Lenovo Group's leader, Yang Yuanqing, has finally reached a defining moment after 12 years of strategic navigation.
At NVIDIA's GTC conference in March this year, when Jensen Huang told Yang Yuanqing "this year will belong to you," sparking widespread online discussion, many were skeptical. However, the answer soon became clear.
On May 22, Lenovo Group delivered a remarkable fiscal 2025/26 annual report. Full-year revenue approached the 600 billion RMB mark, surging by nearly 100 billion RMB in a single year, representing a year-over-year growth rate exceeding 20%. Net profit soared by 42%, doubling the revenue growth rate. All three core businesses—Intelligent Devices, Infrastructure Solutions, and Solutions & Services—achieved double-digit growth. More notably, within the fourth fiscal quarter's revenue structure, AI-related income already accounted for nearly 40%.
This impressive performance, featuring both volume growth and structural optimization, is particularly striking against a backdrop of tariff fluctuations, geopolitical conflicts, and soaring memory chip costs. This significant "expectation gap" propelled Lenovo's stock price to a historic high following the earnings release.
Twelve years ago, Yang Yuanqing pushed through the controversial acquisition of the x86 server business, which struggled to turn a profit for a long time. Now, as the new wave of technology triggered by large AI models surges, Lenovo has precisely caught the rhythm of the AI computing power explosion. The company once labeled with a "PC tag" has completely overcome its weaknesses, transforming from a single hardware giant into a comprehensive smart ecosystem and AI powerhouse.
Recently, at Lenovo's Beijing headquarters, I conducted an in-depth interview with Yang Yuanqing, discussing Lenovo, AI, the "lobster" phenomenon, and the first small program he wrote for the company. This "old captain," who has helmed Lenovo for over 20 years, has steered the company through multiple technology industry cycles. Facing the immense wave of AI, he appears excited yet calm, urgent yet composed, humble yet ambitious. He has a clear grasp of technological evolution logic, market opportunity coordinates, corporate strategy direction, and Lenovo's own strengths and weaknesses.
For the new fiscal year starting April 1, he has set two new "military orders" for Lenovo: first, to achieve revenue exceeding $100 billion within two years, and second, to fully transform into an AI-native company.
The weight of these two goals is substantial. Annual revenue of $100 billion means Lenovo must add hundreds of billions of RMB in new revenue to its financial statements amid a cycle of rising upstream component costs and uncertain global macroeconomic conditions. The concept of an "AI-native company" requires a complete re-examination of Lenovo's organizational DNA, product logic, and workflow. For a global enterprise with over 40 years of history, more than 70,000 employees, and operations in 180 markets, the difficulty of this task is at the level of a business school textbook case.
However, those familiar with Yang Yuanqing know he is steady, pragmatic, and reserved, advocating for "doing one's best and delivering on promises," never setting goals lightly. Since joining Lenovo as a fresh graduate in 1989, his career spanning over three decades has been characterized by a work style of carefully calculating a goal, committing to it, and then delivering.
In 1994, when he took over Lenovo's Microcomputer Business Unit, over 70% of China's PC market was dominated by foreign brands like AST, Compaq, IBM, and HP, with over a hundred domestic brands also competing. Yet, within two years, Lenovo claimed the top spot in China's PC market, a position it has never relinquished. At the end of 2004, he led Lenovo's $1.75 billion acquisition of IBM's global PC business—a deal few were optimistic about, widely described as a "snake swallowing an elephant." Three years later, Lenovo announced that the former IBM PC business had become fully profitable. In 2013, Lenovo became the global PC leader. In 2014, he orchestrated the $2.9 billion acquisition of Motorola's mobile phone and IBM's x86 server businesses, the latter forming the foundation of Lenovo's current AI infrastructure business.
Looking back, each bet he made appeared overly bold at the time. In hindsight, each coincided with a pivotal turning point in trends. I was curious: where does his confidence in this $100 billion commitment come from?
The following is a transcript of the exclusive interview with Yang Yuanqing.
