanoid robotics and embodied AI.The round included state-backed investors such as the Beijing Robotics Industry Development Investment Fund, E-Town Capital, and the Beijing Gaojingjian Industrial Development Investment Fund, along with TH Capital. Strategic investors Baidu and Kyland Technology also participated.The center, a state-backed robotics hub registered in the Beijing Economic-Technological Development Area (Beijing E-Town), was established to accelerate real-world deployment of humanoid robots.It collaborates with firms producing complete robots, core components, and large robot models. Funding will advance embodied AI technologies and support products like the Tiangong robot, described as the world's first full-size, pure electric-drive running humanoid robot, with a tested speed of 12 km/h.Tiangong operates on the Huisi Kaiwu universal embodied AI platform, developed by the center. A newly opened pilot manufacturing and validation platform helps transition humanoid robots from prototypes to larger-scale production.The initiative aligns with China's national plan to build an early-stage innovation system for humanoid robots by 2025, guided by the Ministry of Industry and Information Technology.The center has released Tiangong design documents and APIs, enabling hundreds of institutions to develop add-ons.Beijing is also strengthening shared infrastructure through a national robot inspection and testing platform, a robot standardization technical committee, and a planned Beijing robot industry development fund to accelerate a unified national supply chain. BEIJING HUMANOID ROBOTICS HUB RAISES $100M IN FUNDING· Beijing's humanoid robotics hub raised $100M from state and tech investors· Funding accelerates embodied AI and scaling of the Tiangong humanoid robot· Move supports China's 2025 national strategy for industrial humanoid robots
< Page 7 | Page 9 >