AI News

Yang Yuanqing proposes "private-domain data flywheel": the next AI battle will be fought over data and hybrid AI

Sina TechUpdated Nov 16 21:09

AI summary

Lenovo Group Chairman and CEO Yang Yuanqing proposed the "private-domain data flywheel" at the 2026 Lenovo Capital and Incubator Group CEO Annual Meeting, judging that competition in the AI industry will shift from comparing model parameters to the implementation capability of hybrid AI. In the first quarter of Lenovo's new fiscal year, AI-related business revenue accounted for 35% of total revenue; in the third quarter, its global market share of AI PCs reached 25%, and the Tianxi personal super intelligent agent has been upgraded to 4.5.

Body · AI translation

Related topics top start Related topics top end Futures promotion begin Futures promotion end Miaopai begin Miaopai end Video player start Video player end Market chart begin Market chart end

From November 15 to 16, the 2026 Lenovo Capital and Incubator Group CEO Annual Meeting, themed "Gathering Light, Walking Together, Technology Toward the New," was held in Hangzhou. Yang Yuanqing, Chairman and CEO of Lenovo Group, delivered a keynote speech titled "Accelerating Transformation with Hybrid AI to Create a Shared Future of AI for All," explaining Lenovo Group's latest judgment on AI industry trends and the implementation progress of Lenovo's hybrid AI strategy.

The 2026 Lenovo Capital and Incubator Group CEO Annual Meeting, themed "Gathering Light, Walking Together, Technology Toward the New," was held in Hangzhou from November 15 to 16. Yang Yuanqing, Chairman and CEO of Lenovo Group, delivered a keynote speech titled "Accelerating Transformation with Hybrid AI to Create a Shared Future of AI for All."

In his speech, Yang Yuanqing pointed out that AI is moving from generation to execution, and agents are becoming the core of value. In 2026, agents were written into the Government Work Report for the first time, and a survey released by McKinsey in August this year also shows that the proportion of large enterprises deploying agents at scale is increasing. He said that the success of AI is not only measured by how many devices are sold and how much infrastructure is built, but also by whether agents are actually used and whether they continue to create value. For individuals, a successful agent must truly understand you and ultimately become your AI twin; for enterprises, a successful agent must be able to cross data domains, break down application silos, and let people and agents work side by side. How deeply an agent understands depends on how rich the data is and how deep and long the memory is.

He proposed the "private-domain data flywheel," in which personal devices and enterprise operations generate data, and the data trains agents; the better the agents understand users and enterprises, the more willing users and enterprises are to use them, thereby generating more data. Public data is available to everyone, and models will become increasingly similar. What truly creates a gap is unleashing the potential of users' unique private-domain data through personal intelligence and enterprise intelligence. Yang Yuanqing judged that competition in the AI industry will shift from comparing model parameters to comparing the implementation capability of hybrid AI. "The next battle for AI depends not only on models, but also on data, and on hybrid AI," he emphasized. The data that enterprises accumulate in various applications belongs to the users themselves, not to application software or SaaS service providers. Lenovo Group's role is to provide secure and trustworthy tools and platforms to help users turn data into value.

Yang Yuanqing said that the hybrid AI strategy first proposed by Lenovo Group at the beginning of the generative AI boom is becoming an industry consensus and translating into market results. After setting its best-ever performance in the previous fiscal year, Lenovo Group set another record for its best quarter in the first quarter of the new fiscal year, with AI-related business revenue accounting for 35% of quarterly revenue; its ranking in the 2026 Fortune Global 500 rose to 153rd, a record high.

In the field of personal intelligence, Lenovo is advancing its "one integrated body, multiple devices" strategy. In the third quarter of this year, Lenovo's global market share in AI PCs reached 25%, continuing to rank first. In the Chinese market, AI PCs accounted for more than 50% of Lenovo's notebook sales, with cumulative sales exceeding 7 million units to date. Lenovo continues to expand terminals such as AI hosts and AI glasses. Forty-two AI terminals passed the first batch of L3-level tests, the highest number in the industry; Tianxi personal super agent has been upgraded to 4.5, moving from passive invocation to autonomous execution.

