The 2026 WAIC Qiming Venture Partners · Entrepreneurship and Investment Forum — The Evolution Journey From the Origin of Computing Power to Application Implementation was successfully held.
On July 19, 2026, the "Qiming Venture Partners · Entrepreneurship and Investment Forum — The Evolution Journey from the Origin of Computing Power to Application Implementation", hosted by Qiming Venture Partners, was successfully held at the Shanghai World Expo Center during the 2026 World Artificial Intelligence Conference (WAIC).
This forum focuses on the cutting-edge trends of AI in the fields of entrepreneurship and investment, bringing together top scientists, innovation leaders, and senior investors. It featured 3 keynote speeches, 1 in-depth dialogue, and 1 roundtable discussion, and grandly released three core achievements to jointly decode the paradigm shift in the era of large models: exploring how the industry can move beyond the mere pursuit of parameters to the ultimate exploration of real scenarios, cost reduction and efficiency improvement, and closed-loop business models.
As one of the most forward-looking and influential venture capital institutions in the field of artificial intelligence, this marks the fourth consecutive year that Qiming Venture Partners has hosted this forum. It is also the only forum at this year's World Artificial Intelligence Conference initiated by a venture capital institution, aiming to showcase and discuss the latest breakthroughs of AI moving from the underlying computing power to industrial implementation from an innovation perspective.
Qiming Venture Partners has systematically laid out the artificial intelligence field since 2013. Up to now, Qiming Venture Partners has invested in more than 100 projects in the AI field, with a total investment of 120 billion RMB. Through the full industrial chain layout covering underlying computing power, foundation models, multimodal technology, embodied intelligence, and vertical applications, Qiming Venture Partners has successfully discovered and boosted the growth of many innovative benchmark enterprises, witnessing AI entering the era of intelligent explosion.
In the opening speech themed "Surpassing the Technology Wave: Finding the Business Inflection Point in the AI Era", Zhou Zhifeng, Managing Partner of Qiming Venture Partners, first sorted out the historical context of AI evolution. In his view, technological evolution has drastically reduced the marginal cost of more and more capabilities. The development speed, intensity, and ferocity of the AI industry have surpassed any technological wave in the human industrial and information revolutions over hundreds of years. Qiming Venture Partners attempts to overlook the overall AI landscape with a "helicopter perspective" to capture vivid signals in four major sectors: models, embodied intelligence, computing power, and applications.
Zhou Zhifeng, Managing Partner of Qiming Venture Partners
Looking back over the past year, Zhou Zhifeng observed that the AI industry has rapidly formed market consensus in three major areas.
At the foundation model level, the comprehensive performance of China's leading large models continues to approach the cutting-edge models in the United States. The current gap between the top large models of China and the United States has been shortened to two to three months. The overseas usage scale of Chinese models continues to rise, and the average daily token consumption of Chinese models has increased by 3,000 times within two and a half years, which conforms to the "Jevons Paradox" — the continuous decline of model cost instead leverages exponential demand for computing power. At the same time, the model layer has ushered in an unprecedented capital dividend period, with the primary and secondary markets linked, and capital accelerating to favor leading AI model enterprises.
In the field of embodied intelligence, China has advantages in all industrial chain elements. At the hardware level, a full-size humanoid robot over 170cm has about 1,200 parts. Except for CPU and GPU, the other more than 1,000 parts can all be produced in China. In addition to hardware advantages, China also has massive application scenarios under the most complete industrial categories, data collection advantages, and reserves of algorithm and engineering talents. Over the past two to three years, more than 300 enterprises have been born in this field, and the valuation of many enterprises has exceeded 10 billion and 20 billion RMB. The industry will move towards fierce competition.
At the computing power level, the first batch of domestic computing power pioneers solved the problem of "getting enough to eat". At present, computing power chips with multiple new technical routes cut in from multiple points such as inference to challenge the problem of "eating well". The production capacity of storage, advanced manufacturing processes, and advanced packaging continues to be in short supply, and computing power reserve has been upgraded to a core strategic asset for enterprises. Industry competition is no longer limited to a single chip, but extends to the full system dimensions of interconnection, heat dissipation, and power supply. At the same time, security and trustworthiness are no longer optional options, and have become the three core indicators for large-scale AI implementation, alongside product effectiveness and token cost.
