中美通用人工智能发展路径比较研究：战略竞争与发展道路

Editor-in-ChiefCHEN Dongxiao Exec. Editor-in-ChiefLI Xin Exec. ProducerMAO Ruipeng EditorCHEN Xue © 2026 Shanghai Institutes for International Studies The views represented herein are the author’s own and do not necessarily reflect theinstitutional positions. All rights reserved. No part of this publication may be reproduced or transmitted in anyform or by any means without permission in writing from the Shanghai Institutes forInternational Studies. Please direct inquiries to: Shanghai Institutes for International Studies195-15 Tianlin Road, Xuhui DistrictShanghai 200233, PRCTel/Fax: +86 21 64850100http://www.siis.org.cn This publication can be downloaded at no cost at SIIS website. About the Author Professor of the School of Political Science & International Relations at Tongji University and VisitingResearch Fellow at Shanghai Institutes for International Studies (SIIS) Contents Abstract····························································································································01Introduction·····················································································································02 Part IArtificial General Intelligence (AGI): Definitions and Controversies····· 05 2.1 Technology Roadmap(s) Selected: Strategic Consolidation Centered on Large Models··· 102.2 Organizational Mechanism: A Whole-of-Government, Whole-of-Society Model·············· 112.3 Resource Allocation: From Market-Driven to Strategic Concentration································ 12 Part III China: A General-Purpose Infrastructure Approach······························ 153.1 Technology Roadmap(s) Selected: Application-Driven with Empowering All Industries at the Core······················································································································································· 163.2 Organizational Mechanism: Interagency Coordination··························································· 173.3 Resource Allocation: AI Infrastructurization and Capability Diffusion································· 18 Conclusion························································································································21 Abstract Artificial general intelligence (AGI) has emerged as a critical frontier for the next phase of technologycompetition and global governance. China and the U.S., as two actors of prime importance in thedomain of artificial intelligence (AI), are shaping different approaches to developing AGI. Thisdivergence between the two sides, this paper argues, should be construed not simply as a technologyrace or geopolitical confrontation, but as the result of differing practical choices made by the twocountries in light of their own innovation systems, industrial structures, resource endowments,security concerns and governance traditions. The U.S. takes a model-centric approach whichemphasizes cutting-edge foundation models, the scaling law, advanced chips, large-scale computeinfrastructure, and close coordination between leading tech companies and the federal government.Under this approach, AGI has been increasingly portrayed as a strategic capability closely associatedwith national competitiveness, military superiority, scientific discoveries, and global technologyleadership. In contrast, China’s AGI development underscores infrastructure- and application-oriented features by emphasizing the following areas: the wide diffusion of AI capabilities acrossindustries, the development of public and industrial AI infrastructure, the deep integration betweenAI and the real economy, and a governance framework balancing innovation, security and socialstability. This paper compares the AGI development pathways and governance models of Chinaand the U.S. in terms of technology roadmap, state-market relationship and governance logic. Thestudy finds that the American approach features the concentration of cutting-edge capabilities, thedominant role of the private sector, and security-driven nationwide mobilization, while the Chineseapproach features application diffusion, state-led infrastructure development, and a governancemodel embedded in industrial development. Such differences suggest that the competition over AGIis not only about who will be the first to reach technology frontiers but also about the varying modelsadopted by countries to organize innovation, manage risks, and shape the digital order of the future. Introduction Artificial general intelligence (AGI) has been increasingly viewed as a transformative technologythat is likely to reshape economic production, military power, scientific research, social governance,and the structure of international competition. Unlike mission-specific artificial intelligence (AI)in the narrow sense, AGI is often understood as a form of AI that is widely adaptive, capable ofgeneral problem-solving, and able to play a role in multiple domains. Despite great uncertainty