In the field of enterprise intelligence, Lenovo continues to strengthen "device-edge-cloud" hybrid AI infrastructure and accelerate its layout in AI inference computing and edge computing. Lenovo uses technologies such as Neptune liquid cooling to optimize the cost and energy consumption of infrastructure throughout its lifecycle, matches models to application needs, builds token factories or Token Hubs, and reasonably allocates the proportion of token generation across devices, edge, and cloud to optimize the cost of each token. Yang Yuanqing believes that in the future, 70% to 80% of tokens will be generated at the edge or on devices, and 20% to 30% in the cloud. In the second quarter of 2026, x86 server shipments rose to first place globally for the first time, and the AI server order backlog is sufficient.

At the 2026 Lenovo Capital and Incubator Group CEO Annual Meeting, Yang Yuanqing said that competition in the AI industry will shift from model parameters to the implementation capability of hybrid AI, and that Lenovo Group's role is to provide secure and trustworthy tools and platforms to help users turn data into value.

This year marks the 10th anniversary of the founding of Lenovo Capital and Incubator Group. Yang Yuanqing said that over the past decade, Lenovo Capital and Incubator Group has invested in more than 350 outstanding technology companies, 27 of which have completed IPOs. Currently, more than 70 portfolio companies conduct business cooperation around Lenovo's hybrid AI architecture, covering multiple levels including computing infrastructure, models, and applications. In the past year alone, the amount of ecosystem collaboration between Lenovo Group and its portfolio companies exceeded 10 billion RMB. In Zhejiang, Lenovo Capital and Incubator Group has invested in more than 30 local science and technology innovation enterprises in total, including Supcon Technology, Joulwatt, Zhejiang Humanoid, Jindie Space-Time, and Saide Semiconductor, many of which have already carried out deep cooperation with Lenovo Group in areas such as joint research and development, co-creation of products, supply chain collaboration, and market expansion.

At the 2026 Lenovo Capital and Incubator Group CEO Annual Meeting, Lenovo Capital and Incubator Group signed a strategic cooperation agreement with Zhejiang Venture Capital Group.

At this CEO Annual Meeting, Lenovo Capital and Incubator Group signed a strategic cooperation agreement with Zhejiang Venture Capital Group. The two sides will jointly build a sound ecosystem in areas such as science and technology innovation investment and industrial resources, and held a joint signing ceremony with the Zhejiang Future Industry Fund, the Shanghai Future Industry Fund, Pudong Venture Capital, China Reform Fund, and Fudan Science and Technology Innovation. The Lenovo Future Fund was officially upgraded and launched, and all parties will work together to support cutting-edge scientific and technological innovation and jointly promote the commercialization of scientific and technological achievements and industrial development.

Yang Yuanqing said that Lenovo is undertaking its fifth venture, that is, the "Entrepreneurship 5.0" era of Lenovo Group. Looking to the future, Lenovo is willing to continue to be the most pragmatic partner and most steadfast fellow traveler of entrepreneurs, using a hybrid AI strategy to promote AI inclusiveness, so that AI benefits every enterprise and every person.

The following is the full text of Yang Yuanqing's speech:

I am very glad to meet everyone in beautiful Hangzhou in the late autumn season when the reed catkins first turn white and the parasol leaves are half yellow. Hangzhou is a new highland of the global digital economy. With Hangzhou as its base, Lenovo, on the one hand, deeply participates in the digital construction of Zhejiang, and on the other hand, through Lenovo Capital and Incubator Group, vigorously invests in local science and technology innovation enterprises in Zhejiang. While advancing Lenovo's business expansion, it has helped Zhejiang and Hangzhou upgrade their industries and achieved very positive results.