/01/ Behind the $100 Billion Revenue Target: Surging AI Computing Demand Reporter: The outside world has always viewed you as a relatively steady type of corporate leader. On the first day of the new fiscal year, you proposed turning Lenovo into a $100 billion revenue company within two years. In the current macro environment, this seems like a very bold goal. Why are you so confident about this? Yang Yuanqing: This is not a rash judgment. We have actually been preparing and planning for this for many years. A development goal needs to be broken down into a strategy, refined through continuous exploration and advancement, making it increasingly accurate and executable.
From Lenovo's own perspective, AI has several key elements: data, computing power, and models. Lenovo has a very solid foundation in computing. Whether it's PCs at the personal computing level, or later servers, storage, networking, data centers, and now AI computing infrastructure, these all form the foundation of computing, the cornerstone of the AI era. We have a complete portfolio of computing product technologies, allowing us to build upon this foundation. This is where our strength and capability lie.
From the external environment perspective, the wave of large AI models presents a greater development opportunity for us. Lenovo currently has three business segments—Intelligent Devices, Intelligent Infrastructure, and Solutions & Services. Intelligent Devices are the foundation for personal intelligence development, while Infrastructure and Solutions & Services are the cornerstones of the enterprise intelligence market. AI has equipped these three businesses with new growth engines, giving us more confidence in future development.
Now, everyone can feel that intelligent agent applications, represented by "raising lobsters," are stimulating even greater AI computing demand. Future intelligent devices/infrastructure must not only serve humans but also serve billions of "lobsters" working 7x24 without interruption. This will bring massive market expansion.
In the future, not only will intelligent devices be indispensable, with their market size growing larger, but the enterprise AI market will also surpass the intelligent device market. Previously, when discussing infrastructure, it mostly referred to IT infrastructure, and the overall size of this market never exceeded the terminal market—the overall PC market is about $250 billion, and mobile phones are between $400-$500 billion. The server market, which best represented infrastructure before, was only around $100 billion. But this year, the infrastructure market, including servers, has already surpassed the PC market, reaching roughly $300-$400 billion, and it's still growing, potentially exceeding the mobile phone market. These are all results of continuously growing AI demand.
One of Lenovo's strengths is never being satisfied with current achievements, constantly pursuing higher development goals. When we became the domestic PC leader, we aimed to become the global PC leader. After becoming the global PC leader, we entered the server and infrastructure fields, then expanded into solutions and services. Looking back now, without that early strategic layout, we wouldn't have the foundation for development in today's AI era. Without that early layout, Lenovo's future would truly be concerning. As the saying goes, opportunity favors the prepared. Therefore, a company's foresight and strategic vision determine its future survival and growth value. This is likely the foundation of our confidence in pursuing higher growth and greater achievements.
Reporter: During GTC 2026, Jensen Huang wrote a blog saying AI has a five-layer cake, with the third layer being AI infrastructure. Where do you see the future growth points for AI infrastructure? Yang Yuanqing: The development speed of AI infrastructure today exceeds many people's imagination, even mine as someone with over thirty years in the industry.
The infrastructure market has undergone several rounds of evolution, from traditional computing to cloud computing, and now to AI computing. AI computing itself includes two phases: training and inference.
In the future, the infrastructure market will experience a new round of growth. Currently, training of large AI models still primarily occurs in the cloud, while AI inference requires deploying computing power closer to where data is generated. As AI becomes more ubiquitous, the trend of AI inference shifting to the local and edge is becoming increasingly clear. Therefore, AI inference is one area where Lenovo will focus its efforts in the enterprise intelligence market.
Of course, the growth in AI inference demand does not mean a reduction in the scale of AI training computing power. As long as Scaling Law remains effective, the scale of AI training will also continue to grow, but the inference part will grow faster. Currently, 80% of AI infrastructure is used for training, and only 20% for inference. In the future, this ratio will reverse. Besides AI computing, traditional computing remains indispensable, and the traditional CPU-based cloud infrastructure market will also gradually expand.
If you observe the AI value chain, you'll find computing power gradually moves from centralized to distributed—first from training to inference, then from cloud to edge, and from edge to device. As computing power sinks downward, its scale expands.
