Huawei Steps Up Its Efforts, Lenovo Accelerates: Can AI PC Usher in an "iPhone" Moment?

Recently, Lenovo and Huawei have successively launched new AI terminal products, further integrating intelligent agents with PCs.

Among them, Lenovo unveiled its Tianxi Personal Super Intelligent Agent, along with AI PCs, AI smartphones, and AI tablets equipped with the Tianxi Personal Super Intelligent Agent. Huawei's first HarmonyOS-powered computer also integrated the Xiao Yi Intelligent Agent, marking the first time Huawei's AI assistant, Xiao Yi Intelligent Agent, has been available on a computer.

Since 2023, chip producers and terminal brands have been actively positioning themselves in the AI PC market. With frequent moves in the industry, AI PCs are entering their "iPhone" moment. However, consumer acceptance of existing products remains mediocre.

Several industry insiders told the China Securities Journal that the main challenges for AI PCs lie in the upstream supply chain, including memory and chips, with optimization still needed in cost control and scalability. It is estimated that the industry chain will take two to three years to gradually mature.

▍Real Challenges in the Supply Chain

Li Weichang, General Manager of Lenovo China's Consumer PC and Tablet Business Unit, said in an interview with the China Securities Journal that, unlike the overseas standardized path led by Microsoft, the domestic AI PC ecosystem is fragmented, resulting in a diluted concept of AI PC.

Li believes there is significant confusion in the market regarding the concept. "Some producers market NPU chips as a selling point, but many consumers who buy these so-called AI PCs find they lack AI functionality and do not improve efficiency," he said.

There are also many practical bottlenecks in the industry chain, preventing AI PCs from truly reaching a stage of "natural progression." "For example, the upstream supply chain has not yet achieved cost control and scalability optimization, making it impossible to ensure a good user experience with smaller memory. In addition, the chips currently used by all producers are basically high-end chips, which are expensive and equipped with NPU modules capable of AI acceleration," Li said.

Li used memory as an example, saying, "Large models like 7B require high memory, while the current mainstream is 16GB memory machines. The needs of general users are mainly for simple AI Q&A functions. Installing a 7B model on a device may affect its speed."

However, Li is optimistic about the industry's development. With the scalability of the industry and technological upgrades, such as 32GB memory gradually becoming the market mainstream, the issues will be alleviated. At the same time, model distillation technology is also advancing, aiming to make models smaller and more capable.

▍Low Market Penetration but Rapid Growth

Commenting on the current state of AI PCs, Xu Yue, an analyst at IDC China, believes that AI PCs are in the early stages of development, with low market penetration but rapid growth.

Xu Yue stated that early AI PCs were mainly optimized based on traditional CPUs and GPUs to support some basic AI tasks. Nowadays, PC suppliers have introduced neural processing units (NPUs) specifically designed for AI, and their computing power has begun to rapidly upgrade to over 40 TOPS. These NPUs can run AI tasks more efficiently.

According to Lenovo's internal data, the current activation rate of AI features in high-end models is 20-30%, with some models exceeding 30% and still rising. This indicates that the acceptance among mid-to-high-end users is increasing. In addition, Lenovo is also organizing AIGC campus creation competitions for college students and offering preferential policies to encourage more young people to embrace the AI era.

Li Weichang revealed to a reporter from the Science and Technology Innovation Board Daily that Yang Yuanqing, Chairman and CEO of Lenovo Group, attaches great importance to the AI business. "He personally reviews the overall activity of AI PCs in the China region every week and monitors the trend of weekly active rates. Once a decline in activity is detected, such as a decrease instead of an increase, he will ask us to analyze the specific reasons."

▍When will native AI agent devices emerge?

The integration of AI agents with PCs is becoming a key exploration direction for technology producers.

Recently, Lenovo released multiple super AI agents, including the Tianxi Super Personal AI Agent and the Lexiang Enterprise Super AI Agent. IBM also launched a tool platform on the same day that can build AI agents within five minutes.

Subsequently, Huawei announced that the Xiaoyi AI agent would make its debut on HarmonyOS PCs, integrating the Pangu large model and DeepSeek capabilities.

Li Weichang judged that the value of AI agents as an entry point for human-computer interaction is becoming increasingly prominent. "In the future, we may have multiple AI agents with different functions on a single device, each responsible for different tasks, just like how OpenAI encourages developers to build various 'small robots' or 'small agents' that can freely combine and collaborate on the platform."

This transformation may reshape the business models of the traditional internet era—shifting from an application-centric approach to an agent-centric one.

Li Weichang stated that users may no longer need to download a large number of apps but can complete tasks by invoking different functional agents. Hardware producers will no longer be mere device providers but will become the optimal carriers for these agents to run on.

Currently, the exploration of AI implementation has extended from PCs and mobile phones to lightweight devices such as smartwatches and glasses. However, Li Weichang admitted, "These devices still have inherent limitations, such as communication distance, computing power bottlenecks, weight, and size issues.Therefore, a true "native AI device" is still in the exploratory phase, and a clear product form has yet to emerge. This resembles the evolutionary process of a new species, requiring joint exploration by the industry."

▍AI PC Industry Chain Expected to Mature in 2-3 Years

According to Canalys data, AI PC shipments in mainland China reached 5.8 million units in 2024, accounting for 15% of the overall PC market. Canalys believes this trend is expected to accelerate, with AI PCs projected to account for 34% of PC shipments in the Greater China region by 2025.

"Drawing on the patterns of previous product upgrade cycles, the entire industry chain is expected to mature in approximately 2-3 years," he further explained. "As upstream costs are optimized, particularly with the decline in chip prices, MP-level chips will demonstrate better cost-effectiveness, creating favorable conditions for the widespread adoption of AI PCs."

Xu Yue also believes that with the continuous maturation of technology and reduction in costs, the market penetration rate of AI PCs is expected to continue to rise rapidly. "Chip technology will continue to make breakthroughs, with the performance of NPUs continuously improving. As more AI PCs with computing power exceeding 40 TOPS become available, they will further meet the demands for local large model deployment and more complex AI applications.

At this stage, the application scenarios of AI PCs are mainly concentrated in content creation, smart office, etc., with the potential to expand into diversified scenarios such as education, travel, entertainment, and health in the future.

In addition, localized deployment is the future direction. ""Due to the high operational costs, data privacy, and security risks associated with cloud-based AI deployment, deploying AI technology from the cloud to localized PCs has become an important trend." Xu Yue said.