Core Viewpoint - The report from CITIC Securities indicates that after a significant rise, US and Chinese tech stocks have experienced a pullback, leading to notable market divergence. The firm believes that the current phase is still early in AI application development, and even without achieving AGI, AI will unlock the data dividends accumulated over 30 years of the internet, enhancing efficiency and productivity. They project that global AI Capex could increase 5-7 times by 2030 from 2025 levels, with China's AI Capex potentially rising 7-9 times, suggesting that AI investment has not peaked yet [1][2][11]. Market Performance - AI-driven US stock market has seen a bull run, with the Nasdaq rising up to 140% since the end of 2022, and Nvidia's stock soaring up to 14 times. The Philadelphia Semiconductor Index (SOX) increased by 39% from early 2025 to November 27 [2]. - In China, AI assets have also appreciated, with the A-share market seeing light module and PCB stocks rise by 137% and 65% respectively from the beginning of the year to November 27. Notable stocks like Cambricon and Semiconductor Manufacturing International Corporation (SMIC) have seen increases of 119%, 58%, and 37% respectively from July 25 to November 27 [2]. AI Capex Projections - CITIC Securities anticipates that the combined Capex of the four major US cloud service providers (CSPs) will reach $406 billion by 2025, a 61% increase year-on-year. This Capex level is projected to exceed the net profits of these companies but remain below their operating cash flows, raising concerns about whether global Capex has peaked [3]. AI Token Consumption and Chip Market - The firm has developed a comprehensive framework for estimating AI token consumption and related metrics. By 2030, global token consumption is expected to increase by 100-340 times compared to 2025. Consequently, the scale of inference computing power is projected to grow by 65-220 times, with the AI chip market potentially expanding to over $1 trillion by 2030, which aligns with projections from major tech companies like Nvidia [6][7]. Domestic AI Chip Market - China's AI chip market is expected to grow from $35-40 billion in 2025 to 7-9 times that amount by 2030, outpacing global growth. The domestic AI chip localization rate is projected to rise from 30-40% in 2025 to 60-70% by 2030. This growth is driven by increased investment from Chinese internet companies and advancements in domestic AI capabilities [11]. Opportunities in the AI Supply Chain - The domestic AI chip sector still lags behind Nvidia, creating opportunities in cooling solutions, AI power supplies, and other related fields. Key areas of focus include: - Liquid cooling technology, which is becoming mainstream due to its efficiency in data centers, with a projected market size of approximately $85.8 billion by 2027 [14]. - AI power supplies, which are expected to see significant growth due to rising power density requirements, with market sizes projected at $58.5 billion and $95.7 billion for 2026 and 2027 respectively [14]. - High Voltage Direct Current (HVDC) solutions, which are anticipated to gain traction, with a market size expected to reach $43.15 billion by 2028 [14]. - Optical communication modules, driven by increasing demand for high-speed, low-power solutions, with a projected compound annual growth rate of 22% over the next five years [14].
中信证券:算力需求空间仍然较大 看好国产算力链龙头及核心供应链