In 2025, a profound shift is taking place in the mainstream narrative surrounding artificial intelligence: a new paradigm based on large models as the fundamental capability and Agent (intelligent agent) as the core of interaction is accelerating its penetration into various industries. The explosive growth of technological capabilities is also forcing a comprehensive reconstruction of business logic, organizational forms, and investment strategies.

On September 17th, the “2025 Imagination Conference” initiated by Business Weekly and co – hosted with People and Daily People was held in Beijing. With the theme of “X Rhapsody”, the conference invited many experts and entrepreneurs from the scientific and technological fields such as academia, the investment community, and the industrial circle to jointly explore the cutting – edge technologies in the AI industry and discuss the industry’s development trends.

During the opening speech session of the conference, Rong Bo, the chairman and founder of Boyaa Tianxia Communication Agency, said that we are in a stage where all industries are being accelerated and iterated by artificial intelligence. Those scenarios that once only existed in people’s imaginations are becoming a reality with the help of artificial intelligence. Now, artificial intelligence is no longer just a synonym for technology; it has become the strategic core of international competition. It is like a new “industrial revolution” that is redefining all industries.

â–² Rong Bo, the chairman and founder of Boyaa Tianxia Communication Agency

“The technological breakthroughs and industrial implementation are resonating in China. Countless entrepreneurs with dreams are transforming technology into business realities and conveniences in life,” said Rong Bo. The in – depth integration of large models in fields such as finance, healthcare, manufacturing, and smart cities has spawned a large number of intelligent upgrade demands. The explosive growth of generative AI applications, in turn, has driven the urgent demand for high – performance computing power and advanced chips. Since the beginning of this year, the valuation of the entire AI industry chain has been reconstructed, and the technology mainline has risen strongly. Concept stocks related to infrastructure such as artificial intelligence, semiconductors, and computing power chips have attracted the attention of global investors.

How to seize the AI dividends? Thirteen bigwigs from different fields gave their answers.

The AI competition has quietly shifted, inspiring new industrial prospects

At this conference, observers from academia and the investment community first outlined the development blueprint under the AI technological revolution.

Sun Tianshu, a professor of technology and operations at Cheung Kong Graduate School of Business and an outstanding dean’s chair professor, said that the essence of AI is an “intelligence revolution”. Different from all previous technological revolutions in human history, AI is no longer a “tool” to assist humans but “intelligence” itself, and it is constantly evolving and iterating.

â–² Sun Tianshu, a professor of technology and operations at Cheung Kong Graduate School of Business and an outstanding dean’s chair professor

Sun Tianshu emphasized that as AI enters the second half, intelligence is no longer scarce. What is scarce is the ability to “architect intelligence” in all industries. The industry is calling for “AI architects”.

Liu Mingda, the president of Alxa SEE Ecological Association and the chairman of Mingda Assets, shared his thoughts and practices on “long – termism”. He believes that from the international development process and considering China’s industrial environment, it can be predicted that China’s manufacturing and financial industries also have the soil to nurture world – class “Apple – like” enterprises and are expected to form Chinese profit oligarchs on the global stage in the future. At the same time, China’s “engineer dividend” in AI is yielding huge returns.

â–² Liu Mingda, the president of Alxa SEE Ecological Association and the chairman of Mingda Assets

In Liu Mingda’s view, for the investment concept in the AI era, the key lies in the saying – “Time is the best friend”. Only by going through the cycle and patiently accumulating can we witness the transformation from a tiny seed to a towering tree.

Feng Xiao, the managing director of Shengjing Jiacheng Venture Capital, systematically explained the current cycle of the AI industry, the evolution direction of Agent (intelligent agent) technology, and the optimal commercialization path.

“We speculate that intelligent agents may become the main body of the labor force and the main force in wealth creation. Economic activities will be coordinated, decided, executed, and optimized among intelligent agents (Agent2Agent), which we call the ‘intelligent agent economy’,” Feng Xiao said in his sharing. “The core elements of economic activities such as the labor market, currency, the Internet, social networking, retail, e – commerce, international trade, and going global, as well as future enterprises and organizations, may all be redefined.”

“We boldly envision that in the future e – commerce era, it may neither be the era of Taobao and JD.com where people search, nor the era of platforms like Xiaohongshu and Douyin that push content to attract consumers. Instead, it will be an Agent that represents you, knows what you need, what to buy, and what you like to interact with the Agents of platforms and merchants, and directly complete transactions, logistics, and deliver goods to your hands without human participation throughout the process,” Feng Xiao said.

Regarding the commercialization path of artificial intelligence, he believes that in the era of the intelligent economy, selling results is more in line with users’ needs than selling software or tools. Result – oriented AI RaaS will be the fulcrum for maximizing commercial value.

Chen Shi, an investment partner at Fengrui Capital, summarized the development of AI since 2025 and shared his thoughts on the development trends and entrepreneurial opportunities in the AI industry.

Chen Shi said that in the current AI industry, a “dual – core driving” pattern may emerge. The first core is the co – existence and interaction of “closed – source large models and open – source large models”; the second core is “the United States and China” as the main competitors and driving forces in terms of technology, capital, and policies. AI application entrepreneurs in this era should clarify their business models, build peripheral “scaffolding” around the model, adopt the strategies of rapid iteration and “laying eggs along the way”, and leverage the progress of model technology to continuously optimize product experiences and business models.

