Article | AI Relativity

After DeepSeek's meteoric rise, the public observed a sudden influx of technology vendors rushing to announce the integration of the DeepSeek series of models on their platforms, offering related AI services and capabilities.

This list includes global tech giants like Microsoft and Amazon Web Services, as well as prominent local cloud providers such as Alibaba Cloud, Huawei Cloud, and Baidu Intelligent Cloud. Even the three major telecommunications operators swiftly joined this "traffic surge" surrounding DeepSeek.

In response, skepticism emerged online: Clearly, DeepSeek is in the spotlight, so why are so many technology vendors involved, seemingly detracting from its exclusive "solo show" in AI?

In reality, the industry's development logic transcends the concept of a "one-man show." The truth is that DeepSeek desperately needs technology vendors to "ride the wave." Moreover, only when more vendors or platforms participate in this "traffic surge" can the AI industry achieve broader popularity and robustness.

No "One-Man Show" in AI

DeepSeek's rise did not solely involve technology vendors "unilaterally" capitalizing on its popularity and gaining traffic. The widespread adoption of AI itself necessitates support from various platforms and vendors. Reflecting on the actions of major platforms today, we see that technology vendors or platforms in the market have contributed three critical values to the popularization of DeepSeek.

1. Computing Power: The robust computing, storage, and network resources of cloud platforms provide essential support for the stable operation of DeepSeek.

Users who have tried DeepSeek can attest to the frustration of encountering the "server busy" prompt, ultimately forcing them to turn to other platforms. Although DeepSeek's training costs have significantly decreased compared to the industry, the computational power required for inference often increases with the user base. Simultaneously, DeepSeek faces insufficient server resources, unable to stably and efficiently meet the needs of a vast user base, thereby hindering the popularization and promotion of AI services.

Cloud platforms, leveraging their established computing resources, can better support the operation and promotion of DeepSeek. Currently, the clouds of the three major telecommunications operators, relying on powerful computing, storage, and network resources, such as China Telecom Tianyi Cloud's "Xirang" intelligent computing platform, China Unicom Cloud's "Xingluo" platform, and China Mobile Cloud's COCA native computing power platform, are providing the large-scale computational power necessary for AI model training and inference of DeepSeek, thereby accelerating its adoption across various industries.

2. Scenarios: The scenario access of digital and intelligent service platforms directly expands the market service scope of DeepSeek, accelerating the maturity and promotion of its technology.

Currently, DeepSeek's popularity is rooted in technological innovations and breakthroughs, requiring further commercial verification and implementation to evolve into a mature product. On this basis, the integration of digital and intelligent service platforms becomes even more valuable. Relying on the platform's extensive customer base and diverse scenario services, it can accelerate the commercialization process of DeepSeek.

Aishu, for instance, has launched a data management platform based on the Data+AI architecture of DeepSeek, enabling the integration of business and AI through intelligent agents called Data Agents with business semantics. This platform can develop more business scenarios according to needs to implement AI applications, such as knowledge management, intelligent operation and maintenance, data search and utilization, among others.

This is highly beneficial for DeepSeek. On one hand, Aishu serves over 27,000 customers worldwide, helping DeepSeek swiftly reach a global customer base. On the other hand, with over 20 years of deep involvement in the data field, Aishu boasts strong full-domain data capabilities, capable of integrating multiple data types and providing a high-quality data foundation for AI applications through data engineering and knowledge engineering, truly realizing "DATA for AI".

3. Processes: Technology vendors offer a convenient deployment environment and standardized interfaces, simplifying the deployment and operation processes of DeepSeek, enabling faster and more direct access for enterprises and customers.

While DeepSeek is an exceptional large model, for many industry customers, deploying and using it remains a challenge. In this regard, the deployment environment and standardized interfaces provided by technology vendors significantly alleviate the issue of users having to repeatedly learn, allowing enterprise customers to directly deploy and use DeepSeek in a familiar or simplified environment.

For example, Alibaba Cloud's PAI Model Gallery platform supports one-click deployment of DeepSeek on the cloud, enabling users to complete the entire deployment process without coding. Simultaneously, Alibaba Cloud's Bailian platform provides related model APIs that can be directly called by enterprises. For enterprises already familiar with the Alibaba Cloud platform and processes, this undoubtedly saves the cost of relearning and deployment. Moreover, as customers of cloud vendors, to accelerate the promotion of AI applications, the platform also offers various services and technical consultations, better assisting users in leveraging DeepSeek.

Looking back, the entry of technology vendors is indeed a positive development for DeepSeek. While DeepSeek takes the initial step with its groundbreaking technological innovation, the remaining 99 steps and the "last mile" problem are eagerly addressed by technology vendors, encompassing issues like computing power, scenarios, and processes mentioned above.

Will DeepSeek Impact the Existing Market Landscape?

It is evident that as everyone scrambles to integrate DeepSeek, the capabilities of this large model have effectively become a versatile tool embedded in various platforms.

Consequently, the impact on the existing market landscape will not be substantial.

For enterprises, while DeepSeek's capabilities are impressive, its subsequent applications have yet to be fully explored and verified, insufficient to shake the foundation of enterprise digital and intelligent transformation.

Simply put, the rich data accumulated on past platforms and the familiar cloud infrastructure of enterprises remain crucial foundations for digital and intelligent transformation, rather than AI capabilities alone. Platforms like Alibaba Cloud, Huawei Cloud, Tencent Cloud, Tianyi Cloud, etc., even if they do not integrate DeepSeek, will not lose customers for this reason. Conversely, DeepSeek is more attractive to new customers who have not yet adopted cloud-based digital and intelligent transformation.

Currently, these cloud vendors seem to have sparked a new price war, with services like China Telecom Tianyi Cloud, Baidu Intelligent Cloud, Alibaba Cloud, and Wuwenxinqiong all announcing limited-time free trials.

Among them, Tianyi Cloud's "Xirang" intelligent computing platform offers limited-time free inference API services for the DeepSeek-R1/V3 series of models. Baidu offers a two-week free trial with no limit on the number of tokens used. Alibaba Cloud offers a free trial of 10 million tokens, among other incentives.

That being said, the advent of DeepSeek is also surprising, particularly for the domestic chip market, whose future development landscape may be altered at this juncture.

As we all know, domestic chips still lag behind NVIDIA's high-end chips in terms of single-card computing power and need to bridge the gap through software and hardware co-optimization. Moreover, the stability and accuracy of model inference require further verification, with some users reporting a decline in output quality after domestic chip adaptation. These are the shortcomings of domestic chips, and it is challenging to compensate for hardware deficiencies in the short term. However, DeepSeek presents an optimization path on the software side.

Currently, over a dozen domestic AI chip vendors, including Ascend, Intellifusion, and Bitmain, have announced support for the deployment and training of DeepSeek models, covering the entire scenario chain. The process of localization and adaptation is already accelerating, which will more swiftly promote the technical accumulation of domestic chips in areas like software optimization and memory management, fostering collaboration among upstream and downstream enterprises to gradually improve the domestic chip ecosystem.

Perhaps, when this path is cleared, the market landscape of domestic chips will be rewritten once again.

Conclusion

With its innovative algorithm, DeepSeek achieves high performance at low cost, significantly lowering the threshold for AI applications. As this threshold drops, numerous vendors flock in. As cooperation deepens, market share continues to expand, serving as an important witness to the accelerated rise of the entire local AI industry. Today's "traffic surge" fueled by DeepSeek and "fanned" by numerous technology vendors is destined to become a pivotal node in the history of China's AI development.

*All images in this article are sourced from the internet