Author | Zhu Chaoyang

Source | Lingyi Intelligence

The year 2024 has been a testament to the rapid evolution of large AI models.

According to the "China Large Model Development Index (2nd Edition)" published by Lingyi Intelligence, from August 2023 to October 2024, China's Large Model Index surged from a baseline of 100 points to 260.16 points, marking a 1.60-fold increase with a monthly average compound growth rate of 7.07%.

Figure 1: China Large Model Development Index (August 2023 - October 2024)

Data Source: Lingyi Intelligence

On the technical front, large models have witnessed continuous performance enhancements, significantly boosting operational efficiency and drastically reducing training costs. Underlying architectures, multimodal fusion, end-cloud coordination, open-source initiatives, and other technologies have been refined, making models capable of handling image, audio, and video processing increasingly prevalent.

In terms of capital, data indicates that in 2024, there were 168 financing events exceeding 100 million yuan related to the global large model industry chain, with a cumulative financing amount exceeding 400 billion yuan. Domestic AIGC enterprises raised 37 billion yuan, an over 80% year-on-year increase, with the top six large model enterprises collectively raising over 20 billion yuan.

Policymakers have also played a pivotal role, with central and local governments issuing a series of specific policies and plans to outline the development roadmap for the large model industry. These initiatives aim to promote the application of large models across multiple sectors, enhance infrastructure, and optimize the innovation environment by constructing intelligent computing centers and providing financial support.

The rapid technological iteration, substantial capital influx, and robust policy support have collectively fueled the proliferation of domestic large model application scenarios. Large models have achieved deep integration and innovative applications in education, finance, government affairs, healthcare, energy, transportation, meteorology, and other domains.

This trend is evident in bidding data, particularly winning bid data.

01

Explosive Growth in Applications: Winning Bids Surge 18-Fold in a Year

Based on incomplete statistics from Lingyi Intelligence, there were as many as 1,010 publicly disclosed winning bids for large model projects in 2024, totaling approximately 3.64 billion yuan. In contrast, 2023 saw only 54 corresponding projects with a combined winning bid amount of 390 million yuan.

In just one year, the number of winning bids for large models increased by nearly 18 times, and the total winning bid amount surged by 8.3 times, indicating a remarkable surge in large model applications.

It's worth noting that these statistics are relatively conservative, encompassing only bids with titles containing keywords like "large model" and "big data model" while excluding duplicates. General intelligent computing center projects not explicitly linked to large models are not included in the tally.

In terms of total winning bid amounts, China Mobile, iFLYTEK, and China Communications Services each surpassed 300 million yuan. Baidu, Kunpeng Cloud, Cloudtop Technology, and Henan Kunlun also exceeded 100 million yuan. Regarding the number of winning bids, China Telecom, iFLYTEK, China Mobile, Baidu, and Zhipuhuazhang each secured over 20, with China Unicom Digital Technology, Alibaba Cloud, and Inspur also surpassing 10.

iFLYTEK ranked second only to China Telecom in the number of winning bids, with a total winning bid amount of 470 million yuan, placing second behind China Mobile and exceeding the third place by nearly 150 million yuan. It also won two projects worth over 100 million yuan in Hefei and Dadong, Shenyang.

Notably, China Communications Services, a company owned by the three major telecommunications service providers, won 12 large model projects, including the Dongguan Artificial Intelligence Large Model Center Project (Phase I) worth 318 million yuan, with the rest being internal group demands.

Table 1: Top 30 Winning Bids for Large Models in 2024 (Ranked by Total Winning Bid Amount in Descending Order)

Note: Due to the numerous and complex disclosure channels for winning bid information, as well as incomplete disclosures, the statistics in the table are for reference only.

Data Source: Early Warning for Enterprises, QCC, and Lingyi Intelligence

Among internet giants, Baidu won a substantial number of bids totaling nearly 200 million yuan. Alibaba, ByteDance (Volcano Engine), and Tencent secured varying numbers and amounts of winning bids, but all exceeded 40 million yuan. Huawei won four large model projects with a total amount exceeding 60 million yuan.

Among the "Six Tigers" of large models, only Zhipuhuazhang and Baichuan Intelligence currently have winning bid data for large models. Zhipuhuazhang won a larger number of bids spanning various fields (including communications, education, finance, energy, automotive, etc.), with a total publicly disclosed winning bid amount of approximately 58.33 million yuan. Baichuan Intelligence won two projects, one of which was worth 26.8 million yuan.

02

Multi-field Implementation: Financial Large Models Top Winning Bid Rankings

Winning bid data reveals that the demand for large models stems from a diverse range of sources, with educational and research institutions, as well as communications service providers, demonstrating comparable demand, accounting for nearly 40% combined. The next largest sectors are finance and government affairs, both accounting for more than 5%. Other sectors with significant demand include healthcare, infrastructure, energy, and transportation.

Scientific research institutions, primarily universities, have shown heightened interest and considerable investment in large models, with as many as 192 related bids, accounting for over 19% of the total, and a total bid amount of 370 million yuan, exceeding 10% of the overall amount. Among them, Xi'an Jiaotong University, the Chinese Academy of Sciences, and Zhejiang University have the highest demand, while Beijing University of Aeronautics and Astronautics and Harbin Institute of Technology have the highest bid amounts.

