HomeArticle

A review of 196 large DeepSeek orders in half a year: These 5 provinces have the most hot money.

智东西2025-07-07 19:32
196 government-enterprise major deals, with 1.1 billion yuan in funds pouring in! The half - year development map of DeepSeek since its launch.

According to incomplete statistics from Zhidx, from January 1st to June 30th, 2025, there were at least 196 DeepSeek government and enterprise projects in China with a budget exceeding 100,000 yuan, with a cumulative project amount of over 1.1 billion yuan.

In terms of statistical targets, Zhidx mainly counted projects with "DeepSeek" in the project title and projects with procurement content descriptions strongly oriented towards DeepSeek. The survey found that although there are many DeepSeek government and enterprise project data, most of the budget amounts are not large, mainly small and medium-sized projects within 2 million yuan. As shown in the figure below, among the 196 projects, 80 had an amount of 100,000 - 1 million yuan, 55 had an amount of 1.01 - 2 million yuan, and 19 had an amount of 2.01 - 3 million yuan.

▲Distribution of DeepSeek project amounts in the first half of 2025 (compiled by Zhidx)

Among them, the project with the largest amount is the National Biosafety Monitoring and Early Warning Information Platform project with a budget of 450 million yuan. The tenderer is Shaanxi Chang'an Pioneer Life Science Industry Innovation Center Co., Ltd., and the winning bidder has not been determined yet; the project with the largest amount among those that have completed the bidding is the Group DeepSeek Intelligent Computing Cluster Construction Project of COSCO Shipping Technology Co., Ltd. with a budget of 16.94 million yuan, and the winning bidder is China Telecom.

Most of the 196 projects are integrated projects combining software and hardware. There are almost no pure software projects. Most projects propose the deployment of DeepSeek all-in-one machines or large model computing power clusters. Some projects also propose to combine DeepSeek with embodied intelligent robots.

It is worth mentioning that in some DeepSeek projects, the purchaser also mentioned that the Qwen large model could be used. In addition, the Kimi large model was also mentioned in individual projects.

The distribution of winning enterprises is relatively scattered. Operators such as China Mobile and China Telecom have won relatively more projects. Apart from them, there is no enterprise that has won multiple DeepSeek projects. Well-known manufacturers such as Alibaba Cloud, Tencent Cloud, Baidu, Volcengine, Inspur, and JD.com have all appeared on the lists of bidders and winning candidates, staging many "fights" among cloud giants and battles between cloud giants and AI unicorns.

What are the characteristics of these projects in terms of time, space, and industry distribution? What trends do they represent for the implementation of large models in the industry? Through a comprehensive review of these 196 DeepSeek projects, Zhidx analyzed and discussed these issues.