In the era of Manus accelerating AI, the AI application software sector has sparked a "gold rush".

On the early morning of March 6, the Chinese AI team Monica announced the launch of the world's first universal AI intelligent agent product, Manus. Due to its performance comparable to Open AI's top models, it caused a sensation in the first and second-tier markets as soon as it was released. According to the official website, Manus can help users with travel planning, stock analysis, creating PowerPoint presentations, financial report analysis, and other tasks with just one click. Subsequently, Manus quickly gained popularity in the tech industry, with some media even calling it the "GPT moment of AI Agents." In the second-tier market, there was also a frenzy. By the close of trading on March 6, concept stocks related to AI intelligent agents such as Newcapec Electronics, Hand Enterprise Solutions, and Digiwin Co., Ltd., all hit their limit up in the A-share market. AI application software stocks in the Hong Kong market also saw a significant rise, with the application software sector leading the gains by over 10%, including MING YUAN CLOUD (00909) surging over 23%, TUYA-W (02391) rising by more than 14%, and KINGSOFT (03888) rising by over 11%, with other concept stocks following suit. However, this upward trend did not last long and was broken by external doubts. Due to Manus's limited invitation-only testing, many viewers and even industry insiders were unable to experience the innovative experience brought by this AI Agent, leading to criticism of it being a combination of "starvation marketing and hype." Faced with public pressure, Manus AI partner Zhang Tao clarified multiple times that they had "never invested in any marketing budget" and that the invitation code mechanism was due to "limited server capacity necessitating it." At present, we will reserve judgment on the authenticity of Manus for now. After all, it is undeniable that there is a strong desire in the market for AI Agent products. Since the success of DeepSeek, who can create the next super Agent has become the focus of the industry, and the investment hearts of investors in the second-tier market are eager to try. Who can strike gold in the "first year of commercialization of AI Agents" has become a common issue for investors. Accelerating the AI application era, the industry is entering a "singularity moment." In December 2024, the first version of DeepSeek-V3 went online and was simultaneously open-sourced. On January 20, 2025, the company further released and open-sourced the DeepSeekR1 model, which showed outstanding performance in areas such as mathematics, programming, and natural language reasoning, even rivalling OpenAI's O1 model. According to related media reports, the performance of DeepSeek's large models is on par with leading global models like GPT, while the training costs are far lower than GPT's. The pre-training cost of DeepSeek's R1 model was only $5.576 million, less than one-tenth of the training cost of the OpenAIGPT-4o model. It is clear that the emergence of DeepSeek has significantly reduced model invocation costs and is an open-source model, which is undoubtedly a significant cost advantage for downstream AI applications, potentially promoting the accelerated adoption of AI Agents and various vertical "AI+ applications," allowing for higher ROI for downstream applications and early profitability. Looking at the development process of this AI wave, it can be seen that it is somewhat similar to the wave of mobile internet from 2010-15, following the transmission law from hardware devices, software content to scenario applications. During the surge of the mobile internet from 2010-15, the improvement of communication infrastructure and the emergence of popular products marked the beginning of smartphones replacing feature phones, followed by the flourishing development of the content industry, and traditional industries embracing the wave of the internet. During the acceleration phase, China's content and application side blossomed, and with the gradual increase in smartphone penetration rates, China Mobile Limited's internet user base and internet usage time also significantly increased. According to People.cn Co., Ltd. and the China Internet Network, by 2015, the number of mobile internet users in China had reached 620 million, with the average weekly internet usage per capita increasing from 20.5 hours in 2012 to 26.2 hours in 2015. In this context, social media, instant messaging, e-commerce, online payments, games, and other applications quickly emerged to meet the growing diverse needs of users. For example, in 2014 and 2015, China Mobile Limited's game sales revenues reached 40 billion and 72.4 billion RMB, with year-on-year growth rates of 191% and 81%, respectively. Comparing the transmission law from hardware devices, software content to scenario applications, with the rise of Deepseek in the AI boom, AI applications are moving towards an acceleration period from 1-100. According to BOC International, since 2024, the active users of AI applications in China have grown rapidly, and top applications like Dou Bao have become the second-largest AI application globally in terms of MAU, second only to ChatGPT. With the market focusing on Deepseek in January of this year, by January 2025, its MAU had reached nearly 80 million. Since the second half of 2024, the call volume for the Dou Bao large model has grown rapidly, with daily token usage exceeding 40 trillion in December, a growth of over 33 times since the launch in May. Based on this, it can be seen that the growth slope of AI applications is approaching that of the top mobile internet apps like Meituan during the mobile internet boom of 2014-2015. Top applications like Dou Bao may achieve over 100 million monthly active users by 2025, and top AI applications are moving towards a "singularity moment" of qualitative change from quantitative change. With the number of domestic AI application users entering an accelerated growth phase, the launch of DeepSeek is expected to further accelerate the commercialization of application-side AI. The upward trend effect is here, how should we position ourselves in the AI software application sector? In fact, similar to the mobile internet surge from 2010-15, when mapping toStock market performance, stock market trends also start from hardware devices, gradually transition to software content, and spread to scene applications.Similar to the early stages of the mobile Internet wave 10-15 years ago, in the early stages of the AI application wave 23-24, the performance certainty of hardware end is higher, and the increase in stock price is more significant. Taking the example of NVIDIA, the leader in AI "shoveling people," in the AI wave from the end of 2022 to the present, due to the strong certainty of the computing power sector, NVIDIA has deeply benefited as a global leader in computing power. In 2023, NVIDIA's data center business revenue, where AI chips are located, grew by 137%, and its revenue proportion increased from 39% in 2021 to 78% in 2023. Since October 2022 (data as of 25/02/25), NVIDIA's stock price has increased by up to 1318%. Similarly, the leading A-share company in the optoelectronic module sector, Zhongji Innolight (300308.SZ), also experienced this trend. The performance of the optoelectronic module sector has gradually improved under the drive of the AI computing power outbreak. The net profit attributable to the parent company of A-share optoelectronic module sector has significantly rebounded from -8% in 23Q3 to 101% in 24Q3. Since October 2022, the largest increase in Wind Optoelectronic Module Index is 314%, with the leading company Zhongji Xulie increasing by 898%. Therefore, it is not difficult to speculate that with DeepSeek catalyzing the acceleration of AI application landing, the upward effect of the sector is expected to spread from hardware devices to software content and scenario applications. Specifically, combining the insights of the track rules of the mobile Internet era, sectors with clear application scenarios, easy-to-verify business models, low investment costs in the industrial chain, and high-frequency user needs are more likely to achieve commercialization first. In different vertical fields of AI applications, fields such as office, entertainment, medical, education, finance, and e-commerce are especially worth focusing on. For example, in the field of AI + office, AI technology can be widely used in the office field to optimize work processes, improve productivity, and provide a better user experience. Currently, "AI + office" is in the early stage of cultivating paying users, and the potential for future development cannot be ignored. In the Hong Kong stock market, concept stocks such as KINGSOFT (03888) and KINGDEE INT'L (00268) in the AI + office sector may also benefit greatly from this AI wave. In conclusion, although it is not yet clear whether Manus' success is due to hype or real ability, but with DeepSeek accelerating the landing of AI applications, the investment effects of AI applications are increasingly apparent. In the application field, the vast market and user base in China will provide a fertile soil for AI application innovation, and domestic enterprises have great potential. Investors can follow the development rule of the mobile Internet wave that "the stock market trend will start from hardware devices and gradually spread from software content to scenario applications." After all, the principle of "learning from history" never goes out of date.