CITIC Securities: The valuations of leading AI companies attract long-term optimism about the growth space of China's technology assets

The Zhitong Finance App learned that CITIC Securities released a research report saying that the appeal of AI assets in China is gradually showing, and leading AI companies are expected to receive performance-driven alpha investment opportunities. The valuation of leading AI companies still has a clear comparative advantage over comparable US companies. Once US technology stocks show double pressure on performance and valuation caused by the slowdown in CAPEX growth, the attractiveness of Chinese assets is expected to become even more apparent. The bank said it has long been optimistic about China's technology and the room for revaluation and growth of a wider range of assets, especially after the 2018 Sino-US game put significant pressure on valuations. In the future, every major technological breakthrough and application implementation will have an opportunity to bring about a systematic increase in valuation. The focus is on investment opportunities in the fields of large domestic models and intelligent clouds, Chinese AIDC, domestic computing power processes, intelligent driving, AI agents, and embodied intelligence.

CITIC Securities's main views are as follows:

The origin of the report: DeepSeek has attracted great attention. Investors are concerned about how far can the revaluation of China's technology assets go? DeepSeek triggers a “butterfly effect” with excellent performance, significantly reduced costs, and an open source ecosystem.

It only took 7 days for DeepSeek to have over 100 million registered users, breaking the 2-month ChatGPT record. Currently, DeepSeek has surpassed 30 million daily active users, becoming the most popular AI model in the world, triggering a revaluation of AI assets in China.

Reviewing the revaluation of US technology companies brought about by the launch of ChatGPT: Since 2023, the NASDAQ has risen 100%, the stock price of the leading company Nvidia has risen 8.5 times, and the market capitalization has advanced to the top three in the world. Application-side companies such as Palantir and Applovin have risen to prominence, rising 17.6 times and 47.4 times during the same period, and the valuation gap between Chinese and US technology stocks has widened. Since January, the stock prices of leading companies such as Alibaba, SMIC (688981.SH), BYD (002594.SZ), and Xiaomi Group have risen by 51%, 43%, 37%, and 30% respectively. IDC and cloud infrastructure companies have also seen significant increases, and the valuation gap between Chinese and US technology companies has begun to narrow. DeepSeek is working alone to change expectations of “America's AI cliff-style leadership”, thereby changing global investors' judgment on the value of China's technology assets. The breakthrough of this startup, founded in July 2023 with only 100 employees, is a microcosm of China's scientific and technological innovation power in the context of the undefeated game between China and the US in 2018. The bank believes that it is still in the early stages of the AI era and the beginning of a revaluation of China's technology assets.

Industry impact: DeepSeek brings AI equality, changes the global technology landscape, increases demand for computing power, and ignites expectations for an explosion of applications.

DeepSeek is the first to reproduce GPT-O1 under conditions where computing power is limited. The performance of mathematics, code, and natural language is comparable to o1, and is an independent open source RL-scaling technology. The v3 version uses innovative architectures such as MoE and MLA, and changes in training strategies to achieve low-cost and efficient training. According to the DeepSeek-R1 documentation, its pre-training cost is less than 6 million US dollars. The R1 version explores everything from supervised fine-tuning of SFT to zero-supervision fine-tuning. Through the combination of GRPO and an efficient reward system, performance improvements can be achieved while significantly reducing costs. DeepSeek breaks America's monopoly advantage in AI competition. AI equality will change the global technology industry pattern, and ignite expectations that applications will explode.

Currently, AI applications are on the eve of an explosion, and demand for computing power will take the lead. Big model companies such as ByteDance Doubao and Ali Tongyi Qianwen are expected to quickly follow up and innovate, and vertical models are in full bloom.

Model companies are racing, leading to a “pulse-like” increase in demand for training computing power; demand for inference computing power and local deployment demand for large models of large enterprises are also growing rapidly. The bank is optimistic about investment opportunities in AI cloud computing, AIDC data centers, AI chips, and high-end process companies. On the application side, scenarios such as intelligent driving, AI advertising, AI agents, AI mobile phones, and physical intelligence are expected to gradually be implemented.

Application outlook: The future of AI is not only about chasing more powerful models, but also about solving real-world problems and creating economic value through AI applications.

