The current technological landscape in China is undergoing a significant transformation, particularly in the telecommunications sector, where major players are embracing artificial intelligence (AI) advancements to enhance their servicesAt the forefront of this evolution is DeepSeek, a homegrown AI model that has gained notable traction within the country's primary telecommunications providers, including China Mobile, China Telecom, and China Unicom.

Recently, the Ministry of Industry and Information Technology of China announced on its official platform that during the Spring Festival 2025, the three major telecom operators will fully integrate DeepSeek's open-source large model into their communication servicesThis integration is expected to manifest across various scenarios and products, marking a pivotal shift in the landscape of AI applications in telecommunications.

DeepSeek-R1 has emerged as one of the leading AI language models available globally, showing performance metrics comparable to those offered by OpenAI, yet at a lower costSince its launch on January 20, 2025, this model has captured the attention of investors worldwide and led to its rapid adoption by multiple industry giants during the critical Spring Festival periodThe three telecom operators are also among the first to announce their support for this model.

China Telecom was the pioneer, declaring itself the first operator-level cloud platform in the country to provide a complete stack of DeepSeek model servicesSubsequently, China Mobile followed suit by announcing compatibility with all mainstream versions of DeepSeek, while China Unicom stated its commitment to pre-deploying DeepSeek-R1 across over 270 nationwide backbone cloud pools, establishing its presence in diverse product scenarios.

The telecommunications industry's reception of this integration has been overwhelmingly positiveAnalysts, such as those from Galaxy Securities, have noted that the introduction of DeepSeek can lower the cost of computing resources

Advertisements

This cost-effective deployment can foster a virtuous cycle of stronger capital investment and yielded returns in the telecom sector's AI applications, thus amplifying overall market share.

Among the key strategic goals of these operators, boosting cloud services has become paramountYang Guang, a leading telecom strategy analyst at Omdia, pointed out that over the past two years, telecommunications companies have shifted from being mere information network providers to more advanced digital service providers, substantially increasing investments in cloud computing as a baseline for future strategiesHe emphasized that with cloud computing becoming the backbone for AI models, integrated development is now a viable path for these telecom operators.

One of the primary attractions of DeepSeek for the telecom operators lies in its low operational cost and high performanceTelecom analyst Zhou Guijun noted that as an open-source model, DeepSeek has significantly reduced the costs of training and inference while flexible deployment options, like private clusters and public cloud integration, make it a versatile choiceBy adopting DeepSeek, telecom operators can minimize the high developmental costs typically associated with self-developed models, enabling rapid commercialization of AI services leveraging existing computational resources.

For instance, China Telecom's Tianyi Cloud has made strides in leveraging DeepSeek-R1, effectively streamlining deployment, inference, and fine-tuning by implementing a lightweight distillation model that meets a wide range of demands from complex decision-making to high-load interactive requirementsThis deployment enhances the capability of the cloud platform while also attracting a broader range of corporate clients.

Furthermore, the involvement of telecom operators with DeepSeek has greatly enriched the functionalities of their cloud platformsFor instance, after incorporating DeepSeek, China Mobile Cloud achieved intelligent asset categorization and efficient retrieval, significantly enhancing user experiences

Advertisements

Analysts predict that this pairing will accelerate growth in cloud services offered by telecom operators, especially in the enterprise sector.

As AI continues to merge with the telecommunications industry, it has become evident that establishing a robust ecosystem is crucialWith 5G commercialization and the expansion of AI applications, telecom providers are poised to evolve from traditional communication service providers into comprehensive intelligent ecosystem creators.

Despite the collaboration with DeepSeek, it is notable that all three operators have previously developed their proprietary AI models, such as China Mobile's Jiutian, China Unicom's Lianzhizhi, and China Telecom's XingchenThis prompts questions regarding their rapid shift towards DeepSeek shortly after its launch.

Yang Guang provided insights into this decision; he remarked that while proprietary models focus more on industry-specific scenarios, making direct comparisons with DeepSeek's capabilities inappropriate, the all-encompassing domestic generic models have so far lagged behind globally leading models in performanceThe introduction of DeepSeek-R1, which closely rivals the capabilities of top-tier models like GPT-o1 and does not compete with the telecom operators in cloud service arenas, naturally rendered it attractive.

Moreover, Yang noted that DeepSeek, being a domestically developed leading AI model, aligns with the operators' aspirations to create a fully self-reliant framework from foundational chips to model deploymentObservations from the operators indicate that they have not abandoned their in-house development despite integrating DeepSeekFor example, China Mobile's COCA computing platform supports all iterations of the DeepSeek model while retaining adaptability for proprietary algorithmsLikewise, China Unicom's Xingluo platform allows flexible deployment of multiple model specifications in both private and public cloud scenarios.

According to Zhou Guijun, this strategy can be seen as an attempt by the operators to blend their proprietary models with DeepSeek via an open platform, thus constructing a hybrid AI ecosystem that leverages the strengths of both generic capabilities and vertical applications

Advertisements

Advertisements

Advertisements