Domestic Computing Power Infrastructure Driven by Strong Demand from Tech Giants, as Overseas Cloud Providers Raise Capex Forecasts

Stock News
Aug 17

Huaxi Securities has released a research report stating that domestic computing power and data centers are in a phase of triple resonance: demand driven by capital expenditures from major tech companies, supply released through the expansion of super nodes and domestic chips, and technological advancements through system-level innovation. In the short term, the focus is on the commercial progress of top-tier solutions and orders from major tech companies and operators, while the medium to long term outlook is highly certain with a growth trajectory centered on AI infrastructure, supplemented by optical network upgrades.

The key insights from Huaxi Securities are as follows: At this stage, the parameter scale of large models has exceeded 2 trillion, approaching 10 trillion levels. Single-chip computing power is constrained by chip manufacturing processes and the "memory wall" bottleneck, leading to a widening gap in computing power supply. The limited supply of high-end overseas GPUs is forcing a shift in China from "single-chip catching up" to "system-level innovation." Super nodes, through high-speed Scale-up interconnections, package dozens to thousands of chips into a unified computing unit, directly addressing the system bottlenecks of insufficient single-chip computing power and high interconnection latency. The value chain is no longer concentrated solely on the chips themselves, but is expanding into areas where domestic manufacturers have a comparative advantage, such as switching chips, high-speed backplanes/connectors, liquid cooling, high-voltage power supplies, and whole cabinets. At the same time, intelligent computing centers are evolving from "general-purpose IDCs" to a new infrastructure model featuring "high power density, liquid cooling, green electricity, and modularization." Super data centers in places like Ulanqab, which boasts approximately 90% green electricity utilization and a 90% modularization rate, are becoming core carriers for domestic computing power. Modular delivery significantly shortens construction cycles, further supporting the capital expenditure efficiency of major tech companies' expansions.

Recommended areas of focus and related beneficiaries include: domestic AI chips (Cambricon Technologies, Haiguang Information Technology, Moore Threads, Muxi, etc.); super node systems and servers (Inspur Electronic Information Industry, Dawning Information Industry, Unisplendour Corporation, Ruijie Networks, Centec Communications, etc.); liquid cooling and high-speed interconnects (Envicool, Gaolan Shares, Qiangrui Technology, Accelink Technologies, Huagong Tech, Huafeng Technology); and high-quality computing power leasing and IDC operators (Runze Technology, Dongyangguang, Atnet, Atlas, Aofei Data, etc.).

Risk warnings: Capacity constraints remain for advanced manufacturing processes and HBM; mismatch between inventory and delivery schedules; constraints on power supply and energy consumption; increased volatility in high-valuation stocks; and systemic risks.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10