The Report analyzes new developments and breakthroughs in full-stack technological innovation for generative AI, the latest expansion of model capabilities, a new paradigm for computational efficiency driven by synergy between the foundation layer (comprising computing chips and software frameworks) and the model layer (large-scale AI models), and the model layer built by large models, as well as the comprehensive progress in application-layer innovation marked by intelligent agents. It aims to understand and discover the latest development trends in full-stack generative AI innovation from a patent perspective, offering insights to address existing bottlenecks and identify future innovation pathways.

First, China’s generative AI ecosystem demonstrates deep integration and coordinated growth across the foundation layer (AI chips/software frameworks), model layer, and application layer, forming a development pattern where "demand drives technology, and technology empowers scenarios." From 2017 to 2024, China filed 167,634 publicly disclosed patents in generative AI full-stack technologies, with a 31.6% CAGR in applications. The number of patent applicants surged from 2,257 to 13,075, with an average annual growth rate of 24.6%, showing a revolutionary innovation and development trend of AI technology under the guidance of national strategies, the promotion of local policies, and the full support of technological innovation.

Second, Chinese innovators increasingly prioritize patent quality, with enterprises spearheading autonomous innovation and industrial ecosystem formation. The report shows that Top 10 corporate innovators in terms of the number of patent applications (e.g., Baidu, Huawei, Inspur, Tencent, Alibaba) account for 10% of total patent filings, emerging as pivotal forces in technological breakthroughs, efficiency gains, and scenario-based applications.

Third, the full-stack development of generative AI follows a trinity model of technology-driven innovation, scenario-led demand traction, ecosystem-enabled support, among which representative innovation entities demonstrate the core role of value transmission and pioneering leadership.

Fourth, the accelerated filing of multimodal and reasoning technology patents confirms that in 2024, generative AI models are gradually acquiring real-world comprehension and complex decision-making abilities, marking tangible progress toward Artificial General Intelligence (AGI).

From Electronic Intellectual Property Center, MIIT

March 31st, 2025