DeepSeek brings immediate returns to cloud computing manufacturers.

"Basically, the cloud computing power inventory of major cloud manufacturers, including operators, has been cleared." A researcher from a large model research institution told reporters. This institution came to this conclusion after investigating domestic cloud manufacturers.

On January 20, after the release of the DeepSeek-R1 model, various cloud computing manufacturers launched its full-blooded or distilled version models. It is open source and relatively efficient, which further increases the penetration rate of artificial intelligence in China and promotes cloud computing manufacturers to increase investment in infrastructure.

DeepSeek is now the focus of competition among various cloud computing manufacturers.

"DeepSeek is the best touchstone. Which cloud hosting service has good performance, large throughput and low latency, you can tell directly." Tan Zhiwei, president of Volcano Engine, said in his circle of friends.

The models released by DeepSeek include multiple versions, including a full-blooded version with 671B parameters and a small model with 1.5B, 7B, and 70B parameters. Different parameters, different computing resources are required, and different terminals are deployed. For example, the 1.5B model can be deployed on mobile phones worth several thousand yuan, while the 70B model can be deployed on workstations worth tens of thousands yuan, while the 671B full-blood version model is more cost-effective in the cloud.

Volcano Engine is a cloud and AI service provider under ByteDance. In addition to supporting its own big bean bag model, it has also launched several series of models such as the Dark Side of the Moon and the Smart Score. The Spring Festival has just ended, and the Volcano Engine has launched a full-size DeepSeek model. The DeepSeek model is open source and easy to use. If enterprise-level users want to use a full-blooded version of the big model in the cloud, they need to purchase computing power or storage services from cloud computing manufacturers.

Tan Tie’s statement is a voice to corporate customers and is also aimed at other cloud computing manufacturers.

The competition among cloud computing manufacturers around DeepSeek started during the Spring Festival. In order to launch DeepSeek as soon as possible, many engineering and technical personnel have spent an endless Spring Festival holiday. The competition among cloud computing manufacturers is already fierce, and the price war is one after another. Therefore, Tencent Cloud is worried when connecting to Alibaba's big model. Conversely, Alibaba Cloud will not fully accept Tencent's Hunyuan model, and Baidu Smart Cloud will not favor Tencent or Alibaba's big models.

Open-source DeepSeek looks harmless and beneficial to cloud computing manufacturers, so it has gained wider acceptance. China's technology companies have released hundreds of big models, but none of them is as penetrating as DeepSeek and can jump lightly on every cloud.

DeepSeek has also driven the wave of open source.

Baidu was once a staunch supporter of the closed-source route, but it also announced that it will be fully opened on June 30 after the New YearYuanqiwenxin Mockup 4.5 Series. Baidu Smart Cloud has also connected to the DeepSeek big model and attracts developers to use it at a lower price.

At the 2025 Step Jump Star Ecological Open Day, Step Jump Star, one of the big model Six Little Tigers, announced that it will open source its picture-generated video model in March this year. Previously, Jieyuexingchen has released 11 multimodal models, two of which are open source products.

"Skyu has planned a roadmap for realizing AGI since its establishment. Specifically, there are three stages: simulating the world, exploring the world and summarizing the world." Jiang Daxin, founder and CEO of Stepyue Star, said: "The work of Stepyue is carried out along these three stages. In the past year, we released a series of models, forming a comprehensive model matrix from 100 billion parameters to 100 billion parameters, from language models to multimodal models, from multimodal understanding to multimodal generation, from imitation learning to reinforcement learning, covering multiple key nodes in this roadmap."

Once the model is open source, it has its own vitality, and the open source big model ecosystem can further stimulate the enthusiasm of developers to use.

China's big model industry is undergoing a structural transformation. Some companies such as Zero One Wanwu have abandoned the high-cost general big model research and development and turned to vertical industrial models with lower investment costs and faster commercial effects.

After the DeepSeek-R1 model triggered market enthusiasm, the effectiveness of the "Jevins Paradox" was verified. DeepSeek points to lower computing power demand, but the improvement in computing power usage efficiency has enabled the inference model to quickly penetrate into thousands of industries, thus driving the overall growth of computing power demand.

"The model is updated too quickly, and the best model must be on the cloud." Tan said in an interview at the end of 2024.

OPPO mobile phone has been connected to the big model of Step Stars, and Xiaohongshu has been connected to DeepSeek to enhance its search function. Volcano Engine said it has assisted 60 financial institutions including Guosen Securities, Ping An Securities, E Fund Fund to access the DeepSeek big model.