In Chen Shi’s view, in today’s large – model competition, it’s not about who has the largest model but about who can find a new balance in operating efficiency, cost – performance ratio, reasoning ability, application implementation, and ecosystem construction.

Connecting the collaboration chain, infrastructure synergy is the key
 

Currently, the most concerning issue for AI practitioners is how to make Agents truly bring about an efficiency leap and create visible value. However, for the large – scale application of Agents, there are still key challenges such as the difficulty of “communication” between different systems (difficulty in protocol interoperability) and high computing power costs. Fortunately, the entire industry is going all out to remove these obstacles.

Currently, the accelerated development and implementation of artificial intelligence have spawned a huge demand for computing power. As an advanced form of computing power infrastructure, intelligent computing centers have considerable strategic significance.

â–² Li Li, the secretary of the Party Committee and the president of Runze Intelligent Computing Technology Group Co., Ltd.

In the view of Li Li, the secretary of the Party Committee and the president of Runze Intelligent Computing Technology Group Co., Ltd., the development of computing power presents both opportunities and challenges. Building advanced computing power infrastructure will contribute to the early arrival of the AGI (Artificial General Intelligence) era.

Li Li mentioned that as the foundation of artificial intelligence, intelligent computing centers are also facing rapid iteration. He said that Runze Intelligent Computing Technology Group is also researching and developing a super intelligent computing large – scale scientific device. The new intelligent computing centers require larger scale, higher power, better cooling, shorter connections, and greater environmental friendliness.

What role will the new – generation AIDC (Artificial Intelligence Data Center) play in the urban industrial upgrade? Li Li said that it will deeply penetrate into vertical industry scenarios, promote the intelligent reconstruction of traditional industries through the trinity model of “computing power + algorithm + data”, accelerate the incubation of emerging industrial clusters and the improvement of the development level of regional systems, and stimulate the innovation vitality of the digital industry.

â–² Liu Huifu, the director and chief technology officer of iSoftStone

Liu Huifu, the director and chief technology officer of iSoftStone, shared the secret for enterprises to maintain competitiveness in the AI wave at the conference. As a representative of China’s leading digital technology service providers, Liu Huifu systematically explained iSoftStone’s transformation path and strategic insights in the face of the intelligent wave.

Liu Huifu pointed out that enterprises should pay attention to three points in their understanding of intelligence: AI innovates digital productivity, AI awakens the software pipeline, and AI triggers equipment breakthroughs.

He said that in the “Internet +” and “Mobile +” eras, iSoftStone has long been deeply involved in digital services and is actively embracing the opportunities and challenges in the “AI +” era. iSoftStone is firmly moving towards the path of integrated innovation of “hardware and software integration”. By constructing a full – stack intelligent system, it redefines the new scientific and technological collaboration paradigm of “people, digital, and equipment” in the AI era and releases future digital productivity. First, AI innovates digital productivity by cultivating digital employees with N – fold efficiency on a large scale; second, AI awakens the software pipeline to create a “lights – out factory” for software engineering; finally, AI triggers equipment breakthroughs by deeply integrating embodied intelligence, digital intelligent agents, and human wisdom to build the next – generation intelligent digital system.

Zhang Qing, a senior engineer at the Institute of Industrial Internet and Internet of Things of China Academy of Information and Communications Technology, pointed out that the rapid development of artificial intelligence is also profoundly reconstructing the digital industry architecture, gradually evolving from traditional services oriented towards “people” to services oriented towards “intelligent agents”. The digital industry is also shifting from being centered on “applications/platforms/systems” to a new stage centered on “intelligent agents”. In terms of form, from the traditional simple connection and passive response to the intelligent agent era, through the connection between intelligent agents and knowledge and among intelligent agents, autonomous decision – making can be achieved.

Intelligent agents have been widely used in fields such as scientific research and industry. Zhang Qing believes that in the face of the high complexity and uncertainty of multi – process collaboration in industrial manufacturing, the power of a single intelligent agent is often insufficient, and the interaction and collaboration of multiple intelligent agents are particularly important, becoming a new trend in the development of industrial AI.

Jia Anya, the person in charge of the Racoon family at SenseTime, analyzed the challenges and opportunities of large – model implementation on the enterprise side.

Jia Anya said that AI is transforming from a “productivity tool” to “productivity itself”. With the continuous breakthroughs in underlying models, AI will be more deeply integrated into enterprise processes and data. This means that in the future, the delivery of AI results will no longer only measure the process but will pay more attention to the final results and effectiveness. Enterprises should start from the basic layer, make small and fast progress, and verify the value, so that AI can truly serve business growth step by step.

Deeply explore vertical scenarios to dig for commercial value
 

The breakthroughs in technology and the construction of infrastructure ultimately need to be verified by solving high – frequency pain points in specific fields, creating irreplaceable user experiences, and establishing sustainable business models. The real value reconstruction cannot be achieved without in – depth exploration and practice in vertical scenarios.

â–² Xie Fuqiang, the vice – president of Mobvoi

Facing the current problems of unclear commercialization paths and limited application scenarios for large models, Xie Fuqiang, the vice – president of Mobvoi, shared his insights into the current dilemmas and opportunities in the AI hardware field at this conference.

Xie Fuqiang believes that “Agent is an effective way to break through the current AI commercialization dilemma”. Combining the technological trend of AI Agent, Xie Fuqiang proposed that the core of the success of AI hardware lies in the in – depth combination of “Agent + hardware + vertical scenario”. He said at the conference, “For AI hardware, the real value of hardware is that it can help algorithms and large models penetrate into