UBS: China's accelerated computing development promotes AI progress, optimistic about Alibaba and Baidu

According to the Zhitong Finance APP, China is accelerating its investment in the field of artificial intelligence (AI). UBS published a research report stating that despite uncertainties surrounding imported AI chips, it believes that with national policy support and the push from major technology companies or local suppliers for R&D investment, domestic computing power is still continuously developing. This may continue to drive the development of AI and large models in China. The bank is optimistic about Alibaba (09988) and Baidu (09888), as it believes that progress in self-developed chips will continue to consolidate their positions in the AI value chain, and they will also continue to invest in AI.

UBS mentioned that recent technological advantages include: 1) With ongoing investments in R&D by Chinese internet companies and local GPU suppliers, although there are still gaps at the chip level, improvements are rapidly being made; 2) At the system level, gaps are being bridged through supernode scaling, such as Alibaba's Panjiu 128 supernode and Huawei's Ascend 384 supernode, significantly increasing the number of GPUs per rack, partially compensating for the performance gap of individual domestic GPUs, achieving higher rack-level computing power. The bank believes this design allows domestic chips to support more complex inference scenarios. In the long term, as network technology improves to achieve node expansion to large clusters, it may even support training workloads. 3) AI model developers are optimizing algorithms for domestic GPUs. DeepSeek's latest v3.2 model uses the domestic GPU programming language TileLang, which can better adapt to the local algorithm ecosystem, such as Huawei's Ascend and Cambricon (688256.SH).

UBS stated that most internet companies are accelerating the development of ASICs (Application-Specific Integrated Circuits) to optimize internal workloads and improve cost-effectiveness. Google is one of the earliest companies to develop its own AI chips, having iterated through multiple generations, while Amazon, META, and Microsoft are also developing their own customized AI chips. In China, Baidu has developed three generations of Kunlun chips, and Alibaba has also begun deploying self-developed chips.

After a recent survey of AI chip experts, UBS summarized three key points. 1) Hardware performance: Currently, the computing power of domestic cutting-edge GPUs has matched that of NVIDIA's (NVDA) Ampere, with next-generation products also targeting Hopper, but overall still lagging behind the Blackwell series. 2) Software ecosystem: Some domestic chip manufacturers have established their own software stacks or added CUDA (Compute Unified Device Architecture) compatibility through translation tools, thereby improving engineers' migration efficiency. However, the fragmentation of the ecosystem limits scalability. 3) Supply chain capabilities: Besides chip design quality, China's capabilities in advanced process technology and high-bandwidth memory production are still in the early stages.

In addition to Alibaba and Baidu, UBS is also optimistic about iFlytek (002230.SZ) due to its unique positioning, having made leading progress in combining domestic hardware with large model development. Additionally, the bank prefers Horizon Robotics (09660), Northern Huachuang (002371.SZ), and Zhongwei Company (688012.SH)