/02/ The Future of AI Devices: How Many Computers Are Needed to "Raise a Lobster" Reporter: The nationwide "lobster-raising" craze that emerged earlier this year has brought unprecedented attention to AI devices. How do you view this phenomenon? Yang Yuanqing: People may feel this has an element of chance and suddenness, but I believe there is a certain inevitability to it.
It reflects a major trend: AI development is moving from the cloud to devices, from public to individual, from Q&A to action, from models to agents. Ultimately, it's about the practical application, popularization, and generalization of AI.
Lenovo actually recognized the inevitability of this trend quite early. In May last year, at our Tech World event in Shanghai, we launched Lenovo's super intelligent agent, which shares some conceptual similarities with the "lobster," including local agent deployment, device-cloud collaboration, AI computing power support, cross-device and cross-ecosystem agent coordination, establishing local knowledge bases, autonomous learning, evolution, and action, etc.
I find it interesting that people now say "raise a lobster" instead of "use a lobster." Why "raise"? Because it becomes more personalized, deepening its understanding of the individual, capable of autonomous learning, evolution, and action. The more you "raise" it, the more it understands you, and the more skills it acquires. This aligns with the concept of "Personal AI" that Lenovo has been promoting for a long time.
Lenovo has its own term for this trend: the AI Twin. What we understand as "raising a lobster" is essentially raising another "self." We feed the "lobster" all our personal data, allowing it to truly understand us and act on our behalf according to our thoughts and wishes. This is the essence of "raising a lobster."
Reporter: To "raise a lobster" into another "self," you must feed it all your data; otherwise, it won't fully understand you. An individual's electronic data is often scattered across various personal devices and leaves traces when interacting with cloud-based information and large models. Interestingly, most people currently choose to "raise their lobster" on local devices rather than in the cloud. Yang Yuanqing: That's because our most critical personal data is typically stored locally. People tend to store private personal data on their own devices rather than in the cloud, even if cloud storage is very secure—they might still prefer not to. So "raising a lobster" generally needs to be done locally.
Reporter: There's another phenomenon: many people buy an extra computer specifically to "raise a lobster." Remember when Bill Gates said "a computer on every desk and in every home"? That was for human use. Now, many people buy a computer specifically for "raising a lobster." What's your take on this? Yang Yuanqing: This phenomenon actually reflects some existing concerns about "raising lobsters." Why are people unwilling to use their primary, everyday computer for this? The answer again lies in privacy and security. For the "lobster" to act on your behalf, you must grant it many permissions, including providing various passwords, even credit card swipe permissions. A contradiction arises: insufficient authorization prevents it from working, while excessive authorization allows it to modify files, delete files, or change configurations on our computer, posing security and privacy risks. In extreme cases, it could even delete all user information—a risk no one wants to take.
Secondly, the setup and usage barrier for "lobsters" is quite high; installation and configuration are not easy. So now there are even services specializing in helping people deploy them, and services specifically for uninstalling them.
The third issue is cost. Having it perform tasks for you might be more expensive than doing them yourself, as it consumes a large number of tokens. These are the main problems currently faced by "raising lobsters."
Additionally, there's the issue of resource contention when humans and agents share the same computer. Therefore, many people are willing to use an extra computer to "raise a lobster," which indeed offers certain conveniences at this stage.
Reporter: For Lenovo, is this a significant positive, especially for the Intelligent Devices business? Yang Yuanqing: If everyone needs to buy two computers, of course, that's good news for us, meaning a larger market space. However, I don't think this is necessarily the optimal or final solution for the future development of intelligent devices.
Because using another machine to "raise a lobster" might solve the security problem, but it doesn't address the other issues.
In the future, when everyone's AI Twin, or "lobster," matures, the mode of human-computer interaction and device usage might be such that a person only needs to give instructions. Computers and computing devices would be used more by your "lobster," which would consume computing power and call upon storage, while the human interacts with it solely through natural language. This might be the ultimate form of "Personal AI" development.