This year marks the 10th anniversary of the founding of Lenovo Capital and Incubator Group. At this special juncture, our cooperation with Zhejiang has also taken a new step: Lenovo Capital and Incubator Group and Zhejiang Venture Capital Group have formally reached a strategic cooperation and jointly established a science and technology innovation industry fund, focusing on cutting-edge technology and laying out the future of industry.

The theme of this summit is "Gathering Light, Walking Together, Technology Toward the New." In the new decade when AI is accelerating the reshaping of industry, how can we see the road ahead clearly, find the right direction, and seize the initiative? Here, I would like to share with you Lenovo's latest judgment on AI industry trends, as well as what we are doing and what we can provide to everyone.

To see the trends clearly, first look at what is happening in the industry. Today, agents are becoming an industry consensus. In 2026, agents appeared for the first time in the Government Work Report, which explicitly proposed "promoting the accelerated popularization of a new generation of intelligent terminals and agents." A survey released by McKinsey in August this year also confirmed this trend: among enterprises with annual revenue exceeding 1 billion US dollars, 40% have already deployed AI agents at scale, while a year ago this proportion was only 27%. This shows that AI is moving from generation to execution, and agents are becoming the core of value.

We believe that the success of AI is not only measured by how many devices are sold and how much infrastructure is built, but more importantly by whether agents are actually used and whether they continuously create value. For individuals, a successful agent must truly understand you, possess your panoramic memory, both short-term and long-term, become more and more like you, and ultimately become your AI twin. For enterprises, a successful agent must be able to cross data domains, break down application silos, and integrate enterprise data, business logic, processes, experience, and so on into an enterprise private-domain database/knowledge base, enabling people and agents to work side by side. Behind this lies the agent's understanding of a person and an enterprise. And how deep that understanding is depends on how rich and thick the data is and how deep and long the memory is.

Personal private-domain data comes from the devices, scenarios, and various applications that consumers use in daily life; enterprise private-domain data is scattered across various systems and knowledge bases within the enterprise, such as financial systems, ERP systems, customer relationship systems, supply chain systems, and human resources systems. If the value mining of public data is already in full swing, then the value mining of private-domain data has only just begun.

Personal devices and enterprise operations generate data, data trains agents, and the better agents understand users and enterprises, the more willing users and enterprises are to use them, thereby generating more data. This cycle we call the "private-domain data flywheel."

Public data is available to everyone, and models will become increasingly similar. What truly creates a gap is private-domain data that others cannot obtain and only one owns, using personal intelligence and enterprise intelligence to unleash the potential of such data. Once the private-domain data flywheel starts turning, leaders will become more and more ahead.

At present, more and more enterprises recognize that public intelligence alone cannot solve two major limitations: first, privacy and security. Neither personal data nor enterprise private-domain data can all be placed on public clouds and public AI; second, efficiency. Model parameters are growing larger, computing power consumption is growing higher, and usage costs are growing more expensive, but in fact most scenarios or applications do not need such large models. Therefore, the future direction must be hybrid: public intelligence, personal intelligence, and enterprise intelligence working with and coordinating with one another, each performing its own role and working collaboratively.

Therefore, we judge that competition in the AI industry will shift from model parameters to the implementation capability of hybrid AI. Whoever can make private-domain data truly play a role in a hybrid AI architecture will be one step ahead in the agent era. So the next battle of AI depends not only on models, but also on data and on hybrid AI.

Enterprises with a high degree of informatization have accumulated various applications, and each application has accumulated a large amount of data, but this data does not belong to those application software or SaaS software-as-a-service providers; it belongs to the users themselves. Lenovo's role is to provide secure and trustworthy tools and platforms to help users turn data into value. This is the strategic starting point for our future orientation.

