As the application scenarios continue to expand, the market size and penetration of AIPC will accelerate in the next few years.

after the great breakthrough in AI (artificial intelligence) technology, the industry is all thinking about how to land and realize it. AIPC, the most likely "AI inclusive preferred terminal", has attracted tech giants.

on January 17, Lenovo released a number of AIPC new products such as Lenovo saver and Lenovo xiaoxin. Earlier, on January 4, Microsoft announced that the new notebook computers and PC will be equipped with Copilot keys, which will call out Microsoft's built-in AI assistant Windows Copilot. Intel has also launched the "AIPC Acceleration Plan" and plans to implement artificial intelligence features for more than 0.1 billion PCs by 2025...

IDC predicts that China's AIPC will account for more than 50% of new machine shipments in 2024, and AIPC is expected to usher in the first year of development. Canalys predicts that 60% of PCs will have artificial intelligence functions by 2027. Boston Consulting even predicts that AIPC will account for 80% of the PC market by 2028.

As a new product of PC reform in the AI era, what is the difference between AIPC and traditional PC? What opportunities will it bring to the industrial chain?

The difference between AIPC and traditional PC

according to the white paper on AIPC industry, AIPC has five characteristics: embedded personal model and personalized local knowledge base; With CPU, GPU, NPU local hybrid AI computing power; Have an open AI application ecology; Can use natural language interaction; Have device-level personal data and privacy security protection.

Compared with traditional PCs, in terms of hardware, AIPC integrates hybrid AI computing units on the basis of traditional PC functions to support complex artificial intelligence algorithms and a large amount of data processing. AIPC is usually equipped with high-performance processors, can support heterogeneous computing including CPU, GPU, NPU, the overall computing power support ability is stronger, can run the "personal big model" locally, create a personalized local knowledge base, to achieve natural language human-computer interaction.

It is reported that NPU (neural network processing unit) is a processor or chip specially designed to perform artificial neural network calculations. It is widely used to accelerate artificial intelligence tasks, especially deep learning and machine learning algorithms. NPU is more applied to end-side AI in artificial intelligence landing applications, and will form the core computing power of AIPC together with traditional CPU and GPU, which has also become one of its selling points.

from the perspective of interaction, AIPC improves the convenience and fluency of interaction with users through voice, gesture and other means, reduces the user's use threshold, and changes from the original use tool to a personal AI assistant tailored for everyone.

From the application point of view, AIPC application ecology is more open, can match the pathological analysis, intelligent manufacturing and other high-power, high-intelligence scene use needs.

From a security point of view, AIPC can effectively ensure data security by pooling technical support such as personalized local knowledge bases, local reasoning mechanisms, and hardware-level security chips to deal with the huge amount of data generated when applying large models.

AIPC is expected to drive the C- side consumption boom

"AIPC will be the most important 'first terminal' to promote the landing of large models '." Yang Tianyu, an analyst at Sadie Consulting's Electronic Information and Artificial Intelligence Industry Research Center, told First Finance and Economics that from the perspective of computing capacity, PC is larger than other terminal products such as mobile phones and smart watches, can be equipped with more and stronger processors, memory and other components, and has the ability to support the operation of large models with tens of billions of parameters. From the perspective of computing power application scenarios, PC has a wider application field than other terminal products, it can carry multiple scenes such as entertainment, life and work, and is conducive to promoting the application and promotion of large end-side models. From the perspective of human-computer interaction, PC can support full-mode human-computer interaction, including accurate and professional keyboard and mouse operation mode, as well as intuitive and convenient touch, voice, gesture and other operation modes, which can support large models to better adapt to user needs.

Chen Shuxin, senior research manager at IDC China, holds a similar view. "PCs do have a lot of advantages in landing large AI models. In terms of computing power, storage capacity, multi-modal interaction capabilities, and the most comprehensive multi-scenario applications, PCs have a high probability of becoming the preferred terminal for AI Pratt & Whitney."

As more and more AIPC products enter the market and application scenarios continue to expand, the size and penetration of the AIPC market will accelerate in the next few years.

CCID Consulting expects AIPC penetration to approach 30% by 2025. IDC is more optimistic, expecting AIPC penetration in China to reach 55%/75%/80%/85% respectively in 2024-2027.

At the same time, AIPC is also expected to drive a new wave of rotating machines.

the mainstream replacement cycle in the traditional PC consumer market is 3-5 years, and only about 10% of users have plans to replace PCs within two years in 2022. Driven by AIPC, IDC expects that the number of users who change planes will rise to 20% or more within two years, and the tide of changing planes is expected to advance, while the consumer group is developing towards high-end.

"AIPC can meet the personalized service needs of users, including drawing, writing, etc., and will shorten the user's replacement cycle. Considering the wait-and-see time of some conservative users, it is expected that AIPC is expected to leverage the C- end consumer market on a large scale around 2025." Yang Tianyu told First Financial.

AIPC industry chain to open the imagination space

The increase in AIPC penetration has also brought more imagination to the industry chain.

Yang Tianyu said that from the perspective of the industrial chain, AIPC's most investment value and most noteworthy areas include upstream chip manufacturing and midstream complete machine manufacturing, especially upstream chip manufacturing companies whose main businesses are CPU, GPU, NPU, storage, etc.

in the CPU field, major domestic companies include Huawei, great wall of China (000066.SZ) and haiguang information (688041.SH);GPU field mainly includes Huawei, Cambrian (688256.SH), Haiguang Information (688041.SH), Bi Ren Technology, Yuan Sui Technology, Moore Thread, etc.

The NPU field mainly includes Huawei and Ruixin Micro (603893.SH), which have formed a series of SoC chip products with built-in self-developed NPU, mainly used in end-side and edge computing. Guoke Micro (300672.SZ) said its GK76 series IPC chip integrated self-research NPU engine, which can be used in the cloud to achieve some AI tasks on the edge side. NPU is entirely derived from self-research and focuses on edge computing.

in terms of upstream IP, core original shares (688521.SH) said that in the AIoT field, its neural network processor IP(NPU) for artificial intelligence is the industry leader and has been used by 68 customers in its 120 surplus artificial intelligence chips.

At the same time, the memory upgrade to DDR5, its penetration rate is expected to follow the AIPC penetration rate increase, driving the memory interface chip market. Lanqi Technology (688008.SH) has the third largest market share in the field of DDR5 memory interface chips in the world and the first in China. Juchen Stock (688123.SH) and Lanqi Technology have jointly developed SPD products supporting a new generation of DDR5 memory chips, with built-in SPD EEPROM, which is used to store relevant information of memory modules and all configuration parameters of memory particles and related devices on the modules.

In addition, AIPC, as a high-end product, will gradually increase the volume and price of components such as heat dissipation and structural parts.