Industry News

by Wang Kai

BEIJING, Sept 11 (China Economic Net) - China's next industrial growth stories may not come less from another generation of AI models, but more from the machines and systems that put those models to work.

The shift is highlighted in a white paper released by Frost & Sullivan and LeadLeo during the ongoing China International Fair for Trade in Services (CIFTIS) in Beijing.

White Paper on China's Industrial Development Trends for the Next 50 Years (5th Edition) by Frost & Sullivan in partnership with LeadLeo.

Among others, the white paper identifies key opportunities in China's rapid deployment of AI across the real economy and at scale. The trend is unfolding on three fronts: At the infrastructure level, intelligent-computing capacity is projected to grow nearly sevenfold by 2028, while daily large-model token calls have surged from about 100 billion to more than 140 trillion in just two years, providing the computing power and usage base needed for wider AI adoption. At the physical-application level, the Embodied AI market is expected to grow to trillion-yuan within a decade, as AI moves from digital systems into the physical world. At the industrial level, collaborative robots, humanoid robotics, and eVTOL aircraft are expected to become important growth pillars for manufacturing.

The Next Frontier: Physical AI

China's AI industry is moving from building foundational models to deploying them at scale. The country has developed a full-stack ecosystem covering computing infrastructure, AI models and applications. In 2025, the number of AI-related enterprises reached 757,000, up 40.9% year on year, while related patent applications rose 30.7% to 29,000. As of June 4, 2026, the number of active AI-related companies had reached 2.887 million.

The infrastructure supporting this expansion is growing just as rapidly. China's intelligent-computing capacity is projected to increase from 416.7 EFLOPS in 2023 to 2,781.9 EFLOPS in 2028, representing a compound annual growth rate of 46.2%. Large models already account for around 60% of intelligent-computing demand, while the large-model market is expected to grow at a CAGR of 44.9% between 2024 and 2029. Average daily token calls for large models in China surged from about 100 billion in early 2024 to more than 140 trillion in March 2026.

Graphic from Frost&Sullivan

As leading models become more comparable in areas such as general question answering, coding and multimodal understanding, enterprise buyers are increasingly prioritizing inference cost, response latency, tool-calling capabilities, and industry-specific adaptability over raw benchmark performance.

This shift is opening up a new phase of AI development: from generating content to executing tasks. AI agents are increasingly being integrated into core workflows in sectors including finance, manufacturing and healthcare. “For consumers, AI-powered search, learning assistants and productivity tools are becoming part of everyday digital life,” the white paper notes.

Graphic from Frost&Sullivan

Healthcare is one area where this transition is particularly visible. According to the white paper, AI applications in medicine are developing along four main technological tracks: computer vision, speech processing, natural-language processing and data analytics. Computer vision is currently the most mature, with applications ranging from lesion detection and automated measurement to 3D reconstruction and anatomical segmentation. Natural-language processing is being used to structure medical records, support clinical decision-making and extract data for research. Data analytics is increasingly being applied to drug-target discovery, disease screening and outcome prediction.

The growth projections are striking. Oncology is expected to be the fastest-growing medical AI segment, with a projected CAGR of 71.5% through 2030. AI applications span radiology, endoscopy, pathology and dermatological imaging. Fracture diagnosis is projected to grow at 67.2%, while pneumonia diagnosis, stroke diagnosis, coronary artery disease and infectious-disease applications are also expected to expand rapidly.

Beyond software, the next frontier lies in Physical AI — systems capable of perceiving, making decisions and acting in the physical world — and AI for Science (AI4S), which applies AI to areas such as hypothesis generation, experimental design and scientific simulation. Frost & Sullivan projects China's Physical AI market to grow from $38.5 billion in 2025 to $337.5 billion by 2035.

Intelligent manufacturing: Robotics, UAVs

For 16 consecutive years, China has remained the world's largest manufacturing economy accounting for 30.2% of global manufacturing value-added. In 2025, China had 101 World Economic Forum-designated Lighthouse Factories, the highest number in the world. Smart-manufacturing equipment capacity exceeded 4.5 trillion yuan, while the CNC rate for critical processes reached 68.6%.

Robotics provides a clear example of the transformation. China produced 773,000 industrial robots last year, more than three times the 218,000 units produced in 2020. Its share of global industrial robot installations has risen to 54%, from roughly 30% a decade ago.

Graphic from Frost&Sullivan

Collaborative robots are emerging as one of the fastest-growing areas. Their applications are expanding beyond basic human-machine interaction into precision manufacturing and commercial services. Improvements in payload capacity, flexibility and intelligence are opening up new uses in footwear, fast-moving consumer goods, biopharmaceuticals and food processing. Designed to operate safely alongside human workers without protective fencing, they serve as a critical hardware foundation for flexible manufacturing.

Humanoid robots is expected to become another major growth market. The white paper projects China's humanoid robotics market to reach the trillion-yuan level by 2045. Adoption is likely to proceed in stages, beginning with industrial manufacturing and logistics, followed by commercial services, healthcare and rehabilitation, and eventually household and specialist applications. The next three to five years could therefore prove critical for turning humanoid robotics from a technological concept into a commercially scalable industry.

Graphic from Frost&Sullivan

The underlying technology is also evolving. Large AI models are increasingly being integrated with advanced robotic hardware to create embodied-intelligence systems capable of completing a perception-decision-action loop. World models and diffusion-based technologies could further improve their reasoning and generalisation capabilities.

Industrial manufacturing and services are likely to become the two main application engines. Electronic assembly and precision manufacturing are expected to lead industrial adoption, while healthcare rehabilitation, eldercare and retail logistics could drive growth in service applications.

Another emerging opportunity lies in China's low-altitude economy, covering drones, electric vertical take-off and landing (eVTOL) aircraft and related infrastructure.

Urban air mobility and drone logistics shall be reshaped, the White Paper notes. On the Shenzhen Shekou–Zhuhai Jiuzhou Port route, for example, an eVTOL journey could take about 15 minutes, compared with 70 minutes by ferry, three hours by bus and 1.5 hours by ride-hailing. Under the white paper's assumptions, an eVTOL service carrying 10% of 100,000 daily passengers could require about 250 aircraft, with an estimated investment payback period of 1.99 years. Lower acquisition and ticket costs could shorten that period to 1.78 years.