This process won't happen overnight; it will be a gradual transition. The current situation is that people are still using computers and various applications like Word, PPT, or WeChat on phones, while AI capabilities begin embedding into these applications. Simultaneously, a person might use another device to run their "lobster," with the human and the "lobster" each doing their own thing peacefully. But in the future, as your "lobster" is raised better and better, the tasks you do yourself will gradually decrease, and you'll let the "lobster" take on more responsibilities.
As a provider of devices, we need to understand customer needs and think about how to meet them. For users, if one device can solve the problem, that is certainly the most convenient. Therefore, we now need to figure out how to meet the needs of "human-lobster coexistence" on a single device, allowing both to coexist safely. For the "lobster," i.e., the intelligent agent or AI Twin, we must ensure only necessary permissions are granted, avoiding over-authorization. This is an issue we need to consider now. We can't simply think that selling two devices means a larger market; we must consider how to truly meet user needs and what the future development direction is.
Reporter: Following your logic, future AI devices will mainly evolve in terms of hardware/software architecture, but won't see a massive change in market size? Yang Yuanqing: I believe understanding future AI devices starts with understanding the concept of "Personal AI."
Future AI will undoubtedly be personalized and customized. Each person will have only one "intelligence" of their own—their "lobster" or "AI Twin." On one hand, public intelligence cannot replace personal intelligence. On the other hand, it's not the case that there will be as many "personal intelligences" as there are devices—one on the phone, another on the computer. That's impossible. Because each person is unique, with only one set of characteristics, habits, preferences, and thoughts. A person's "Personal AI" must see what you see, hear what you hear, and then think what you think and act as you wish.
To achieve this, data scattered across various devices must be integrated to form a personal private database. Then, the intelligent agent learns, reasons, and evolves based on this foundation. This unique "Personal AI" can then represent the unique you and take actions on your behalf. This is why Lenovo's strategy in the "Personal AI" field is called "One Body, Multiple Devices."
Therefore, devices will remain indispensable in the future, and the market will grow larger with an increasing variety of devices. Once everyone's AI Twin matures, a single person, through human-computer interaction, can command an army of intelligent agents. For example, with today's "one-person companies," people will purchase more devices. Whether it's more PCs, more phones, tablets, wearables, or other new categories of devices. Because one more device means one more unit of productivity.
IDC recently made a prediction: although the overall PC and mobile phone markets are under pressure from component price increases affecting sales volume, AI PCs and AI phones—two device categories closely linked to AI—are expected to see growth rates exceeding 30% this year compared to 2025 sales, and achieve roughly 2.5 times growth by 2030. If we add the continuously emerging new types of devices, the incremental growth of the AI device market will be even larger.
Reporter: Hardware carries AI, and AI reconstructs hardware. Within the concept of "Personal AI," what new roles will devices play? Yang Yuanqing: I summarize it into four points.
First, devices are the touchpoints for human-AI interaction and engagement. Any AI application requires a hardware carrier to be used by us. Computers, tablets, phones, glasses, headphones, necklaces, rings—all could become hardware touchpoints between humans and AI. User input to AI and AI output to the user are both completed through these touchpoints.
Second, devices are for sensing, the place for data input and accumulation. For AI to understand you better, it must see what you see, hear what you hear, experience what you experience, and understand what you are doing. For example, articles you write or PPTs you create on a PC, photos you take, messages you reply to, or call records on your phone—all are data. There are also various wearable devices like smart glasses, necklaces, rings; these devices are data input points, or "sensing ends." I'm wearing a digital ring right now that can record some body data.
Third, devices provide computing power. Not all computing power is in the cloud; devices themselves can provide computing power. For example, on PCs or phones, even without a network, locally deployed models can run for inference. In home scenarios, there are edge computing devices capable of providing greater computing power and running larger models. In most cases, we complete tasks through device-cloud collaboration—some tasks are done on the local device, others in the cloud. The more data from various devices, the more the inference results will align with personal needs.
Fourth, devices represent trust. This is very important because devices belong to users. When devices are in users' hands, data is in users' hands, and "Personal AI" is more controlled and dominated by users themselves.
