CITIC Securities: Focus on investment opportunities in internet and AI applications with the efforts of large models and edge AI

According to the Zhitong Finance APP, CITIC Securities released a research report stating that in observing global technology market investments in 2025, from a market perspective, Chinese technology assets are considered to have better investment cost-effectiveness compared to U.S. assets. Within the Chinese technology sector, the top choice is the Chinese concept internet sector, urging investors to pay attention to the performance turning points brought by short-term macro recovery and policy stimulus, as well as the valuation reshaping opportunities brought by the sustained prosperity of the AI ecosystem in the medium to long term. The report also sees promising investment opportunities in China's domestic AI industry chain.

CITIC Securities' main viewpoints are as follows:

Market Review: Since 2024, the U.S. stock technology sector has led the global market, while technology stocks in the A-share and Hong Kong markets have outperformed the broader market, but the Hang Seng Technology Index valuation has returned to the level before 924.

U.S. Stocks: In 2024, U.S. technology stocks continued to lead globally, with the Nasdaq rising 34% year-to-date. By sector, the S&P North America Software Index rose 19% in Q3 2024, while the Philadelphia Semiconductor Index rose 5% during the same period, with software significantly outperforming. Over a longer period, from early 2023 to the end of 2024, the S&P North America Software Index increased by 109.9%, aligning closely with the Philadelphia Semiconductor Index's increase of 110.2% during the same period. The firm believes that the recent performance of the U.S. technology sector reflects the market's shifting interest in AI towards application levels. In November, AI service provider Palantir rose 61%, and AI advertising service platform Applovin rose 99%, with companies that have significant potential for AI application and rapid growth entering a phase of dual expansion in performance and valuation. Additionally, the performance of the U.S. technology sector is also spreading to small and mid-cap companies; in September 2024, the market capitalization growth of the top ten U.S. technology stocks accounted for 77% of the Nasdaq's year-to-date market capitalization increase, which decreased to 71% by December 2024.

Hong Kong Stocks: The Hang Seng Technology Index increased by 17% in 2024, with the majority of the increase driven by the market performance in September. In terms of structure, Pop Mart, Xiaomi, and Meituan had the highest increases for the year, with respective increases of 371%/95%/91%. After October, the Hang Seng Technology Index corrected, and as of December 20, its valuation level had returned to the level before 924, positioned at the ~10th percentile since 2021.

Domestic: In 2024, the A-share technology sector outperformed the broader market, with the Shanghai Composite Index rising 14%, and the electronic/communication/computer/media/automobile indices rising 17%/28%/17%/19%/19%, respectively. Since September, the technology sector's overall rebound has been stronger than the broader market, with the computer sector, which had previously low performance growth expectations, showing significant elasticity in this round of market performance. In September, the Shanghai Composite Index rose 17%, while the electronic/communication/computer/media/automobile indices rose 19%/24%/35%/26%/23%, respectively. For the year, the communication industry benefited from the overseas AI development driving sub-industries such as optical modules and AIDC, leading to a strong overall increase. The two main themes are AI and domestic self-reliance, with significant increases seen in overseas AI leading industry chain companies and domestic AI self-reliance supporting companies such as computing chips and data platforms. From the perspective of active allocation ratios of public funds, as of Q3 2024, the overall allocation ratio of the electronic sector has remained at historically high levels, while the allocation ratio of the communication industry has continued to expand throughout the year Progress in the Artificial Intelligence Industry: Reinforcement Learning Becomes a New Path for Model Performance Improvement, Software and Edge AI Implementation Accelerates, China's AI Industry Expected to Develop Independently.

Large Models: Competitive Landscape Clarified; Growth Rate of Model Parameters Slows Down. Reinforcement learning has become a new path, constrained by the scale of computing clusters and the supply of high-quality data. Since the release of Claude 3.5 in mid-June, the release speed of large parameter models has lagged behind market expectations. By December 2024, with OpenAI releasing the o3 model, the reinforcement learning paradigm will become a new method to continue enhancing model performance and improving mathematical and logical capabilities. OpenAI has improved the model's accuracy and its ability to answer complex mathematical questions by extending reasoning time and applying chain-of-thought reasoning methods. Domestic manufacturers are also accelerating the penetration of large models in the B-end and C-end by reducing model costs. The cost of ByteDance's large model continues to decline, with C-end DAU rapidly increasing at a compound monthly growth rate of over 15%. The token usage in the B-end has also exceeded market expectations, rising rapidly from 120 billion times in May to 4 trillion times in November.

