Exclusive: China Large Model Winning Project Monitoring and Insight Report (January 2025)

At the beginning of 2025, DeepSeek created a whirlwind of technological innovation led by Chinese big model manufacturers, adding new impetus to the popularization of big model technology in the personal and enterprise markets.

In 2024, China's large-model bidding and tendering market achieved rapid growth. As the value of intelligent transformation of smaller scenarios in the early stage gradually emerges, government and enterprise customers will inevitably further expand their exploration of the implementation of intelligence in multiple scenarios in 2025, driving the overall market size to continue to maintain rapid growth.

In 2025, Intelligent Hyperparameters will use monthly and quarterly important nodes as cycles, and based on data from public big model winning projects, comprehensively track and analyze the trends and progress of big model commercialization, providing valuable market insights for related companies and practitioners.

The "China Large Model Winning Project Monitoring and Insight Report" will analyze multiple dimensions such as overall market size, industry coverage, project types, representative manufacturers, representative projects, and geographical distribution to depict the commercialization progress of the large model industry.

Key insights:

1. The large-scale model bidding market has a very good momentum at the beginning of the year. In January 2025, the number of large-scale model-related winning projects was counted to be 125, and the project disclosure amount reached 1.267 billion yuan, ranking second in the data of the past 13 months.

2. Application, application, application. In January, application projects still accounted for the largest proportion of data. In terms of specific application scenarios, digital humans & customer service, knowledge questions & answers & knowledge platforms, and intelligent agents are the top three scenarios. Intelligent agents appear more and more frequently in the names of large model projects. At the same time, the number of large model projects has exceeded that of computing power projects for the fourth consecutive month.

3. Five major industries dominate the large model bidding market. In January, based on the number of large model bidding projects, communications, education and science, government affairs, energy, and finance are still the top five industries, which is consistent with the data for the whole year of 2024.

4. There are more market opportunities and large companies have more competitive advantages. In 2025, as many places replicate the "investment for projects" model, large companies may have more competitive advantages in super-large projects.

5. Non-first-tier economic regions are more likely to have super-large-scale projects. Overall, economically developed regions have more active large-scale model bidding activities, but some super-large-scale model projects are more likely to appear in non-first-tier economic regions.

The following is the full text of the January 2025 monitoring report:

1. Overall market

In January 2025, the number of large-model related winning projects counted was 125, of which 31 projects did not disclose the amount (for the convenience of statistics, the amount was calculated as 0), and the remaining projects disclosed an amount of 1.267 billion yuan, ranking second highest in the data of the past 13 months.

From the above figure, we can see that although the monthly data in January 2025 is not as good as the monthly peak data in December 2024, considering the rules of the bidding market, it can at least be said that the large-model bidding market has a very good momentum at the beginning of the year.

In January 2025, the disclosed amount of large model winning projects continued to remain high, mainly due to the emergence of three projects with amounts of hundreds of millions of yuan. Among them: Baidu Smart Cloud won the Shenyang Data Circulation Application Infrastructure Project (Section 1) with an amount of 364 million yuan; Jiangsu Yungongchang Information Technology Co., Ltd. won the Shenyang Data Circulation Application Infrastructure Project (Section 2) with an amount of 265 million yuan; Fujian Shida Group won the national-level Fuzhou New District Artificial Intelligence Scenario Application Project with an amount of 211 million yuan.

In 2025, we expect that the intelligent computing center projects across the country will be closely related to big models, or AI big model platforms will become the standard configuration of intelligent computing centers. Because compared with a simple computing power center, the combination of computing power + big models is easier for local government and enterprise customers to actually use.

In January 2025, there were 12 large orders with a project value of more than 10 million yuan, accounting for about 86% of the total amount of these large-value projects. Judging from the data of the past few months, the number of large-value projects has been steadily increasing.

In January 2025, the median disclosed amount of the winning projects was RMB 1.387 million, higher than the median of the projects in 2024 (RMB 1.25 million). As the application scenarios of large models become wider and more complex, the further increase in the median project amount will be a very certain trend.

