China Galaxy Securities: The Value of AI Applications Continues to Be Realized, Domestic Supernodes Expected to Welcome a Volume Turning Point

date
15:53 07/09/2026
avatar
GMT Eight
The bank expects that the delivery of domestically-produced supernodes will accelerate in the second half of the year, and upstream parts of the supply chain such as chips, servers, power supplies, liquid cooling, and connectors are expected to continue benefiting.
China Galaxy Securities has released a research report stating that the artificial intelligence sector experienced slight fluctuations and minor gains in August, with significant divergence within sub-sectors. The bank believes that as we enter the mid-year report disclosure window, there is severe differentiation among sub-sectors within the AI field. There has been some switching between high-end and low-end segments within computing hardware, with certain players gradually transitioning to AI application sub-sectors. Currently, the market is more focused on existing stock competition, and further observation of order status, performance, cash flow, and other conditions for individual AI stocks is needed, suggesting a potential move towards distinguishing genuine opportunities from those that are not. The main points from China Galaxy Securities are as follows: The first national-level special policy, primarily focusing on AI service providers, has been issued, and AI applications are expected to usher in a value reassessment turning point. On August 31, the Ministry of Industry and Information Technology issued a notice regarding the implementation of a special action to cultivate AI application service providers. The policy outlines 1) Establishing a resource pool for service providers: The goal is to exceed 2,000 service providers nationwide by the end of 2026 and no less than 3,000 by the end of 2027, forming a multi-tiered team; 2) Enhancing supply levels: Organizing service providers to lead the establishment of AI application service groups for model resonance and computing-power coordination, collaborating with upstream and downstream units to create integrated solutions. 3) Promoting large-scale applications: Exploring models such as first purchase and first use, risk compensation, etc., to increase procurement of large models, intelligent agents, and tokens, thus improving the quality and efficiency of industry applications; 4) Strengthening support and guarantees for service providers: Supporting providers in enhancing connections with national computing power network hubs, national interoperability nodes, and Chinese computing power platforms, making use of policy tools such as computing power vouchers to reduce computing power costs, and encouraging service providers to build frontline deployment engineer (FDE) teams to engage with users directly and ensure scene implementation. The bank believes this is the first national-level special policy focused on AI application service providers in China, marking a shift in industry focus from model competition to implementation and delivery, with an emphasis on bridging the service layer between large models and industries, thereby facilitating the final step in AI application delivery. AI applications are expected to arrive at a value reassessment turning point. Domestic super nodes are moving from concept validation to system delivery stage, with a focus on investment opportunities in the industrial chain. At the WAIC 2026, Huaweis Atlas 950 Super PoD was showcased for the first time. This product is based on a single cabinet with 64 cards and can be scaled up to 8192 NPUs; in addition, companies such as Alibaba, Dawning Information Industry, H3C, Baidu, ZTE Corporation, and MuXi have launched their own super node products. The bank believes that internet companies are gradually moving from product validation to bulk procurement stages and remain the main demand source for domestic super nodes. The bank expects accelerated deliveries of domestic super nodes in the second half of the year, benefiting upstream segments of the industrial chain such as chips, servers, power supplies, liquid cooling, and connectors. NVIDIAs earnings guidance exceeds expectations, with strong growth in AI infrastructure demand continuing. On the evening of the 26th (Eastern Time), NVIDIA announced that its revenue for the second quarter of fiscal year 2027 reached $96.221 billion, an increase of 106% year-on-year and 18% quarter-on-quarter, with net profit of $59.69 billion, up 126% year-on-year and 2% quarter-on-quarter. Additionally, NVIDIA provided guidance for approximately 70% revenue growth in fiscal year 2028, and if there are no constraints on the supply chain and capacity, growth may accelerate; the bank believes NVIDIAs optimistic guidance indicates that the global AI industry is still in an upturn period, reflecting strong downstream demand and suggesting that CSP companies' capital expenditures are likely to continue rising. Upstream supply chains such as wafer fabs, optical modules, storage, PCBs, semiconductor equipment, liquid cooling, and power supplies are expected to benefit first, while midstream computing power leasing and cloud vendors are expected to gain short-term benefits in the context of supply shortages, with AI applications, agents, and enterprise AI applications likely to benefit long-term. Investment Recommendations It is recommended to pay attention to the following sub-sectors and companies: 1. Domestic chip design manufacturers: Cambricon, Hygon Information Technology, Suiruan Technology, BIREN TECH, ILUVATAR COREX, Moocube Technology, Loongson Technology Corporation; 2. Super node layout manufacturers: Dawning Information Industry, Inspur Electronic Information Industry, ZTE Corporation, Unisplendour Corporation; 3. Server ODM manufacturers: Digital China Group, Huaqin Co., Ltd., Talkweb Information System; 4. Domestic computing power industry chain supporting: Ruijie Networks, Kehua Data Co., Ltd., SUGON DATA ENERGY, Sichuan Huafeng Technology, and Shengke Communication; 5. Computing power leasing and token factories: iSoftStone Information Technology, Range Intelligent Computing Technology Group, Beijing Sinnet Technology, RunJian Co., Ltd., Dawei Technology (Guangdong) Group, Guangdong Hec Technology Holding, Xingyun Technology; 6. Domestic large model manufacturers: Z.AI, MINIMAX-W; 7. Xinchang and AI application manufacturers: Beijing Kingsoft Office Software, Inc., Fujian Fuxin Software Development Joint Stock, Intsig Information, China National Software & Service, Iflytek Co., Ltd., Wuhan Dameng Database, SDIC Intelligence Information Technology, Wondershare Technology Group, Servyou Software Group, Hithink RoyalFlush Information Network, Shenzhen Fortune Trend Technology, MOBVISTA, MARKETINGFORCE, State Power Rixin Technology, and others; 8. Industrial software and physical AI: Zhejiang Supcon Technology Co., Ltd., 51WORLD, MOMENTA-W, Shanghai Suochen Information Technology, ZWSOFT, Digiwin Co., Ltd., Empyrean Technology, Primarius Technologies, and Semitronix Corporation. Risk Warning: Risks of technological iteration falling short of expectations; risks from intensified competition among technology giants; regulatory risks; supply chain risks; risks of downstream demand falling short of expectations.