AI Software Applications: At the application level, American tech companies continue to promote the accelerated implementation of large model technology.

1. The software function release speed of large model companies has accelerated, adapting to various ecosystems and application environments. Google launched the Gemini 2.0 Flash version, Anthropic released Compute Use; OpenAI launched the ChatGPT iOS version and Canvas.

2. B-end SaaS companies have rapidly advanced their understanding and technical level regarding Agents. Companies like Workday and Salesforce are accelerating the launch of Agent products and beginning to explore the monetization of AI Agents, with company valuation levels also rising in tandem.

3. Project-based AI companies and AI advertising platforms are leading in implementation speed. In FY24Q3, AI service provider Palantir achieved revenue of $726 million, a year-on-year increase of 30%, significantly exceeding previous guidance, mainly due to sustained investment in government AI projects. AI advertising platform company Applovin also exceeded expectations in FY24Q3 revenue and profits, enhancing the efficiency of its advertising platform through AI technology. Applovin's penetration rate among e-commerce clients has rapidly increased, and the disruption of business models in vertical industries by AI is evident. Chinese companies are also actively following up in related directions, and with the rapid iterative engineering capabilities of domestic tech companies, the independent construction of the AI ecosystem is expected to accelerate further.

Edge AI: In autonomous driving, Tesla continues to iterate on its autonomous driving algorithms, with FSD V13 expected to be pushed out soon. Tesla's Robotaxi plan is set to begin mass production in 2026. In the field of humanoid robots, with the gradual maturity of the robot components supply chain and the introduction of large models, NVIDIA, Tesla, and Figure are continuously iterating on the commercialization progress of robots, with multiple potential catalysts expected in 2025 Looking back at the domestic market, several technology companies have launched AI glasses and AI headphones, and the AIoT market continues to ferment, with blockbuster products emerging one after another. Technology companies such as Xiaomi and Li Auto are also actively exploring the application of AI in autonomous driving and smart cockpit fields.

Investment Logic: Looking ahead to 2025, the bank is optimistic about the performance of the Chinese technology sector driven by AI.

Chinese Technology: The Hang Seng Tech Index valuation level is at the 10% low percentile since 2021, and the foreign ownership ratio is also at a historical low. The strengthening of buybacks continues to enhance the EPS of internet companies, providing a strong margin of safety, while positive changes in macro policies will bring a double boost to performance and valuation for the sector, providing significant upside potential. Cyclical sectors such as e-commerce, local life services, transportation, freight, online recruitment, and real estate service platforms will benefit significantly, and targets with a solid pattern and high performance certainty will enjoy valuation premiums. In the medium to long term, the domestic AI industry chain represented by ByteDance's Doubao AI ecosystem continues to iterate, with multimodal capabilities and accuracy continuously rising, and ongoing penetration in both B-end and C-end markets. Chinese tech companies like Tencent and Alibaba also have rich vertical data and application scenarios. By 2025, the landing of more AI software applications or edge products will strongly boost the valuation levels of Chinese tech companies.

US Tech: Looking ahead to 2025, the bank believes that the core focus of global tech investment will still be on AI, urging investors to pay close attention to investment opportunities in the software SaaS sector. For the software SaaS sector, the bank judges that many negative factors that previously suppressed the software sector are gradually easing or turning into positive catalysts, while IT spending by European and American companies is beginning to gradually recover after the elections. Agent-driven AI + software monetization is expected to accelerate in 2025. For the hardware sector, AI data centers are expected to continue to be a growth driver for the performance fundamentals in 2025, while the valuation levels of related companies are at a high level.

Risk Factors: Investors are advised to fully pay attention to the risks brought by changes in the external geopolitical environment and the intensification of trade frictions on technology industry investments. Potential tariffs and trade policies in the United States may hinder the export of domestic optical modules, lithium batteries, complete vehicles, and components to North America or weaken product competitiveness; changes in geopolitical situations may prevent IC design companies in mainland China from using advanced process capacity at overseas foundries for tape-out. The bank also warns of risks such as: the pace of macroeconomic recovery being slower than expected; relevant industrial policies not meeting expectations; the pace of Federal Reserve interest rate cuts being slower than expected; progress in core technology and product R&D being slower than expected; the speed of AI application landing being slower than expected; and cloud vendors' capital expenditures being slower than expected