"Yang Yuanqing's Bold Investment in Super Intelligent Agents"

Yang Yuanqing, born in 1964, is the Chairman and CEO of Lenovo Group, and will reach the milestone of 60 years old in 2024.

While many people might retire at this age, Yang Yuanqing remains as vibrant as a seasoned warrior returning from numerous battles. This time, the typically cautious executive is prepared to fully commit to the realm of super intelligent agents.

On May 7, 2025, Lenovo Group hosted its annual Innovation Technology Conference at the Shanghai World Expo Hall. Yang Yuanqing energetically jogged onto the stage, showcasing his vision of "super intelligence" to the world.

He categorizes super intelligent agents into three distinct levels:

1. Perception and Interaction: Utilizing text, speech, gestures, eye-tracking, and other modalities, these agents capture real-time user status and environmental data, enabling precise interpretation and responsive action across various devices.

2. Cognition and Decision-Making: Super intelligent agents access personal and corporate knowledge bases across devices and ecosystems. Through knowledge accumulation, experiential reflection, real-time learning, and efficient feedback, they establish a traceable, correlated long-term memory that integrates endpoints, edges, and clouds. This allows them to uncover logical relationships, comprehend complex intentions, and apply learning from one scenario to another.

3. Autonomy and Evolution: This represents the ultimate goal for super intelligent agents. By integrating user intent with prior knowledge, they decompose complex tasks into sub-tasks and assign them to relevant domain agents. They autonomously orchestrate execution sequences, plan steps, invoke cross-ecosystem tools based on agent interconnection protocols, and evolve continuously through self-learning. Critically, they optimize decision-making logic via self-learning.

In essence, super intelligent agents enhance human learning, entertainment, and interaction efficiency, bolster decision-support capabilities, and evolve through autonomous learning.

As Lenovo's CEO, Yang Yuanqing's objective in promoting super intelligent agents is to drive hardware sales. At this conference, Lenovo unveiled a range of hardware products, spanning laptops, tablets, and smartphones (see image below). With slim profit margins in PCs and other hardware, advancements in software (operating systems), chips, and applications propel hardware upgrades, boosting Lenovo's sales and profits.

Due to the steep decline in large model inference costs and the rapid AI adoption (epitomized by the emergence of DP), China has entered the era of embodied intelligence. From traditional consumer electronics to appliances, automobiles, and even drones, AI transforms them into diverse intelligent agents. Lenovo's super intelligent agents serve as the systemic foundation for Lenovo's markets in individuals (Tianxi), enterprises (Lexiang), and governments (cities).

For humans to seamlessly utilize super intelligent agents (or large models like OpenAI and DeepSeek) and enable their autonomous learning and evolution, powerful computing servers or terminals are essential. This will foster widespread hardware upgrades and replacements, significantly boosting Lenovo's performance and profits, while also helping it capture market share from competitors lacking super intelligent agents, solidifying its position as the global PC leader (AI PC).

A Deja Vu Return

This scenario likely evokes a sense of deja vu for Yang Yuanqing.

Thirty years ago (1994), Yang Yuanqing, who had been with Lenovo for five years, opted to stay and lead Lenovo's proprietary brand division, foregoing an overseas study opportunity. At that time, the PC market was poised for explosion:

Firstly, the user-friendly Windows 95 operating system with a graphical user interface (GUI) was about to be launched (bundled with a sales strategy), drastically lowering the barrier for ordinary users to adopt PCs (originally requiring programming skills), thereby accelerating PC popularization. (Now, Yang Yuanqing defines super intelligent agents as 'cognitive operating systems,' enabling users to interact and operate via voice and other means, fully liberating humanity.)

Secondly, the massive PC adoption made previously expensive CPUs more affordable, further reducing PC prices and fostering a virtuous cycle. (With AI embedded in various intelligent agents and becoming a standard configuration, it will inevitably drive down GPC prices and reduce computing costs, accelerating intelligent agent adoption.)

Thirdly, the advent of email, portals, and other services allowed people to browse vast amounts of information and communicate online via PCs, expanding PC application scenarios. (Currently, AI+terminals lack application scenarios and popular apps akin to the PC era, slowing down their adoption.)

From 1989 to 1994, Yang Yuanqing served as an agent for international computer brands like HP and Compaq. He continuously learned from these brands while making localized improvements. For instance, HP and Compaq generally adopted a national general agency approach for distribution, whereas Lenovo PCs sought cooperation with numerous small distributors across provinces and cities, creating a denser and broader channel network.

