DeepSeek’s Popularity May Mean Clear Skies Ahead for China’s Cloud Compute Industry, as U.S. Catchup Is Imminent

Huawei’s cloud computing unit has partnered with Chinese Artificial Intelligence (AI) hardware firm SiliconFlow to offer DeepSeek on Huawei’s Ascend platform. DeepSeek is the open-source Chinese alternative to Western leaders’ offerings of OpenAI’s GPT and Google’s Gemini, having generated buzz in the claim that it uses fewer advanced chips to train the model and is thus a cheaper substitute to its competitors—a realization that greatly affected NVIDIA’s market performance. Although it is a tech leader in its own right, China’s success comes as the United States—the global hub for AI and cloud—has limited the export of AI semiconductors into the country due to fears of their misappropriation for military use, thereby hindering innovation in the sector. This has forced Chinese tech firms to become more creative in their endeavors, and the nation’s cloud computing services are benefitting from this innovation. As a foundation upon which AI may be executed, cloud adoption and growth in the country will only grow as AI gains further momentum—giving China an advantage over the United States if DeepSeek continues to generate interest at its current rate.

China’s cloud computing industry, which has benefitted from US$6.12 billion in government investment for data center building in recent years, is only on the rise, largely driven by AI initiatives.

This growth is signified by the popularity cloud firms have garnered as a result of their AI-related releases and acquisitions. Several weeks ago, Alibaba Cloud released a new generation of its proprietary Tongyi Qianwen Large Language Models (LLMs), which the firm says are used by 90,000 corporate clients in China, undoubtedly driving business to its platform. Chinese state-owned semiconductor and cloud firm Tsinghua Unigroup, which has also acquired a stake from HPE in its locally based full-stack cloud AI platform, H3C, is also looking to raise US$1 billion on the Hong Kong Stock Exchange as it saw its shares grow by 45% over the past year post-acquisition.

Hyperscalers have also been gifted a new client magnet with DeepSeek—on top of Huawei, Tencent Cloud integrated its R1 model into its Hyper Application Inventor service, which allows developers to deploy it within 3 minutes. Chinese firms are not the only ones doing this. Microsoft launched R1 support on its Azure cloud platform, shortly followed by Amazon Web Services (AWS), and surprisingly, NVIDIA with its NVIDIA Inference Microservices (NIM) tool and container platform. Further, DeepSeek, and even Alibaba’s Qwen 2.5-Max model have shown that AI can be more affordable, compounded by the fact that the latter firm slashed the prices of Generative Artificial Intelligence (Gen AI) products by 85% at the end of 2024—giving Chinese firms the upper hand. With the United States’ Graphics Processing Unit (GPU) embargo on the country a mere setback now, it wouldn’t be surprising to see one of these leading Chinese hyperscalers announcing their own GPU solutions at MWC.

The Biden-Harris Administration in early January announced that it would require U.S.-based data center firms to keep at least 50% of their controlled advanced chips within the country or a short list of allied countries—prohibiting more than 7% of installations occurring in any one country. Although this move is largely aimed at limiting the potency of AI harnessing by markets, this will ultimately limit the reach that hyperscalers will be able to have in other regions, depleting market share for other countries to exploit, as this limits the capacity of U.S. firms to penetrate new markets if their local base remains stagnant.  

In any case, large emerging economies in Asia, such as Thailand, Malaysia, and Indonesia, are likely targets for China’s cloud business expansion. These countries already see a larger share of cloud computing availability zones from China-based businesses over U.S. hyperscalers. It is abundantly clear that the Chinese government is also appealing to this group, as it has vowed in its AI Capacity-Building Action Plan that it would support the Global South through AI-led initiatives—likely including the provision of data center infrastructure.

Overall, China is domestically successful in cloud initiatives, helped by DeepSeek and AI integration, as well as synergies with local players across the ecosystem.  However, there are blind spots to be filled by Chinese players that can be improved upon.

Pushing the AI agenda has helped Chinese firms innovate cloud computing platforms as a whole, but they cannot rest on their laurels, especially considering the rate of innovation in supporting both infrastructure and software globally in this tech arms race.

• Target the Right New Markets: The aforementioned sanctions list imposed by the U.S. government very broadly highlights the regions in which the country has a stronghold—namely the select nations the United States has allowed its chips to be sent to. The list of countries, including much of Southeast Asia and Middle Eastern countries such as the United Arab Emirates (UAE) and Saudi Arabia, which may or may not already contain U.S. hyperscalers as incumbents, should be targeted over other regions. Offering unlimited access to AI and hyperscaler resources not only strengthens China’s position in the global tech industry, increasing trust for its local companies, but it may also become the most fruitful endeavor when choosing new markets, given the perceived openness and cooperation of the country compared to the United States. If Chinese hyperscalers like Huawei were also able to significantly expand and improve upon their GPU offerings in the future, this would only make their cloud propositions stronger.
• Bundle and Simplify: When it comes to entering a new market, especially a developing one, price and simplicity matter. Indeed, AI is a prime value proposition to spearhead cloud initiatives, especially the AI-in-a-box offerings that Huawei has pushed for. However, AI is not the only cloud-compatible technology—this model should be replicated with similar solutions such as the Internet of Things (IoT). Even further, these offerings could be segmented by vertical or use case to better encourage adoption. In any case, creating these solutions, especially when made up of components from fellow Chinese vendors, will allow for cheaper and easier adoption of cloud services, with less need to convince firms of business value.
• Target Essential Local Tech Ecosystems: The strength of operating in the cloud market is that most big firms will require these services to remain competitive, especially considering the rapidly accelerating rate of data generation. Hyperscalers can harness this by targeting essential firms in the local market for multi-year contracts. This can be more traditional government and defense-related contracts, or perhaps consumer goods and services as the digital economy grows in especially developed countries. For example, last year, Alibaba signed a 5-year cloud and AI deal with GoTo, the leading e-commerce and last-mile delivery company in Indonesia. It also signed a contract with Tencent Cloud, which will use the firm’s Platform-as-a-Service offering to support its end-to-end application services. The latter hyperscaler also invested US$500 million in Indonesia to build its third data center in the region.