"The release of the inference model has led to a massive influx of players, and the cloud computing power has been eaten up." The above-mentioned researcher told reporters.

Compared with leading companies such as Alibaba Cloud and Tencent Cloud, cloud computing companies at the lower end and below need such opportunities to change the decline in revenue.

Youkede's revenue reached 2.9 billion yuan in 2021, but its revenue has declined continuously in the following two years, with only 1.5 billion yuan in 2023. In the 2024 interim report, UCID's revenue also fell to 730 million yuan, and it suffered a loss of 100 million yuan. Qingyun Technology also had a sluggish revenue after revenue of 429 million yuan in 2020, with revenue of 330 million yuan in 2023.

In this round of revaluation of China's technology assets, UCID and Qingyun Technology's market value have doubled since the beginning of 2025, but now they also need to tangible performance double.

DeepSeek is inspiring a round of investment in cloud and AI infrastructure among major Chinese technology companies.

Alibaba Group CEO Wu Yongming announced on February 24 that he will invest more than 380 billion yuan in the next three years to be used for cloud and AI hardware infrastructure. This investment exceeds Alibaba Group's total investment in the past decade.

On February 25, Baidu announced that it would acquire YY Live, the domestic video entertainment live broadcast business of Huya Group for about US$2.1 billion. After the new agreement was reached, the approximately US$1.6 billion previously deposited into the escrow account has been returned to Baidu in full. Baidu said it will invest US$1.6 billion in cloud and AI infrastructure construction.

In January this year, the first phase of the Suzhou Fenhu Intelligent Computing Center of Mobile Cloud was put into operation, with a total investment of 2.68 billion yuan in the first phase. The total planned investment of Mobile Cloud's Suzhou project is 6 billion yuan. After all, it can carry 24,000 cabinets and 500,000 server installation capacity, and finally the computing power can reach 10EFLOPS.

"Volcano Engine's computing power reserves are definitely very sufficient, which is also one of the important reasons why we are confident of taking on industry services at large scale, low prices and high throughput." Tan Zhi told reporters in an interview at the end of 2024.

Compared with Alibaba Cloud, Baidu Cloud and other companies, Volcano Engine is a latecomer in the cloud computing market. "If you want to do it faster, you must be the first in new technological changes. What is the new technological change? In fact, it is AI and big models, so our goal is the first in this area, and we are moving in this direction." Tan Tai said.

Domestic computing power is expected to take this opportunity to embed deeper into the artificial intelligence ecosystem.

"DeepSeek's popularity has a strong demand for our computing power." Mobile Cloud told First Financial News. Mobile Cloud will make some fine-tuning and distillation of DeepSeek's model to meet the needs of companies in specific industries. Mobile cloud provides a nationwide deployment environment. This can especially impress some central state-owned enterprises, whether it is local or cloud deployment.

According to Mobile Cloud, the computing power it provides largely meets the needs of central state-owned enterprises. In February 2025, Mobile Cloud assisted China Petroleum in completing the training, adaptation and privatization deployment of DeepSeek full-stack domestic production, which is inseparable from Ascend's computing power support.

Nvidia's GPU meets more than 90% of the needs of cloud computing companies, but the emergence of DeepSeek provides the possibility for domestic computing power. Mu Xi, Moore Thread or Huawei Ascend GPUs have been accepted by more and more cloud computing platforms or terminal companies.

Chinese chip companies also seized this opportunity and took advantage of the situation.

Models in DeepSeek SeriesShortly after the launch of the model, Kunlun Chip completed the full version of the adaptation, and announced on the day the start of the Year of the Snake that it had lit up its new generation of product P8 million card cluster; on February 20, Kunlun Chip officially announced that it had achieved a full-blooded version of the stand-alone deployment on the P800.

Since DeepSeek launched the "Open Source Week", it has opened sourced four code libraries one after another. Moore threads have now completed the adaptation of DeepSeek open source libraries FlashMLA and DeepGEMM.

The United States has imposed restrictions on China's import of high-end chips, which forces Chinese companies to achieve technological breakthroughs. At present, Nvidia GPUs have advantages in computing power performance and ecological maturity, but with the advancement of domestic chip research and development, the improvement of algorithm capabilities, and the industry's emphasis on supply chain security, the country has the hope to gradually reduce the risk of single dependence on Nvidia.

"This process requires time and technology accumulation, but it is an irreversible trend. The future computing power base is more likely to show a 'multi-dimensional coexistence' form rather than the absolute dominance of a certain manufacturer." Moore Thread previously told reporters.