Lenovo was the first to propose the hybrid AI strategy when generative AI first exploded, putting it ahead of the industry. At the time, our judgment was: AI cannot remain only in the cloud; the future will certainly be the coexistence and mutual complementarity of public intelligence, personal intelligence, and enterprise intelligence. Public intelligence provides general capabilities, while personal intelligence and enterprise intelligence meet needs such as personalization, private-domain deployment, security, and response speed. AI ultimately empowering individuals and enterprises will certainly be built on top of the intelligent capabilities individuals and enterprises already have. In other words, the future of AI is "human intelligence + AI artificial intelligence."

Today, this vision is increasingly becoming an industry consensus and is gradually translating into market results. In the previous fiscal year, Lenovo delivered the best fiscal-year results in its history; in the first quarter of the new fiscal year, we set another record for the best fiscal quarter in our history. AI-related business revenue accounted for 35% of the group's total revenue; ISG, the Infrastructure Solutions Business, became an important engine of the group's growth and profitability. Lenovo's share price hit an all-time high, and it rose to 153rd place in the Fortune Global 500, its highest ranking ever.

These results are not accidental, but the outcome of our strategic investment and planning over many years. More than a decade ago, Lenovo entered the infrastructure field through the acquisition of IBM's x86 server business, and has continued to invest since then, gradually building full-stack AI capabilities from devices and infrastructure to solutions and services. In the infrastructure field, we have also expanded from traditional enterprise infrastructure to cloud computing, then AI computing, and now we are rapidly building out AI inference computing and edge computing. These bets, which were not widely favored at the time, are now translating into real competitiveness.

Yang Yuanqing introduces Lenovo Group's hybrid AI progress at the 2026 Lenovo Capital and Incubator Group CEO Annual Meeting.

In the field of personal intelligence, Lenovo's strategy is "one body, multiple devices." AI PC is the first step, and we have already made solid progress in this area. In the third quarter of this year, Lenovo's global market share in AI PCs reached 25%, continuing to rank first worldwide; in the Chinese market, AI PC sales accounted for more than 50% of Lenovo's total laptop sales, with cumulative sales exceeding 7 million units. At the same time, we are continuing to expand new-form devices such as AI hosts, which are mainly used by agents, and AI glasses, and we ranked first in number in the first batch of national AI device tiered certification, continuing to define new standards for AI devices.

The Tianxi personal super agent has been upgraded to 4.5, moving from merely being called upon passively to being able to execute autonomously, with user activity and ecosystem scale continuing to grow. As we said earlier, the success of personal intelligence depends on how deeply the agent understands the user, and that in turn depends on personal data. Behind Tianxi's growing understanding of you is precisely Lenovo's continuous mining of data value. We are steadily moving toward the goal of "letting everyone have their own AI twin." These devices and the user base also provide a natural entry point for entrepreneurs to build personal agent applications.

In the field of enterprise intelligence, Lenovo continues to strengthen hybrid AI infrastructure based on the "device-edge-cloud" architecture. On the computing foundation, Lenovo mainly does three things: first, use devices, servers, storage, and networking to build more efficient data centers, relying on our unique Neptune liquid cooling and other technologies to help optimize the cost and energy consumption of infrastructure across its entire lifecycle; second, let different enterprise workflow loads run on different models, not necessarily all the most advanced and largest models, but load according to the application, thereby building a token factory or Token Hub so that the cost of every token is optimal; third, help enterprises build a computing architecture that combines "device-edge-cloud" and enable token generation to be rationally allocated on this architecture, because we believe that in the future 70-80% of tokens will be generated at the edge or on devices, with only 20-30% in the cloud. Such allocation can greatly optimize token costs. The performance of the past few quarters is enough to prove our achievements in this area: x86 server shipments rose to first place worldwide for the first time, and AI server order backlog is ample, providing solid support for the large-scale delivery of AI infrastructure.

At the same time, Lenovo's enterprise super agent "Lexiang," on the basis of serving Lenovo itself well, is being deployed in many industries such as manufacturing, healthcare, transportation, agriculture, and government affairs, moving from intelligent Q&A to closed-loop task execution.