Within the "One Body, Multiple Devices" system, I believe the core is the "Body," your personal AI Twin, while devices provide these values: they are touchpoints, sensors, computing power sources, and trust anchors.
Reporter: Devices are becoming more important. If AI increasingly takes over computers, do you think the original software ecosystem can remain unchanged? Yang Yuanqing: Definitely not. AI will disrupt many things, and software is first in line. Whether in the form of licensed copies or SaaS, it will be changed, or at least easily changed. Accompanying this is the reshaping of enterprise business processes.
For example, if you need to make a plan today for "visiting a customer in Shanghai tomorrow," you might need to pull work schedules from a calendar tool, customer information data from a CRM, past transactions from an ERP, and then integrate the materials somehow. But the future process is: you can directly converse with the model using the data accumulated by these applications to directly generate what you need—be it a document or a presentation deck. In extreme cases, AI could rewrite a CRM for you or directly generate a functional software.
Of course, not all needs require rewriting software from scratch; imagine the token consumption. The future scenario might involve interactions between intelligent agents, with them redistributing responsibilities among themselves, which would be more efficient.
/03/ Lenovo's Moat Reporter: If the judgment about the future development of the device and infrastructure markets is from the demand side, then from the supply-side competitive perspective, what is Lenovo's moat? Yang Yuanqing: Lenovo has two core competitive strengths: persistent technological innovation and highly efficient, resilient supply chain management. These are the cornerstones that have enabled Lenovo's sustainable development over the years and constitute its moat.
First, the supply chain. Cyclical fluctuations in global manufacturing, price volatility of components, and changes in logistics rhythms are daily challenges for a multinational manufacturing enterprise. In such an environment, the ability to manage the supply chain becomes particularly crucial. Over the past year, despite multiple pressures including tariff challenges, tight supply and rising costs of key components, and geopolitical crises, our performance still reached new highs. This relies on the highly efficient and resilient global supply chain operational capabilities Lenovo has accumulated over the years. This includes our "China + N" global manufacturing footprint and "global resources/local delivery" business model, our globally leading intelligent operational management capabilities, and the deep mutual trust and mutually beneficial ecosystem relationships we have with upstream and downstream partners.
"The industry's strongest supply chain" is not just a slogan; it's a capability system for "remaining calm in change and advancing steadily." We believe that regardless of industry cycle changes, this capability system can continuously deliver value.
Next, the foundation in hardware innovation. The future trend of AI development is that hardware's importance is increasing. Agentic AI won't be the endgame of AI; next will come Physical AI and embodied intelligence. The further AI moves into the physical world, the deeper its reliance on hardware and the greater the need for understanding hardware. Lenovo has decades of accumulation in the hardware field, from PCs, phones, tablets to servers, data centers, and the global supply chain. In the AI era, the importance of these capabilities will become even more apparent, and I believe the space for unleashing our advantages will be greater.
Reporter: Lenovo operates in both consumer AI and enterprise AI, spanning the ToC and ToB markets. Could this lead to neglecting one for the other? Yang Yuanqing: I actually think this is one of our advantages. ToB enterprise intelligence and ToC personal intelligence actually share many commonalities. At their essence, both are about using private domain data, applying suitable combinations of models and computing power, to enhance efficiency in an augmented manner.
Many capabilities of personal intelligence and enterprise intelligence are intertwined. For example, the computing power used by personal intelligence doesn't only come from personal devices; it also requires "device-edge-cloud" collaboration. Here, the "edge" for an individual might be larger-scale storage or higher-computing-power devices in the home, used to run larger-scale models to support personal devices. In the future, there will also be personal clouds providing computing power support. This is why we now emphasize hybrid artificial intelligence. Even in personal intelligence scenarios, public intelligence is needed.
The same applies on the enterprise side. An enterprise isn't just a data center or edge computing; it's also composed of individual employees, each using devices. So enterprises also need to consider that after models are trained in the cloud, they can be loaded onto the edge or even deployed to devices, allowing employees to complete task processing on the device without relying entirely on the cloud. This would be a more efficient, secure, and reliable way to implement AI.
