Huawei, Tencent and other cloud vendors have successively launched the DeepSeek large model

Author | Huang Yu

Editor | Zhou Zhiyu

During the Spring Festival of 2025, Chinese AI startup DeepSeek made waves in the global tech community with its open-source models DeepSeek-R1 and V3 series, achieving a technological breakthrough by matching OpenAI at "3% cost," marking the "DeepSeek moment."

In just one week, global cloud giants rushed to "land" and launched DeepSeek's large models to share in the spoils of this global tech storm.

On February 4th, Huawei Cloud announced that, after intensive efforts from the silicon-based flow and the Huawei Cloud team, they jointly launched the DeepSeekR1/V3 inference service based on Huawei Cloud's Ascend cloud services.

Recently, Tencent Cloud's TI platform also announced the availability of the DeepSeek series models, including the fully functional V3 and original R1 models, with a parameter count reaching 671 billion; as well as a series of models distilled from DeepSeek-R1, with parameter scales ranging from 70 billion to 1.5 billion.

It is reported that the TI platform fully supports one-click deployment of the DeepSeek series models. Additionally, to facilitate developers with a zero-threshold experience, the TI platform has temporarily opened free online access to the R1 model. The TI platform also provides capabilities for model service management, operational monitoring, and resource scaling, helping enterprises and developers efficiently and stably integrate DeepSeek models into their actual business.

In addition to Huawei Cloud and Tencent Cloud, ByteDance's Volcano Engine has also announced support for various sizes of DeepSeek open-source models, which can be used through deployment on the Volcano Engine machine learning platform veMLP and invocation in the Volcano Ark.

Baidu Intelligent Cloud, Alibaba Cloud, 360 Digital Security, and others have also participated in this "arms race."

Of course, it is not just domestic cloud service providers; DeepSeek, which is currently making waves in the global tech community, has been added to the "shopping cart" of more international tech giants.

It is reported that Microsoft, NVIDIA, Amazon, Intel, AMD, and others have also recently launched DeepSeek large model services.

Over the past year, the cloud service market has been filled with both opportunities and challenges, with a price war raging from the beginning to the end of the year. The immense market opportunities brought by AI large models are a battleground for competitors.

Li Qiang, Vice President of Tencent Group and President of Enterprise Business, stated that in the past two years, large model training has created a massive demand for GPU computing power. Although the growth in demand from large model training slowed last year, the increasing shift of enterprise users and startups towards large model applications has also generated significant demand on the inference side.

There are substantial opportunities brewing from the underlying training to the upper-level applications of large models. Although the revenue from AI large models for cloud vendors is still minimal at present, in the long run, it will become an important growth engine Therefore, behind the "collective launch" of the DeepSeek large model by cloud vendors is the desire not to miss the significant business opportunities brought by DeepSeek in an increasingly competitive market environment, as well as the important strategic layout made for the future.

By offering policies such as "zero-code deployment" and "limited-time free access," cloud vendors are essentially competing for the traffic entry point of future AI applications. Whoever can bind the most developers will gain an advantage in the next round of AI application explosion.

DeepSeek has already become another blockbuster AI native app following ChatGPT.

The official DeepSeek app will be launched on January 10, 2025, and then benefited from the high performance and low cost of the R1 model released on January 20, combined with the information dissemination during the Spring Festival, resulting in a dramatic increase in product attention.

From the day of the product launch, in terms of daily active users, DeepSeek surpassed ChatGPT on the 5th day, reached 2.59 million daily active users on the 15th day, which is twice that of ChatGPT, making it the fastest-growing AI native application globally. On the 18th day, it reached 15 million daily active users, while ChatGPT only reached 15 million daily active users on its 244th day.

On January 27, the DeepSeek app topped the free app download rankings in the Apple App Store in both China and the United States.

In addition, the inference model DeepSeek-R1 was released as open source, and in tasks such as mathematics, coding, and natural language reasoning, its performance is comparable to that of OpenAI's official version o1. Meanwhile, through algorithm iteration and architecture upgrades, DeepSeek has reduced the costs of its general and inference models to less than one-tenth of similar models from OpenAI.

DeepSeek has successfully rewritten the global AI competitive landscape.

Ying Ying, the chief analyst of computers at CITIC Construction Investment, pointed out that DeepSeek has fully open-sourced the model weights, and the MIT License open-source agreement it follows is very permissive, allowing other developers to use the model for commercial purposes and perform model distillation, which has been praised by Facebook's chief AI scientist, Yang Likun, as "the victory of open-source models over closed-source models."

It is foreseeable that as the barriers to application development are lowered, breakthrough "killer" AI applications will emerge more quickly