This week has seen the Hong Kong stock market maintain its upward momentum, officially entering a technical bull market, albeit the trajectory of technology stocks has shifted from a consistent rise to a more volatile ascendance. This change leaves investors—both domestic and international—questioning the sustainability of valuation expansions driven by innovations like DeepSeek.
As the week progresses, major internet giants in China prepare to unveil their earnings, with Baidu’s report on February 18th falling significantly short of expectations, resulting in a dramatic drop in their stock prices. Meanwhile, Alibaba, which had rebounded nearly 50% from its lows last year, is set to release its report on February 20th at 8:30 PM, followed by Tencent's earnings on March 18th.
A significant factor contributing to the reevaluation of the market's tech industry valuations rests on the horizon of cloud computing services. According to the Director of Greater China Equities at UBS Wealth Management, Li Zhiying, cost-effective models are expected to accelerate the adoption of AI technologies, thereby increasing enterprise demand for AI-related services. However, it should be noted that income from major language models and AI-related cloud services currently represents only a small single-digit percentage of the overall revenue for these firms. This means that, in the prevailing economic slowdown, the upcoming earnings season will be a critical indicator of revenue and profit trends.
The landscape surrounding cloud services offers a promising outlook for valuation expansion. The groundbreaking innovations introduced by DeepSeek have ignited a new wave of investments, applications, and commercialization in artificial intelligence, thereby prompting many international investment banks to reassess their ratings of the Chinese stock market.
For instance, HSBC recently upgraded its outlook for Chinese stocks from neutral to bullish, citing the emerging self-innovation in AI technologies and commercial opportunities resulting from them. Morgan Stanley noted that investors are beginning to realize that the development of AI in China may not necessarily require massive capital investment. They believe that even with constraints on high-end GPU supplies, China has the capacity to narrow or even close the technology gap. This improved growth outlook, along with potential confidence boosts, is expected to drive up the fair value of Chinese stocks by 15% to 20%, potentially leading to over $200 billion in investment inflows.
The Hang Seng Tech Index has rebounded nearly 25% from its lows this year, with prominent firms such as Alibaba, Tencent, and Kingsoft Cloud notably drawing attention. Kingsoft Cloud, in particular, has surged an astounding sevenfold since last November, with cloud services being the common theme among these firms.
According to Thomas Chong, head of internet and media for Jefferies in the Asia-Pacific region, there is a significant low valuation on China's internet sector, suggesting that the cloud valuations of the so-called BATs (Baidu, Alibaba, Tencent) will undergo a reevaluation based on their respective scales and technological capabilities. He highlighted that the advances in large language models (LLMs) would lead to this reevaluation, differentiating this industry surge from the broader macroeconomic recovery story expected in 2024. Chong anticipates that the development of AI agents and applications will accelerate due to the pervasive integration of AI across industries, improvements in user insights and feedback, and elevated conversion rates. Consequently, we can expect to see more subscription-based models focused on improved conversion and AI agents being launched in the near future.
The importance of cloud services in relation to large models cannot be overstated, especially concerning deep learning models that require significant compute resources during the training and inference stages. These models involve vast amounts of data and numerous parameters, necessitating powerful computing capabilities and effective distributed computing architectures. Cloud service providers like Alibaba Cloud, AWS, and Azure are uniquely positioned to furnish these resources while allowing for flexible scalability based on demand.
For American tech giants, cloud services often serve as a key driver of performance acceleration. For instance, AWS represented approximately 17% of Amazon's total revenue in 2024; Google Cloud accounted for 12% of Alphabet's revenue, while Microsoft's Intelligent Cloud segment made up a staggering 43% of Microsoft's overall sales.
Chinese internet behemoths still have ample room for improvement in this sector. Data reveals that Alibaba Cloud, Huawei Cloud, and Tencent Cloud command close to 60% of China's cloud computing market share collectively. Alibaba Cloud, in particular, is a crucial revenue stream for Alibaba Group; during the third quarter of 2024, it accounted for 12.5% of Alibaba’s total revenues.