The difference between the two lies more in the delivery method. Personal intelligence is more often delivered as a product, ready for users to use out-of-the-box, which is an area Lenovo excels at. Enterprise intelligence is delivered as projects, starting from "what do you want to do," and then helping the customer achieve their goals around the project. In this regard, Lenovo also has a complete set of technology and service frameworks as our project delivery execution system. This framework includes the computing power foundation, the ability to integrate enterprise data, an Agent generation factory, different skill modules... all integrated to form solutions for delivering "enterprise intelligence" to customers.
From a technology perspective, AI model training in the cloud and AI inference on the edge and devices influence and interact with each other. From a market perspective, the three growth engines—incremental device sales, incremental infrastructure sales, and incremental service revenue—are igniting successively, taking turns to provide momentum, which is unprecedented in Lenovo's history. This is the most straightforward arithmetic behind Lenovo's $100 billion target.
/04/ From Code 30 Years Ago to AI-Native Today Reporter: At the pledge meeting for this fiscal year, you also mentioned that Lenovo aims to become an AI-native enterprise. Lenovo is a large technology company with over 40 years of history and annual revenue potentially reaching around $80 billion. I want to ask, how does Lenovo understand "AI-native"? Yang Yuanqing: AI-native means not treating AI as an add-on or supplement, but rather rethinking, constructing, and operating a company based on AI. In other words: AI First. This process is inevitably challenging, but such a restructuring process is not unfamiliar to Lenovo.
Reporter: What were the previous restructuring processes like? Can you take us back? Yang Yuanqing: For example, the development process of Lenovo's IT information systems could be described in today's terms as "informationization-native."
Over thirty years ago, when I first joined Lenovo after graduate school, I worked in sales for workstations and HP peripherals. Every week, the finance department would give me a form listing available products and their quantities. But sometimes, when a customer went to finance to pay and then to the warehouse to pick up the goods, they found nothing available. Why? Because the form said there were five items, but someone else had already sold five, and the form hadn't been updated, so the sixth person arrived to find no stock.
I felt this made doing business very difficult. As a computer science major, I programmed a software myself to record basic "inventory-sales" data. This way, if the system showed stock when a customer paid at finance, they would definitely find goods at the warehouse. That was the earliest attempt.
Later, as enterprise informatization developed, we began building systems like ERP and CRM. However, these systems were built for different purposes and objectives, and after completion, they became information silos, unable to connect with each other, leading to operational inefficiencies.
Reporter: But restructuring these traditional IT assets is often very difficult. Yang Yuanqing: True. But in the AI era, a single model can rewrite all software and can re-engineer an enterprise's business processes. Both information flow and business flow could be restructured. If we remain stuck in fixed thinking, doing things the old way, we will inevitably be eliminated by more efficient, AI-native enterprises. So we say we must use AI thinking to re-examine the enterprise's business processes, work methods, organizational structure, as well as job roles and talent systems.
Reporter: For a company like Lenovo with over 70,000 people and operations in 180 markets, the difficulty of becoming "AI-native" is completely on a different scale compared to a startup. Yang Yuanqing: It's certainly not easy. For a company of 70,000 to pivot means re-examining every product, every business process, and every job responsibility. But once this pivot is completed, the efficiency it unleashes will be enormous. Lenovo currently has a business scale exceeding 500 billion RMB and a full-stack AI business portfolio, which is the crystallization of over forty years of technological innovation,卓越运营, customer trust, and partnerships. This "gold mine" is the springboard for reaching new revenue heights.
Reporter: Finally, a question of general concern. You wrote Lenovo's first business IT program. In an era where software is being rewritten by AI, many colleagues feel anxious. As a first-generation programmer, what's your perspective? Yang Yuanqing: I believe simple coding work will definitely be replaced by AI. But the more important value of a programmer lies in planning and design. If no one defines what needs to be done, what the direction is, or how the process should be designed, AI doesn't know what to do.
Of course, AI is also constantly advancing. What we consider must be done by humans today might also be partially replaced by AI in the future. But human roles will also continuously move upward, taking on higher-level tasks. So we must keep learning and must develop more creative thinking.
Comments