Recent releases of nine cutting-edge models, including GPT-5.5 and DeepSeek-V4, have intensified the global artificial intelligence wave, with China's large model industrial landscape demonstrating high concentration in core regions and emerging innovation hubs. In the recently released "2026 Top 100 Most Valuable Chinese Token Suppliers" list by Deben Consulting, 17 enterprises from Guangdong province were ranked, placing second nationally after Beijing. Previously, the "Top 50 Chinese Cities in Large Model Industrial Competitiveness" list issued by China Economic Information Service showed Guangdong leading with seven cities included. These two rankings reflect Guangdong's substantial capabilities in the next phase of AI development from both corporate and urban perspectives.
As a major manufacturing province, Guangdong views AI as a critical factor for industrial transformation and upgrading. On April 22, Guangdong launched its "Action Plan for Promoting High-Level AI Application Across All Domains, Times and Industries," outlining 63 specific areas across seven key directions to integrate AI into social life. From computing power保障 to scenario openness, pilot base construction, and comprehensive application promotion, Guangdong is establishing a value chain from technology to industry, deeply empowering various sectors.
A high-density large model industrial belt is emerging in Guangdong. Among the seven ranked cities, Shenzhen, Guangzhou, Dongguan, and Zhuhai entered the top 20. Shenzhen, with a comprehensive score exceeding 90, leads the first tier. The city maintains领先地位 in large model R&D and innovation, driven by tech giants like Tencent and Huawei. Tencent's intelligent assistant "Yuanbao," based on its Hunyuan model, ranks among China's top AI applications, while Huawei's "Pangu" model solves core challenges in industries like coal mining, steel, and meteorology. Shenzhen aims to achieve 1 trillion yuan in AI terminal industry scale by 2026, launching over 50 popular AI terminal products.
Guangzhou focuses on original innovation and scenario incubation, nurturing 19 AI research institutions and 12 industrial parks. Twenty-four enterprises from the city were listed as global unicorns, including several in AI, while 12 enterprises accounted for nearly one-third of the national disruptive technology innovation stars list.
Dongguan, Foshan, and Zhuhai serve as secondary engines, all ranking within the top 20. These cities share advanced manufacturing bases and are undergoing transformation from manufacturing to smart manufacturing. Dongguan, for instance, promotes "AI+" integration with advanced manufacturing through its national AI application pilot base. Using a "challenge-based" mechanism, the city connects enterprises, developers, and investors, recently releasing 128 scenario demands across 241 companies and 10 industries to bridge the gap between technological breakthroughs and industrial application.
NVIDIA founder Jensen Huang noted that the Guangdong-Hong Kong-Macao Greater Bay Area uniquely concentrates both mechatronics and AI technologies. Overall, Guangdong's AI industry has formed a complete梯队. In 2025, the province's core AI industry scale exceeded 300 billion yuan, growing over 40% annually and representing about one-quarter of the national total. 132 large models passed national备案, ranking third nationally, with over 430 vertical industry models deployed. Guangdong leads with three national pilot bases in consumption, healthcare, and energy sectors.
The data foundation remains robust. In 2025, Guangdong's data center operational rack scale reached 1.42 million, ranking second nationally, with intelligent computing scale exceeding 70 exaflops. Over 4,000 data enterprises operate in the province, with Pengcheng Laboratory pioneering the release of a 1.1TB high-quality Chinese corpus. Thirteen 400G all-optical networks connect Shaoguan clusters with the Greater Bay Area, enabling low-latency computing power coverage. Shenzhen has established a 14,000 petaflop fully autonomous intelligent computing cluster, providing solid computing foundation for domestic large models.
The large model industry is transitioning from technological competition to application-oriented deployment. Guangdong's manufacturing sector, encompassing all 31 major categories and accounting for about one-eighth of national scale, provides an ideal testing ground with its 10 trillion-yuan level industrial clusters including new-generation electronic information, green petrochemicals, and smart home appliances.
AI transformation is becoming routine in Guangdong's production lines. At Honor's AI smart factory in Shenzhen, 85% of processes are automated, with over 60% of equipment self-developed, producing a smartphone every 28.5 seconds. At GAC Aion's smart eco-factory, a global lighthouse facility for new energy vehicles, over 600 robots collaborate to roll out a new car every 53 seconds.
Guuangdong hosts over 1,600 core AI enterprises, forming full-stack industrial chains represented by Huawei's Ascend ecosystem and Tencent's Hunyuan model. Numerous industrial large models and intelligent agents undergo iterative upgrades, bringing new engines for manufacturing transformation.
In the token supplier ranking, 17 Guangdong enterprises including Huawei, Tencent, Industrial Fulian, X-Chuang Data, Aofei Data, Hongjing Technology, Desay SV, and CloudMinds appear, mapping a comprehensive token economy spanning chips, servers, data centers, smart driving, edge computing, and super applications. Compared to fundamental model development, token suppliers are closer to application, indicating the region's AI industry has progressed from model creation to service provision.
This full-stack computing system transcends technical parameters. If the large model industry represents a gold rush, Guangdong possesses not only manufacturing scenario "gold mines" but also the capability to produce "shovels." Market-oriented and specialized in segmented scenarios, Guangdong maintains solid industrial foundations and diverse applications, positioning its AI technology at the global forefront.
The April 22 action plan marks a policy shift from cultivating supply to activating demand. Guangdong's 15th Five-Year Plan proposes developing "AI+robotics" into high-tech, high-growth, large-scale industrial clusters, forging additional trillion-yuan and hundred-billion-yuan clusters. Under this framework, AI serves not as an independent industry but as an enabler for all sectors, similar to electricity's transformative role.
Guangdong also plans to incorporate Hong Kong and Macao, collaborating to strengthen computing power supply and build 100,000-card level intelligent computing clusters, aiming to exceed 1,000 exaflops by 2030. This requires more than 14-fold growth in intelligent computing capacity within four years.
Challenges remain, including relative scarcity of top talent and need for deeper AI integration with application scenarios. The key risk involves how many applications can form sustainable commercial闭环 after policy incentives and capital enthusiasm subside, and how many enterprises can transition from experimentation to routine use. As industry competition evolves from model development to token supply, Guangdong has demonstrated a trajectory from conceptual emergence to practical implementation, potentially setting a benchmark for China's intelligent economy.
Comments