In contrast to the formation of the three major consensuses, the "way of opening up" of AI in the application field is different from expectations. He believes that although China is the fastest-growing generative AI application market in the world with 515 million users, the commercialization process is slower than market expectations. The traditional "scale first, monetization later" model of the Internet and mobile Internet no longer applies. Expanding AI users will continuously increase token costs, and there is no marginal cost advantage. It is difficult to grow bigger by only improving the experience on existing businesses. Only business models that reconstruct the supply-demand relationship can give birth to phenomenal products.
"In the next stage, AI will move from technically usable to sustainable in value. Building a sustainable business ecosystem is the core prerequisite for the birth of phenomenal products," Zhou Zhifeng stated.
Subsequently, Zhou Zhifeng and Hu Qi, Executive Director of Qiming Venture Partners, jointly released Qiming Venture Partners' 2026 Top 10 AI Outlooks:
Foundation Models
Outlook 1: In the next 12-24 months, top-tier models will internalize most of the "add-on" capabilities, including task planning, tool invocation, multi-agent collaboration, and partial harness engineering capabilities.
Outlook 2: Multimodal models will further evolve toward interactive world modeling, becoming a key technical path for AI to obtain environmental perception, long-term planning, and large-scale implementation in the physical world.
Embodied Intelligence
Outlook 3: The effective data of leading robotics companies will jump from the "10,000-hour level" in 2025 to the "million-hour level" in 2027, with human first-person perspective data occupying an absolute dominant position.
Outlook 4: Dexterous hands will see accelerated development. Dexterous hands with tactile perception will gradually mature and continue to reduce costs, forming a high-low matching with the two-finger gripper solution, and gradually penetrate on a large scale in complex operation scenarios.
AI Infrastructure
Outlook 5: The focus of AI computing power demand will shift to inference, and the production capacity of storage, advanced manufacturing processes, and advanced packaging will be increasingly tight. AI infrastructure will continue to face structural shortages in the next two years, and computing power asset reserves will be upgraded to the core strategy of AI enterprises.
Outlook 6: AI infrastructure will enter the stage of system-level competition, with competition dimensions covering key links such as chips, interconnection, heat dissipation, and power supply. In the next two years, computing power chips and super-node large clusters based on new architectures are expected to emerge, to achieve low-cost and high-efficiency token production.
Outlook 7: Security and trustworthiness, along with product effectiveness and token cost, will become three key variables affecting the large-scale implementation of AI for enterprises, and secure and trustworthy AI will be upgraded from an optional item to a mandatory item.
AI Applications
Outlook 8: In the next 12-24 months, the business models of AI applications will accelerate to break away from the freemium logic of the Internet era, shifting to pricing based on results and value. The core metric for measuring AI companies will shift from user scale to the commercial value created per unit of intelligent cost.
Outlook 9: In the next 12-24 months, the commercialization of AI applications will focus on vertical scenarios and high-paying users. The efficiency side (save time) will take the lead in breaking out before the consumer side (kill time). With the continuous decline of token costs and the innovation of interaction paradigms, phenomenal AI consumer applications will gradually emerge.
Outlook 10: In the next 12-24 months, AI-Native organizations will move from concept to empirical evidence, and a group of enterprises will achieve several times the per capita output of traditional organizations.
As a global leading R&D team for multimodal models and world models, Shengshu Technology open-sourced the world's first world action model with a unified architecture based on video generation large models in 2025. Luo Yihang, Co-founder and CEO of Shengshu Technology, shared the team's latest explorations in multimodal generation and world models in his keynote speech titled "General World Model: Connecting Virtual and Real Worlds, Building the Foundation of Physical Intelligence". He stated, "AI is moving from understanding and generating the digital world to understanding and acting in the real physical world, and the world model is exactly the core foundation of this transformation. Shengshu Technology is the world's first general world model company that realizes the unification of the digital world and the physical world, building a general world model from understanding to generation to action, providing intelligence for digital generation and physical actions."