Moreover, leading cloud platforms including Baidu, Huawei, and Tencent have announced support for DeepSeek R1, offering competitive pricing structures. For example, Baidu is pricing the DeepSeek R1 API at 2 RMB and 8 RMB per million input/output tokens, respectively, and is currently providing a free version until February 19, 2025. The impact of these developments on fiscal performance for cloud services is yet to be determined.
The rising hype surrounding DeepSeek has resulted in skyrocketing stock prices for major internet firms, yet whether this translates into measurable profits remains uncertain. Therefore, the implication of these dynamics on future performance guidance is crucial. Chong remarks that overseas investors are increasingly expressing focus on Alibaba's valuation, especially in light of macroeconomic stability in China. The recent share price movements of Alibaba and Tencent reflect significant aspects of their respective cloud business valuations, with future reassessment hinged on management’s commentary regarding outlook and strategy; immediate corrections in revenue and profit are not anticipated. The entertainment sector is viewed as more defensively positioned, featuring greater visibility.
Jefferies maintains an optimistic stance on the outlook of Alibaba Cloud. According to the latest rankings of open-source models from Huggingface, Alibaba's Qwen model occupies a notable position within the top ten models, with at least seven models based on Qwen2.5-72b. Additionally, Alibaba Cloud's Qwen2.5-MaxMoE model, which was pre-trained using a dataset exceeding 20 trillion tokens, has outperformed DeepSeek-V3, Llama-3.1-405B, and GPT-4o across several benchmarks, providing a pleasant surprise to the market. Its global cloud infrastructure is robust, with the Tongyi model family catering to the needs of users and enterprises by offering various foundational models and applications. Consequently, Jefferies has raised its target stock prices for Alibaba’s American and Hong Kong shares to $150 and HKD 145, respectively—an adjustment from $144 and HKD 139—and based on a projected fourfold increase in cloud business sales multiples, assigning a valuation of approximately $60 billion to Alibaba Cloud, up from $45 billion.
Turning to Tencent, the recent integration of DeepSeek has led to a noticeable reevaluation of its market valuation. Why is this development pivotal? Jefferies posits that Tencent's significant investments in its mixed-precision model, coupled with its decision to incorporate third-party AI models into its flagship product, marks a shift towards a more open, user-centric AI strategy. This indicates Tencent's focus on functionality and user experience over exclusivity. Although such a move may not yield immediate profit enhancements, it fortifies WeChat's essential role in the daily lives of its 1.3 billion users, creates an intangible asset moat, and lays a foundation for sustainable long-term monetization through higher engagement rates.
Moreover, UBS is currently placing greater emphasis on the execution of performance amidst these developments. Li Zhiying has noted that leading companies might accelerate the integration of DeepSeek models into their respective products, which will benefit various operational firms. For instance, advertising technology and recommendation systems may be strengthened, enhancing returns on investment (ROI) and increasing user engagement, ultimately benefiting numerous internet companies. Nonetheless, the growing demand may heighten competition among participants, and the business volume growth might potentially be offset by declining prices of API tokens, serving as the minimal unit for text processing.
Most importantly, Li Zhiying emphasized that while the long-term impact of AI adoption and applications driven by large language models in China could yield substantial benefits, the current contribution of cloud services to these internet giants remains limited. Currently, revenue from large language models and AI-related cloud services makes up only a small single-digit percentage of their overall revenues.
In conclusion, the fundamental drivers of China’s internet platforms still hinge on core businesses, namely e-commerce, gaming, and advertising. Given the backdrop of an economic slowdown, the upcoming earnings season will serve as an essential barometer for revenue and profit trends. Given the strong correlation between the performance of these platforms and the broader Chinese macroeconomic landscape, it is imperative to stay vigilant regarding U.S. trade policy as well as China's responses.