Alibaba has unveiled an even bigger AI story.
```
Author | Chai Xuchen
Editor | Zhou Zhiyu
Last year, Wu Yongming described AI's greatest imagination as connecting to the real world. This year, Alibaba's story in AI is even bigger: artificial intelligence will be able to rapidly amplify human capabilities.
“The ultimate goal of artificial intelligence is to develop super artificial intelligence (ASI) capable of self-iteration and completely surpassing humans.” Wu Yongming said this at the 2025 Yunqi Conference on September 24th. In today's world, an AI-driven intelligent revolution has just begun.
In Wu Yongming's view, in the past few centuries the Industrial Revolution amplified human physical power through mechanization, and the Information Revolution amplified human information processing through digitization. “But this time, the intelligent revolution will far exceed our imagination. General artificial intelligence (AGI) will not only amplify human intelligence but also unleash human potential.”
In the past three years, the industry has already felt the power of AI, with concrete numbers to prove it.
In just a few years, AI's intelligence has rapidly risen from the level of a high school student to that of a PhD. The speed at which AI is penetrating industries has surpassed that of any technology in history. Wu Yongming revealed that right now, token consumption is doubling every two or three months. In the past year, total global investment in AI has exceeded $400 billion, and in the next five years, global cumulative AI investment will exceed $4 trillion.
Wu Yongming bluntly said, this is the largest ever investment in computing power and R&D in history, which will inevitably accelerate the birth of more powerful models and the penetration of AI applications. Riding on this momentum, Wu Yongming announced Alibaba would continue its ALL IN stance, advancing the three-year, 380 billion RMB AI infrastructure construction plan, and will continue to increase investment even further.
This has filled the market with great expectations for Alibaba's future, with Alibaba's Hong Kong stock price soaring over 7%. The capital market’s reaction indirectly shows that Alibaba is about to enter a blue ocean of huge potential. At the event, Wu Yongming told an even sexier story: “AI will not stop at AGI, it will move towards super artificial intelligence (ASI) capable of self-iteration and surpassing human intelligence.”
Wu Yongming believes the goal of AGI is to free humans from 80% of daily work, so people can focus on creation and exploration. ASI, as a system that completely surpasses human intelligence, may create groups of “super scientists” and “full-stack super engineers.”
“ASI will, at unimaginable speed, solve scientific and engineering problems unsolved today, such as conquering medical challenges, inventing new materials, solving sustainable energy and climate issues, even interstellar travel, and more.” Wu said, ASI will drive technological leaps at exponential speed and lead us into an unprecedented intelligent era.
The road to ASI is gradual, and Wu divides it into three stages: “intelligence emergence,” where AI gains generalized intelligence through learning massive human knowledge; “autonomous action,” where AI masters the use of tools and programming to assist humans; and finally, “self-iteration,” where AI connects to the physical world and achieves self-learning, ultimately “surpassing humans.”
The first stage, “intelligence emergence,” has been the main thread of recent years; the development of the Internet provided the foundation for emergent intelligence.
The Internet has digitized almost all human knowledge in history. The information carried by these languages and words represents the entirety of human knowledge. Based on this, large models first gained generalized intelligence by understanding the world's knowledge set, showing general conversational abilities, able to understand human intentions, answer human questions, and gradually develop the ability to reason multi-step problems.
The second stage is “autonomous action,” which is precisely the stage the industry is now in. Wu Yongming believes that in this stage, AI is no longer limited to language communication, but has the ability to act in the real world. Under human goal-setting, AI can break down complex tasks, use and create tools, independently interact with both the digital and physical worlds, thus having a huge impact on the real world.
The key to crossing the current stage lies in the ability to use tools (Tool Use).
“With Tool Use, AI can, like humans, call external software, interfaces, and physical devices to perform complex real-world tasks. In this stage, as AI can help humans greatly improve productivity, it will quickly penetrate logistics, manufacturing, software, business, biomedicine, finance, scientific research, and virtually every industry,” said Wu Yongming.
In Wu’s vision, in the future, natural language will be the source code of the AI era, and anyone will be able to create their own Agent with natural language. “In the future, there will be more Agents and robots working alongside humans than there are people globally, creating a huge impact on the real world. In this process, AI will connect the vast majority of real-world scenarios and data, creating conditions for future evolution.”
Once these conditions are met, AI will enter the third stage—“self-iteration,” with the defining characteristic of “surpassing humans.”
In Wu Yongming's view, this stage has two thresholds. First, AI needs to access the full, raw data of the real world.
“Currently, the fastest-advancing areas in AI are content creation, mathematics, and code. But in other areas and the broader physical world, AI still lacks wide, raw data that interacts with the physical world.” Wu Yongming bluntly said that to achieve breakthroughs surpassing humans, AI needs to obtain more comprehensive, raw data directly from the physical world.