Luo Yihang, Co-founder and CEO of Shengshu Technology
Facing the era of production-grade agents, the development of foundation models is moving from pursuing the upper limit of model capabilities to a new stage that balances inference efficiency, engineering reliability, and deployment cost. Zhu Yibo, Co-founder and CTO of Step Star, stated in his keynote speech titled "The Integration and Promotion of AI Infrastructure and Model Capabilities" that entering the agent era, the core goal of AI infrastructure is shifting from simply reducing inference costs to pursuing the comprehensive optimization of intelligence, speed, and cost at the same time. With the rapid growth of token demand and the continuous expansion of model scale, relying solely on computing power expansion is unsustainable. The industry needs to promote the co-design of models, systems, and chips to release more intelligence per unit of computing power. For intelligent terminals, the infrastructure also needs to further support cloud-edge collaboration, long-term operation environments, etc., so that real-time execution on the edge side and complex cloud inference can cooperate efficiently, becoming the key foundation for the large-scale implementation of agents.
Zhu Yibo, Co-founder and CTO of Step Star
Subsequently, Tang Wenbin, Founder and CEO of Forceling, released the latest global progress of RoboChallenge, including the latest achievements of the embodied foundation model leaderboard, the results of real-robot challenge competitions at top international academic conferences such as CVPR and ICRA, and the latest progress of EAI Bench officially released at the AI for Good Global Summit in Geneva. At the same time, it announced a new batch of members of the RoboChallenge International Organizing Committee and ecological partners. Relying on an open and shared real-robot evaluation platform, it connects models, ontologies, developers, and industrial partners to jointly build a global open ecosystem of embodied intelligence, promoting China's embodied intelligence technology, standards, and ecosystem onto the international stage.
Tang Wenbin, Founder and CEO of Forceling
Tang Wenbin said: "Forceling, together with Qiming Venture Partners, will build RoboChallenge into a public welfare and neutral embodied intelligence evaluation platform. We insist on driving innovation through evaluation, promoting standard construction through openness, and actively promoting deep industry-university-research collaboration. We want more scenario parties to clearly see the current technical boundaries of embodied models, and accelerate the entry of robots into reality."
As a domestic leader in invasive brain-computer interfaces, NeuroXess's brain-computer interface system has taken the lead in entering the "Green Channel" for innovative medical devices of the National Medical Products Administration. In the dialogue with Hu Xubo, Managing Partner of Qiming Venture Partners, Zhao Zhengtuo, Founder of NeuroXess, shared the latest progress of brain-computer interfaces. He believes that brain-computer interface (BMI) is the "key hub" that breaks human biological limits and enables efficient human-machine connection. He put forward an insight of great industrial value: brain-computer interfaces will become a key data source for the evolution of embodied intelligence. Through the ultimate engineering of the invasive path, NeuroXess allows subjects to precisely control physical devices with their minds, which can not only replace damaged functions, but also make them "brain trainers" for embodied intelligence, efficiently generating scarce high-precision control data in real scenarios. He predicts that in the next 8-10 years, with the miniaturization and low-invasiveness of invasive brain-computer interface systems to the size of "ear studs", this technology is expected to enter the lives of ordinary people.
Hu Xubo, Managing Partner of Qiming Venture Partners; Zhao Zhengtuo, Founder of NeuroXess
At present, the iterative speed of artificial intelligence technology continues to accelerate, and industrial application scenarios are fully implemented. The opportunities and challenges brought by technological innovation are becoming increasingly diversified and complex, and global artificial intelligence governance has entered a development process of rule improvement, model innovation, and global collaboration. Professor Wang Dong, Executive Director of the Institute for Global and People-to-People Exchanges of Peking University, on behalf of the research team, released the report "The International Landscape of Artificial Intelligence Governance". Supported by Qiming Venture Partners, this report is based on the general trend of global digital transformation, providing solid theoretical support and practical references for bridging governance differences among countries, building multilateral dialogue bridges, and forging international consensus on the safe development of artificial intelligence.