The key to enterprise intelligence likewise lies in data. Let me give you two examples. Take our new-generation call center: from initially having only a chatbot answer users' questions, it has shifted to relying on technical/product knowledge bases and on existing historical Q&A and service data, with service agents independently judging problems and orchestrating end-to-end service. Some faults the agent can repair itself, while others are handed to humans; for the parts handed to humans, the agent will also automatically allocate repair parts and automatically dispatch tasks according to engineers' busy/idle status, forming a closed loop of workflow.

Another good example is the order and quotation system we have just completed development of and are actually using at Lenovo. You could say we have used AI to reconstruct the order quotation process: integrating scattered transaction, finance, marketing, supply, and customer data, enabling the order quotation agent to give effective recommendations, while also integrating processes that previously required five levels of approval. Routine product specifications, supply/production matching, credit management, and legal terms review are handled by the agent, with humans brought in only for non-routine or extraordinary requests, thereby shortening order quotation time by 85%.

These are all practices through which Lenovo mines data value in its own business. These proven capabilities can all be productized in the future to serve enterprises with the same needs.

Ten Years of Lenovo Capital: Two-Way Empowerment, Co-Creating Value

This year marks the 10th anniversary of Lenovo Capital and Incubator Group. Over the past decade, Lenovo Capital has invested in more than 350 outstanding technology companies, 27 of which have completed IPOs. Currently, more than 70 portfolio companies are conducting business cooperation around Lenovo's hybrid AI architecture, covering multiple layers including computing infrastructure, models, and applications. In the past year alone, the ecosystem collaboration amount between Lenovo Group and its portfolio companies exceeded RMB 10 billion.

In Zhejiang, we have cumulatively invested in more than 30 local science and technology innovation enterprises, including Supcon Technology, Joulwatt, Zhejiang Humanoid, Jindie Space-Time, and Saide Semiconductor. Many of them have already carried out in-depth cooperation with Lenovo in joint R&D, product co-creation, supply chain collaboration, and market development.

Take Hangzhou-based Supcon Technology, for example. It connects and governs the private-domain data that has long been "dormant" on industrial sites, and then, through a time-series large model self-developed for the process industry, turns it into reusable production knowledge, helping industrial enterprises improve efficiency and reduce energy consumption and costs. This is a typical scenario of unlocking data value.

There are many more pioneers like Supcon Technology in Lenovo Capital's ecosystem. They not only deepen their presence in niche fields but also form synergies with Lenovo. Lenovo has full-stack AI capabilities, a global supply chain, and industrial resources, while entrepreneurs have cutting-edge technologies and scenario insights. By complementing each other's strengths, the two sides can turn technology into industrial value.

Entrepreneurship 5.0, AI for All, Moving Forward Together

Since its founding in 1984, Lenovo has gone through more than four decades, with every decade feeling like a new startup. Today, like all of you, we are embarking on our fifth startup, which is Lenovo's "Entrepreneurship 5.0" era.

And the defining mark of this era is AI. Lenovo is determined to turn the ideal of hybrid AI into reality, promote AI for All, and let AI benefit every enterprise and every person. Dear entrepreneur friends, Lenovo is willing to continue to be the most pragmatic partner and the most steadfast fellow traveler of entrepreneurs. Here, I also sincerely invite everyone to visit the 2027 Lenovo Tech World held at the Sphere in Las Vegas during CES, the global consumer electronics show, on January 6 next year, to see our latest progress in hybrid AI and to envision together an intelligent future in which human wisdom and AI coexist.

Thank you all! news_keyword_pub,stock,sz300024&sh688141&sh688777 iframe below the main text iframe below the main text General statement

Sina statement: All meeting transcripts are compiled from on-site stenography and have not been reviewed by the speakers. Sina.com publishes this article for the purpose of conveying more information and does not mean that it agrees with the views or confirms the descriptions. General statement QR code at the end of the article start QR code at the end of the article start

QR code at the end of the article start QR code at the end of the article start Editor name and work code

Editor name and work code

Source

Sina Tech