Besides absorbing raw data, the second threshold is achieving self-learning.
Wu Yongming believes that as AI penetrates more real-world scenarios and understands more real-world data, models and agent capabilities will become stronger; there will be an opportunity to build their own model training infrastructure, optimize data flows, and upgrade model architectures, enabling self-learning. This will be a critical moment for AI development.
This means that future models will, through continuous interaction with the real world, acquire new data and receive real-time feedback. Leveraging reinforcement learning and continuous learning mechanisms, they will self-optimize, correct biases, achieve self-iteration and intelligent upgrades. “AI will iterate itself into intelligence surpassing humans—a nascent super artificial intelligence (ASI) will take shape.” said Wu Yongming.
Once a certain singularity is crossed, human society will feel as if the accelerator button has been pressed: the speed of scientific advancement will outpace our imagination, and a new burst of productivity will propel human society into a brand new stage. This road towards super artificial intelligence is becoming ever clearer before our eyes.
It is worth mentioning that the influence of AI iteration will eventually radiate to thousands of industries, sparking a huge transformation in the IT industry. Wu Yongming put forward several judgments, which may be crucial support for Alibaba’s future market re-evaluation.
First, Wu thinks “Large models are the next-generation operating system,” taking the place of today's OS.
“In the future, almost all tool interfaces connected to the real world will be linked to large models; all user demands and industry applications will be executed through large model-related tools. Large models will devour software.”
Wu said, as the next-generation operating system, large models will allow anyone to use natural language to create unlimited numbers of applications. In the future, almost all software dealing with the computing world may be Agents generated by large models, not today’s commercial software. Such imagination comes with it: “Potential developers will increase from tens of millions to hundreds of millions.”
Meanwhile, AI models will rapidly penetrate all devices. “Future models will run on all computing devices, with persistent memory, edge-cloud collaborative operating status, and may even update parameters and self-iterate at any time, similar to today’s OS running in various environments,” Wu asserted.
Based on the above vision, Alibaba has made a strategic choice: increases in infrastructure investment and opensources Tongyi Qianwen, with the overt ambition of “building the Android of the AI era.”
As software rapidly branches out, the underlying hardware market will also be upended. Wu Yongming’s new judgment is—Super AI Cloud is the next generation of computer.
First, large models will become the foundational system, with everyone having dozens or even hundreds of agents. These agents work and collaborate around the clock, requiring massive computation resources. And the computing paradigm in data centers is also revolutionizing, shifting from traditional CPU-centric computing to AI computing centered on GPUs.
The new AI computing paradigm needs denser computing power, more efficient networks, larger cluster scale. All this requires ample energy, full-stack tech, millions of GPUs and CPUs, with chips, storage, databases, and network operating efficiently together, processing worldwide requests 24/7. This needs ultra-large-scale infrastructure and full-stack technical accumulation.
Wu Yongming believes that only super AI cloud can bear such massive demand. He asserts that in the future, there may only be 5-6 super cloud computing platforms globally.
“In this new era, AI will replace energy as the most important commodity, driving the daily work of thousands of industries. The vast majority of AI capability will be generated and delivered in the form of tokens on cloud computing networks. Tokens will be the electricity of the future.”
With the blueprint drawn, Alibaba’s own positioning for this new era is to be a full-stack AI service provider, offering world-class smart capabilities and a global AI cloud computing network.
According to the vision, in the future, the Tongyi Qianwen model side will become a global open source AI foundational “system”; the Bailian platform will offer one-stop model services, allowing developers to easily utilize model capabilities and create/use agents; underpinning it at the hardware and network layer, Alibaba Cloud’s self-developed core storage system, network architecture, and compute chips are the most solid bedrock of this new story.
Wu Yongming revealed that Alibaba Cloud is making every effort to build a brand new AI supercomputer, capable of collaborative innovation in fundamental infrastructure and model architecture, ensuring that calling and training large models on Alibaba Cloud achieves the highest efficiency.
Looking back, Wu Yongming still believes the speed of AI industry development far exceeds expectations, and the industry’s demand for AI infrastructure also far exceeds our expectations.
The passionate story is right in front of our eyes: “Compared to 2022, the first year of GenAI, by 2032 the energy consumption of Alibaba Cloud's global data centers will increase tenfold.”
Wu Yongming promises that in preparation for the ASI era, Alibaba will actively advance its three-year 380 billion RMB AI infrastructure construction plan, and will continue to invest even more.
Risk Warning and DisclaimerThe market has risks, investment needs caution. This article does not constitute personal investment advice, nor does it take into account individual users’ special investment goals, financial situations, or needs. Users should consider whether any opinions, views, or conclusions in this article are appropriate for their particular situation. Investing accordingly is at your own risk. ```