From technological innovation to application implementation, from changing expectations to creating value, from low-quality domestic involvement to high-quality global development, Chinese technology enterprises still have a long way to go. Currently, America's big model companies are still significantly ahead, and they still have strong advantages in cutting-edge fields such as multimodality, world models, and reinforcement learning. Looking forward to the future, the US will continue to raise the upper limit of AI technology. China's greater opportunity lies in combining AI technology with traditional industrial infrastructure to solve practical problems and create economic value.

According to the bank's judgment, applications in fields such as intelligent driving, AI advertising, and AI agents will be the first to be implemented, contributing to revenue growth and changing the market pattern in 2025. The innovation of Apple's iPhone 17 in the third quarter is worth watching. The AI phone may bring a new round of end-side intelligence opportunities, and big local models such as Alibaba's Tongyi Qianwen are expected to participate. Embodied intelligence benefits from China's solid manufacturing base and is expected to become another advantageous industry in China. In terms of industry applications, AI+ Internet, AI+ finance, AI+ healthcare, AI+ industry, and AI+ transportation are expected to take the lead. Looking at the business chain, AI+ audit, AI+ supervision, and AI+ research are expected to significantly improve operation management efficiency. In the farther future, where the physical world and the digital world are deeply connected and mutually influenced, the future of the metaverse is expected to arrive at an accelerated pace.

Revaluation logic: R&D spending in the technology sector goes from a “negative valuation” to a “positive valuation.”

In the post-pandemic era, China's technology assets continued to be undervalued, mainly due to: 1) the slowdown in profit growth, leading to insufficient investor confidence in the profit growth of the main business; 2) China's AI was viewed as “intergenerational backwardness,” and “R&D investment” in new businesses was interpreted as a “loss” and eroded the valuation. In the past few years, Chinese tech giants have often been valued by PE in the overall law. As a result, losses due to R&D of new businesses are reflected in negative valuations in the total market value. The increase in profit pressure from the main business has intensified the valuation differentiation between Chinese and US tech giants. The revaluation brought about by DeepSeek is ostensibly an increase in the PE valuation multiples of leading companies. In depth, it is a shift in China's AI assets from negative valuation to positive valuation. In the future, if economic expectations improve, there is room for improvement in the valuation ratio of the main business; once the AI business achieves significant revenue contributions, it is also expected that there will be a shift in PS segment valuation. According to the bank's judgment, the revaluation of China's technology assets is still in its early stages. In the future, every major technological breakthrough and application implementation will have an opportunity to bring about a systematic increase in valuation. However, at the same time, the experience of the PC and mobile internet wave shows that even within a major technology cycle, there will be twists and turns; the experience of revaluing AI assets in US stocks also shows that continuous improvement in valuation requires quarterly implementation of performance as verification.

Investment opportunities: China's AI assets are the first to launch the Internet, focusing on computing power and domestic chip manufacturing processes, intelligent driving, and AI agents.

China's AI is being iterated at an accelerated pace, and large domestic models and smart cloud companies are expected to benefit significantly. The Internet is not only an important implementation scenario for AI, such as WeChat's access to DeepSeek, the accelerated implementation of AI advertisements, and increased demand for AI content generation; it is also a provider of intelligent clouds and domestic large-scale models. Under the grand narrative of Chinese AI, the revaluation of Internet companies will continue, but the sustainability of the revaluation depends on the quarterly verification of the performance of leading companies. In terms of computing power, similar to Nvidia's position in the US stock asset revaluation in the past two years, SMIC will play an important role in the revaluation of China's technology assets in the next 1-2 years. SMIC's pricing power, revenue growth rate, and profitability of advanced manufacturing processes are expected to become important direction guides for the market. On the application side, intelligent driving is expected to double the penetration rate in 2025, and the Chinese automobile industry is beginning to explore globalization. Traditional enterprises have begun experimenting with deploying DeepSeek locally to accelerate the implementation of large AI models. Fields such as AI+ finance, AI+ healthcare, AI+ industry, and AI+ transportation are worth paying attention to, and AI agents are expected to become an important form of application implementation. Furthermore, AI will hopefully feed back the digital upgrading of the industry and bring performance opportunities for sensor, PC, server, and IT service companies.

Risk factors:

The risk of increased geopolitics and friction; the risk that macroeconomic growth will fall short of expectations; the risk that AI applications will not be implemented as quickly as expected; the risk that the performance of leading technology networks will not meet expectations in the coming quarter, etc.