Before starting, I would like to give special thanks to the developers who support the tech industry in China and globally. Today marks the 10th anniversary of the Yunqi Conference, which originated as Alibaba Cloud’s developer conference. It is the developers who have driven the growth of cloud computing, AI, and the tech industry in China and around the world. So before beginning, I want to express my highest gratitude to the developers.
The world today is witnessing the dawn of an AI-driven intelligent revolution. Over the past few centuries, the Industrial Revolution amplified human physical capabilities through mechanization, and the Information Revolution amplified human information processing through digitization. This new intelligent revolution will exceed our imagination. Artificial General Intelligence (AGI) will not only amplify human intelligence but also unlock human potential, paving the way for the arrival of Artificial Superintelligence (ASI).
In the past three years, we have already clearly felt the speed of this progress. In just a few years, AI intelligence has grown from the level of a high school student to that of a PhD, even winning gold medals in the International Mathematical Olympiad. AI chatbots have achieved the fastest user penetration rate in human history. AI is penetrating industries faster than any technology in history. Token consumption doubles in just a few months. In the past year, global AI investments exceeded US$400 billion, and in the next five years, global AI investment is expected to exceed US$4 trillion. This represents the largest investment in computing power and R&D in history, inevitably accelerating the development of more powerful models and wider AI adoption.
Achieving AGI—a system with human-level general cognitive abilities—now appears to be a certainty. However, AGI is not the endpoint of AI development; it is a brand-new starting point. AI will not stop at AGI; it will progress toward ASI, surpassing human intelligence and capable of self-iterative evolution.
The goal of AGI is to free humans from 80% of routine work, allowing us to focus on creation and exploration. ASI, as a system surpassing human intelligence, could produce a generation of “super scientists” and “full-stack super engineers.” ASI will tackle unsolved scientific and engineering problems at unimaginable speeds, such as conquering medical challenges, inventing new materials, solving sustainable energy and climate issues, and even enabling interstellar travel. ASI will drive exponential technological leaps, leading humanity into an unprecedented era of intelligence.
We foresee three stages on the path to ASI:
Stage One: Emergence of Intelligence ——“Learning from Humans”
The past decades of internet development laid the foundation for intelligence emergence by digitizing nearly all human knowledge. This data represents the totality of human knowledge. Large AI models first develop generalized intelligence by understanding this global knowledge collection, emerging with general dialogue abilities, understanding human intentions, answering questions, and gradually developing multi-step reasoning. Today, AI is approaching top human levels in various academic disciplines, such as gold medal standards in international math competitions. AI is increasingly capable of engaging with the real world, solving real problems, and creating real value—this has been the main theme in recent years.
Stage Two: Autonomous Action ——“Assisting Humans”
In this stage, AI moves beyond language communication and can act in the real world. Under human goal setting, AI can break down complex tasks, use and create tools, autonomously interact with both digital and physical worlds, and make substantial real-world impacts. This is the stage we are currently in.
The key to achieving this leap is that large models now possess Tool Use capabilities, enabling them to connect to digital tools and complete real-world tasks. Just as human evolution accelerated through tool use, AI can now employ external software, APIs, and physical devices to execute complex tasks. With these abilities, AI can dramatically boost productivity across logistics, manufacturing, software, business, biomedicine, finance, research, and nearly all industries.
Additionally, improvements in large model coding capabilities allow AI to solve increasingly complex problems and digitize more scenarios. Current agents primarily handle standardized, short-cycle tasks. To tackle longer, more complex tasks, advanced coding abilities are essential, allowing agents to autonomously understand requirements and perform coding and testing, similar to a human engineering team. Developing these capabilities is essential on the path to AGI.
In the future, natural language will be the source code of the AI era. Anyone will be able to create their own agent simply by stating their needs in natural language. AI will write the logic, call tools, and build systems, completing almost all digital-world tasks while controlling physical devices through digital interfaces. Eventually, the number of agents and robots could exceed the global population, working alongside humans to impact the real world. AI will connect to most real-world scenarios and data, laying the groundwork for its evolution.
Stage Three: Self-Iteration ——“Surpassing Humans”
Two critical elements define this stage:
1. Access to complete real-world raw data
Currently, AI advances fastest in content creation, mathematics, and coding—fields where knowledge is fully human-defined and text-based. In other domains, AI mainly accesses summarized human knowledge, lacking interaction with raw physical-world data. To surpass humans, AI must obtain more comprehensive and raw data directly from the physical world. For example, designing a next-year car based solely on survey data cannot match the insights AI could gain if it had access to all the car’s real-world data. Direct interaction with real-world data is essential for AI to uncover deep patterns beyond human cognition.
2. Self-learning
As AI penetrates more physical-world scenarios and understands more data, models and agents can build their own training infrastructure, optimize data flows, and upgrade architectures autonomously—achieving self-learning. Continuous interaction with the real world allows AI to iteratively optimize and upgrade itself, eventually forming an early ASI.
Once a technological singularity is reached, human society will accelerate dramatically, with technological progress surpassing our imagination. New productivity breakthroughs will propel society into a new stage. The path to ASI is becoming increasingly clear.
Our Key Judgments:
1. Large models are the next-generation operating system.
Large models will replace current OS platforms as the central hub linking users, software, and AI computation. Natural language will serve as the programming language, agents as software, and contexts as memory. Large models, through interfaces like MCP and A2A protocols, will enable multi-agent collaboration, akin to PC-era buses and APIs.
2. Super AI Cloud is the next-generation computer.
The AI Cloud will provide massive computation resources for dozens or hundreds of agents per user. Computing paradigms are shifting from CPU-centric to GPU-centric AI computing, requiring dense compute power, efficient networks, and large-scale clusters. Only super AI clouds can support such enormous demand. Globally, there may eventually be only 5–6 super cloud platforms.
AI will become as fundamental as energy, driving daily work across industries. Tokens, representing AI capability, will flow through the cloud. Alibaba Cloud positions itself as a full-stack AI service provider, delivering developer-friendly AI services worldwide.
Key Achievements:
Tongyi Qianwen: Open-sourced over 300 models, covering all modalities and sizes, downloaded over 600 million times globally, with 170,000 derivative models.
Bailian Platform & AgentBay: Supporting model customization and agent development.
Hardware & Network: Alibaba Cloud operates one of the world’s leading vertically integrated AI cloud platforms.
Alibaba Cloud is building a new AI supercomputer with the most advanced infrastructure and models, ensuring maximum efficiency for developers. With a three-year US$480 billion investment plan in AI infrastructure, Alibaba Cloud is preparing for the ASI era, anticipating a tenfold increase in data center energy scale by 2032 compared to 2022.
Human-AI Collaboration:
With ASI surpassing human intelligence, humans and AI will collaborate in entirely new ways. Early examples include coding systems overnight with AI agents. In the future, every household, factory, and company may have numerous agents and robots working 24/7, potentially requiring hundreds of GPU chips per person. Just as electricity amplified human physical power, ASI will exponentially amplify human intelligence.
Finally, this is only the beginning. AI will reconstruct infrastructure, software, and applications, becoming a core driving force of the real world and igniting a new intelligent revolution. Alibaba will continue investing and collaborating with partners and clients to integrate AI deeply into industry and co-create the future. Wishing everyone a fulfilling and enjoyable Yunqi Conference. Thank you!
Editor: Zhongxiaowen