2. Distribution of large model project types

Currently, there are many types of bidding projects related to big models, but overall we divide them into four categories: computing power (computing resources required to run big models), data (big model-related data annotation or data resources, etc.), big models (various vertical big models, AI middle platforms or big model bases, which may include computing power), and applications (the integration and implementation of big models in clear scenarios, which may include computing power, AI big model platforms, application software, etc.).

The project categories counted in January 2025 are as follows:

In January, application projects still accounted for the largest proportion of data. In terms of specific application scenarios, digital humans & customer service, knowledge questions & answers & knowledge platforms, and intelligent agents are the top three scenarios. Intelligent agents appear more and more frequently in the names of large model projects.

The number of large model projects still exceeds that of computing power projects. According to our statistics, this situation has lasted for 4 months, which continues to support our previous judgment: large models are accelerating their implementation in more and more vertical industries, or are beginning to play the role of middle-end infrastructure in many medium and large enterprises.

The amount of computing power projects accounted for a relatively large proportion in January, accounting for about 77%. This is related to the new development trend of the intelligent computing center mentioned above.

3. Industry distribution of large model projects

Among the large-scale model winning projects counted in January, the prominent industries include: communications, education and science, government affairs, energy, and finance. The top five industries remained stable, consistent with the data for the whole year of 2024.

In January, the communications industry ranked first in the number of large-scale model winning projects. Local branches or subsidiaries of China Telecom, China Mobile and other companies continued to carry out a large number of bidding projects.

The number of large-scale model projects in the education, science, energy, and government affairs industries is similar, but the government affairs industry is far ahead in terms of the amount of projects. Super-large projects in the government affairs industry should continue to appear frequently in 2025, and may become the most eye-catching industry in the new year.

The number of projects in the financial industry was not outstanding in January, and the number also declined month-on-month compared with the monthly average data in the fourth quarter of 2024. However, it is worth noting that among the winning projects in the financial industry, large model and application projects accounted for a very large proportion, and the application scenarios were relatively diversified.

4. Competition among manufacturers

In the monthly and quarterly reports, Intelligent Hyperparameters, as usual, focuses on tracking the performance of six well-known domestic general large-scale model manufacturers.

In January, due to the limited scale of the projects, the number of winning projects for manufacturers was not large. Among them, Baidu and iFLYTEK were still the two leading large model manufacturers in terms of the number of winning projects, followed by Alibaba Cloud, Volcano Engine, Zhipu and other manufacturers.

Since the fourth quarter of last year, Alibaba Cloud, Volcano Engine, Tencent Cloud and other vendors have made significant efforts. Entering 2025, as many places replicate the "investment in exchange for projects" model, large companies may have more competitive advantages in super-large projects.

However, judging from the plans announced in many places, many first- and second-tier local governments are also working hard to cultivate the big model-related innovation ecosystem in their own regions, so new big model companies may also have some opportunities in the regional market.

5. Regional distribution

In January, the winning projects were spread across 26 provinces and municipalities in China, of which 59 projects were in the first-tier economic regions (Beijing, Guangdong + Shenzhen, Shanghai), accounting for about 47%, followed by Zhejiang, Fujian, Shandong, Jiangsu and other places.

Judging from the monthly data from 2024 to the present, the coverage area of ​​large-model winning projects has been steadily increasing. Overall, economically developed regions have more active large-model bidding activities.

However, some large-scale model projects with extremely large amounts of money are more likely to appear in non-first-tier economic regions. For example, the extremely large projects in January came from Liaoning, Fujian, Gansu and other places.

Note: Some of the winning bids in the report did not disclose the winning bid amounts. For the convenience of statistics, the project amounts are calculated as 0. There must be cases where projects are missed. The project classification and industry classification are not strict, which may affect the final statistical analysis results. Please be aware of this. For reference only.