The opening of CES is imminent, and chip companies are racing on the AI PC track

The 2025 International Consumer Electronics Show (hereinafter referred to as CES) will be held in Las Vegas, but chip companies such as Intel, AMD, and Qualcomm have taken the lead in launching new end-side AI chip products for personal computer PCs.

With the support of PC hardware innovation, large AI models that originally ran on cloud servers are expected to be migrated to PC on-premise on a large scale, and AI applications on PCs will be enriched. The hardware innovation brought by the device-side AI chip will bring a new PC application and interactive experience, and make the entire PC industry glow with a "second spring". It is understood that unlike the previous PC main chip, the main chip of AI PC not only integrates CPU and GPU, but also incorporates NPU into the center, so as to better support the local operation of AI applications on the PC.

The above-mentioned chip companies have preemptively held their own new product launch conferences before the start of CES to seize the future incremental market of AI PCs. These newly announced chip products will enter the AI PCs of OEMs such as HP, Dell, and Lenovo in the first and second quarters of 2025 and be sold to consumers.

Race on an AI PC

On January 6, Intel held a press conference and announced that the second-generation Core Ultra processors applied to AI PCs, including 200V, 200HX, 200H, 200U, etc., are aimed at different positioning notebook AI chip series. These chips all integrate Intel's CPUs, GPUs, and NPUs to provide the hardware foundation for running generative AI applications on laptops. At the press conference, Michelle Johnston Holthaus, interim co-CEO of Intel, announced that the first Intel 18A process chip, the Panther Lake processor, will be released in the second half of 2025.

The PC market has developed since 20 years ago, with x86 architecture chips dominating, and Intel chips dominating the x86 architecture PC chip market. According to Mercury Research data, although the market share has fallen in recent years, Intel has occupied more than 70% of the market share in the x86 PC market all year round, while AMD has a share of about 25%.

"At the moment, Intel is in a relative lead in AI PCs." Xing Muzi, assistant vice president of the semiconductor group of Yunxiu Capital, told reporters. According to the research report of iResearch, Intel's AI PC main chip Ultra series has entered the AI PC products of major manufacturers at home and abroad, such as Microsoft, Dell, HP, and Huawei. However, in the fast-growing market of AI PCs, the future development pattern is still difficult to say, Intel is not short of strong competitors, at the opening of CES, AMD and Qualcomm have launched the latest main chip products for AI PCs.

On the same day, AMD announced the Ryzon AI Max, AI 300 and 200 series processors, all of which are SoCs for AI PC scenarios. AMD claimed at the press conference that the chips in the Ryzon AI Max series can run 70 billion parameters of large language models on the computer. In addition, Qualcomm also released a new product for AI PCs, the Snapdragon X platform, which is different from Intel and AMD's x86 architecture, Qualcomm uses the ARM architecture. According to Qualcomm employees told reporters, this is the fourth platform of Qualcomm's X series for PCs, which is supplied to PC terminals priced at $600.

"AI PC shipments have grown rapidly over the past year and are expected to exceed 40 million units for the whole of 2024. In order to better meet the needs of users running device-side AI models and improve the multimedia content and business office processing capabilities of PCs, the hardware performance of AI PCs is also constantly improving, among which the NPU computing power is upgraded to more than 40TOPS, and the storage capacity is generally increased. Xing Muzi told reporters.

In addition to the above three companies showing new products during CES, Nvidia was previously rumored to announce CPU products for AI PCs at CES, but at today's conference, Nvidia CEO Jensen Huang did not show such products. "The newly announced NVIDIA RTX50 series products are independent GPUs, and some AI PCs will not only be equipped with the main chip, but also with independent GPUs to further improve the AI processing capabilities on the device side." Xing Muzi told reporters. For example, some Dell sub-brands, such as Alien and Lingyue's high-end gaming laptops and high-end business laptops, are equipped with Intel's Ultra series main chips and NVIDIA RTX40 series discrete graphics cards, with prices of up to RMB 40,000.

Many of the questions are expected to be solved

Compared with traditional PCs, AI PCs are expected to bring users a new operating experience. According to iResearch, compared with users interacting through mouse, keyboard, touch screen, etc., AI PC is expected to allow users to interact through gestures, eye tracking, voice interaction, etc., bringing a new user experience. Once AI applications such as personal agents are widely used on AI PCs, AI PCs may become a new traffic portal. However, in order to achieve a satisfactory user experience, manufacturers must work computing power, power consumption, and cost to win the recognition of users.

To achieve a natural and smooth new user experience, high requirements for end-side computing power are indispensable. Intel, AMD, and Qualcomm are focusing on increasing the computing power of the main chips, such as the newly launched Intel Core Ultra 200H series processors, whose GPUs, CPUs, and NPUs can deliver up to 99 TOPS of computing power.

On the other hand, running chips with low power consumption and achieving long-term battery life of AI PCs are also the goals pursued by these chip manufacturers. For example, AMD claimed at the launch event that its Ryzon AI 300 series can achieve more than 24 hours of battery life. Counterpoint believes that how to provide sufficient computing power under limited power consumption and achieve a balance between computing power and power consumption has become the core competitive factor of a manufacturer.

In addition to improving computing power and reducing power consumption, how to reduce costs and then reduce prices has also become an "examination question" in front of these chip manufacturers. Due to the high cost of the above-mentioned end-side AI chips, the product price of the entire AI PC has been directly pushed up. At present, the AI PC on the market is mainly for high-end business people, electronic enthusiasts and game enthusiasts, and these products are often expensive, such as the price of AI PC with alien brands ranging from 20,000 ~ 40,000 yuan. However, Intel, Qualcomm and other manufacturers have announced products for mainstream PC product positioning at the press conference, and the products are available for PC products priced at 4,000~8,000 yuan, and these PC products are expected to be shipped in the first half of this year.

According to Canalys, global AI PC shipments will grow at a compound growth rate of 44% from 2024 to 2028, with laptops accounting for the vast majority of the share and desktops accounting for a small share. While achieving a high compound growth rate, technical difficulties need to be further solved. "With the increase in the penetration rate of AI PCs, the computing power, data transmission speed, and power consumption of AI chips on the device side need to be further improved, and more powerful chip computing performance and better memory bandwidth are needed to win the favor of more users." Xing Muzi told reporters.

(This article is from Yicai)