The International Innovation College of Beihang University has two bidding demands for the large model computing power base project, totaling 70.19 million yuan. In the first phase of the AI industry-education integration innovation platform project of Harbin Institute of Technology, there are three bids directly related to large models, totaling approximately 28 million yuan. The cross-modal large model and embodied intelligent AI system project of Tongji University has a bid amount of 11.907 million yuan.

The three major telecommunications operators and their affiliated companies won 185 large model projects, accounting for 18.3% of the total, but the disclosed total winning bid amount was only 140 million yuan, accounting for 3.8% of the overall amount. In terms of individual bid amounts, only three reached the ten million yuan level.

Figure 2: Proportion of Winning Bids for Large Models in Various Fields in 2024

Data Source: Early Warning for Enterprises and Lingyi Intelligence

Figure 3: Proportion of Total Winning Bid Amounts for Large Models in Various Fields in 2024

Data Source: Early Warning for Enterprises and Lingyi Intelligence

Finance stands as one of the earliest and most widely applied fields of large models. According to a report by the Shanghai Industrial Research Institute of China Mobile, the penetration rate of large models in the financial sector has surpassed 50%, ranking first among all industries.

In terms of the number of winning bids, financial large models rank second only to education, research, and communications, with 63 projects, primarily involving banks and securities firms, as well as auto finance, trusts, futures, and other areas. TPG Fintech has the highest demand. Banks with significant demand include Bank of Beijing, Bank of Shanghai, and China Minsheng Banking Corp. Securities firms with notable demand include China Merchants Securities, Guotai Junan Securities, Haitong Securities, and Guosen Securities.

Institutions with more winning bids include Baidu, ECCOM Network (a subsidiary of CEC Digital), Volcano Engine, and iFLYTEK. Zhongke Jincai, a case company featured in the "China Large Model Development Index (2nd Edition)" report, won the bid for the software and hardware procurement project for the deep learning large model training and inference platform of China Minsheng Banking Corp. Ltd. Additionally, it secured a non-financial large model project, the 2024 large model knowledge base customization service project of China Telecom Co., Ltd. Xiangtan Branch.

In September 2024, the candidate winners for the procurement project of the intelligent extraction and writing platform based on large models for the Beijing Branch of Postal Savings Bank of China were publicly announced, with Zhongke Jincai ranking first. The platform aims to facilitate writing and extraction scenarios such as investment research, due diligence report generation, and meeting minutes, providing structured industry data for various report writings and citations, sourced from various open-source reports.

In terms of the total winning bid amount, the GPU computing power resource procurement project for the intelligent operation center of China Everbright Bank has the highest amount, exceeding 18 million yuan, with Nantian Electronics as the winner. The second is the Kunpeng chip server project for the procurement of information technology and innovation computing power for the 2024 large model application system construction project of Shanghai Pudong Development Bank, with a total bid amount of over 16 million yuan and Huaxun Network as the winner.

Furthermore, the three bids for the automotive finance large model platform of Changan Automobile Finance Co., Ltd. totaled over 8 million yuan. Bidders with winning bid amounts exceeding 5 million yuan also include Huichang Finance, WeBank, and Shenzhen Stock Exchange.

In the government affairs sector, the number and amount of winning bids for large models rank fourth, with more than 50 projects and an amount exceeding 280 million yuan. Among them, the Tianjin Artificial Intelligence Vertical Large Model Innovation Community Upgrading and Reconstruction Project has a bid amount of 110 million yuan. Projects from the All-China Federation of Trade Unions, Shenzhen Bao'an District, Nanjing Government Cloud, and Beijiang Public Safety Video Intelligent Mining and Large Model Application are all in the ten million yuan range.

A clear trend emerges: Local government-led infrastructure construction for large models has entered an intensive promotion stage, encompassing large model application bases, industrial clusters, intelligent computing centers, and training factories. There are at least 35 related winning bid projects, with a total winning bid amount of up to 1.41 billion yuan, accounting for approximately 39% of the overall amount. Among them, more than six projects exceed 100 million yuan, and eight projects reach the ten million yuan level.

Projects exceeding 100 million yuan include, in addition to the aforementioned Dongguan Artificial Intelligence Large Model and the two projects won by iFLYTEK, the East Data to West Computation Guian New Area Computing Power Industrial Cluster Supporting Project - Large Model Construction Project (Phase I), the large model intelligent computing server and RoCE switch framework of Beijing Zhiwang Digital Technology Co., Ltd., and the Chengdu-Chongqing Region Artificial Intelligence and Industry Large Model Fusion Application Center.

In the energy sector, including power grids, pipeline networks, and petrochemicals, there are also more than 30 winning bid projects for large models. Among them, the AI-enabled large model software and hardware framework project of the National Pipeline Network Digital Platform has a bid amount exceeding 150 million yuan, with Cloudtop Technology as the winner. In the healthcare sector, numerous winning bid projects can be observed, involving various scenarios such as impact analysis, auxiliary medical treatment, psychological assessment, report analysis, pre-consultation, and cloud accompaniment for diagnosis.

Additionally, there are over 30 winning bids for large models in the transportation sector (including automobiles, buses, railways, civil aviation, ports, docks, etc.), and winning bids in the meteorology, agriculture, and disaster prevention sectors are also noteworthy.

End.

This is the first article in our annual review of large models, offering an overview of the winning bid landscape for large models. Subsequent articles will delve deeper into specific industries. If you have any questions or other needs, please feel free to contact us!