International brands like HP typically sold previous-generation products at exorbitant prices in China (exceeding RMB 20,000 to 30,000). Yang Yuanqing and his capable assistant Liu Jun (currently Executive Vice President and President of Lenovo China) foresaw Moore's Law early on, understanding that with chip and operating system iterations, PCs updated and replaced rapidly, and costs declined swiftly with increasing sales volumes. Hence, they adopted a strategy of small but frequent updates, designing the Lenovo E series with Intel's latest CPUs and Microsoft's latest operating systems, priced at RMB 10,000, maintaining low inventory (no pressure to sell new-generation products), and becoming the market leader in China in 1997.

Since 2017, Lenovo Group has been promoting an AI strategy, awaiting a market inflection point akin to 1995.

In 2023-2024, global AI (large models) witnessed significant advancements, but China lagged due to insufficient supply (or even a cutoff) of NVIDIA GPUs, consistently trailing in the competition with large models like OpenAI. In early 2025, the independently developed DeepSeek released its fully free R1 inference model, drastically reducing inference costs and enhancing model capabilities. Subsequently, Tongyi Qianwen, Tencent Yuanbao, Baidu Wenxin, and others launched free large models, with companies like Lenovo gaining access.

At this conference, Yang Yuanqing even unveiled an inference acceleration engine – a joint research and development achievement of Lenovo, Tsinghua University, and Wuwen Xinqiong, aimed at addressing the industry's core pain points of lightweight large models and improved inference efficiency.

It is reported that in February 2025, Wuwen Xinqiong completed the adaptation and optimization of DeepSeek-R1 and V3 on seven hardware platforms, including Biren, Haiguang, and Moore Threads. Lenovo and others can access DeepSeek series models and diverse heterogeneous domestic computing power services through the Infini-AI heterogeneous cloud platform.

In other words, Lenovo's super intelligent agents, leveraging Wuwen Xinqiong, can call upon the most suitable computing resources based on user-end needs. For instance, solving basic math problems doesn't necessitate a billion-parameter large model and extensive computing power, whereas enabling a large enterprise to train its own small model based on existing data may require the most advanced algorithms and computing power.

Aiming for the Productivity Battlefield

'In the future, we will be an enterprise centered around intelligent agent business. It's exhilarating, and I sincerely hope we can realize this vision, enhancing everyone's life and boosting enterprise efficiency.' Yang Yuanqing revealed in a media interview post-conference.

Yang Yuanqing's ability to envision and implement this stems from the 'insights' garnered from years of experience and setbacks.

Yang Yuanqing became Lenovo Group's second-generation leader by making PCs number one in China. He laid the groundwork for the smartphone business as early as 2010. Leveraging its brand and channel advantages, Lenovo collaborated with major operators and became the fourth-largest global mobile phone market share and the second-largest in China within four years, alongside ZTE, Huawei, and Coolpad, known as 'Zhonghua Kukuan'. In 2014, Lenovo even acquired Motorola Mobile. The 2024-2025 fiscal Q3 report shows that Lenovo Group's smartphone business achieved revenue growth of 155% and 28% in the Asia-Pacific and Europe-Middle East-Africa markets, respectively, excluding North America and Latin America. Without Lenovo's continuous exploration and investment in its mobile business, it would have a gap today in constructing an intelligent agent ecosystem spanning from PCs, tablets, to mobile phones.

Historically, Yang Yuanqing internally ventured into FM365, which later proved unsuccessful. This also reflects Lenovo's hardware-selling and government/enterprise customer-serving genes. The 2024-2025 fiscal Q3 report shows that revenue from the Intelligent Devices Group (IDG) was RMB 99.1 billion, a year-on-year increase of 12%. Lenovo's global market share in the personal computer business rose to 24.3%, nearly 5 percentage points ahead of the second place. The share of AI PCs in notebook sales in the Chinese market reached 15% ahead of schedule. Revenue from the Infrastructure Solutions Group (ISG) surged nearly 60% year-on-year to RMB 28.3 billion. The AI server (dominated by X86 servers) business has entered a stable revenue period by assisting more enterprises in building vertical large models. Revenue from the Services Business Group (SSG) reached RMB 16.2 billion, a year-on-year increase of 12%, with an operating profit margin maintained above 20%.

The super intelligent agents launched by Lenovo Group this time target Lenovo's three core business groups, accelerating the update of AIPC for individuals, enabling enterprises to expedite the construction of their own vertical large models (requiring AI server purchases), and allowing cities to build urban intelligent agents by integrating various large models and product services.

'In the longer term, I aspire to let AI compete in productivity, while humans excel in creativity.' Yang Yuanqing's 60-year-old bold investment in super intelligent agents is aimed at the productivity (2B) battlefield with the largest budget and where Lenovo has the most advantages, rather than the 2C market for individuals like Tongyi Qianwen, Tencent Yuanbao, and ByteDance Doubao.