The AI Era: The Future War of PC Giants

Recently, the AI PC hardware has already possessed AI acceleration capability, but the moves of terminal manufacturers in leveraging the advantage of computing power and deploying local large models are still unclear.

In the exploration process of AI PC, Lenovo has taken the lead in defining the concept of edge-side models and personal intelligent agents. Huawei continues its consistent strategy of building an ecosystem, emphasizing the interconnection of all things and the synergy between the edge and the cloud. Apple appears relatively cautious, emphasizing a unified underlying architecture and self-developed chips.

As the demand for AI chips in PC increases, power consumption issues are also beginning to emerge. The X86 architecture personal PC market led by Intel is welcoming a new batch of competitors, which may completely overturn the hardware and application ecosystem of PCs.

In just one week, the PC market has reignited with the three giants, Huawei, Apple (AAPL.NASDAQ), and Lenovo Group (00992.HK, hereinafter referred to as "Lenovo"), consecutively dropping major bombshells.

On April 11, Huawei released the new MateBook X Pro, which for the first time integrates its self-developed PanGu large model and has established partnerships with over 100 AI large model partners.

On April 15, the Apple M4 series chip was exposed, and the all-new Mac computer series may debut by the end of this year, with the new products being able to execute more complex AI instructions.

On April 18, Lenovo unveiled multiple PC products at its Innovation and Technology Conference, including the dual-screen Yoga Book 9i Yuanqi Edition, all of which are equipped with the Lenovo Xiaotian AI intelligent agent.

Although these products focus on different categories such as ecosystems, chips, and intelligent applications, they all point to the same new concept—AI PC (Artificial Intelligence Personal Computer).

There is already a basic consensus on the definition of AI PC in terms of hardware—it integrates CPU (Central Processing Unit), GPU (Graphics Processing Unit), and NPU (Neural Network Processor) to form a hybrid computing unit.

This configuration endows devices with powerful local AI processing capabilities, driving the execution of AI applications such as image recognition, speech recognition, and natural language processing, enhancing computing performance and user experience.

Terminal manufacturers are rushing to launch new products, in fact, to seize the golden time of 2024.

From 2019 to 2021, the peak of home office work triggered a wave of computer upgrades, and then the PC market has been in a downturn. According to the market research firm Canalys, the global PC shipment volume has continuously declined year-on-year for seven consecutive quarters, until a turnaround occurred in the fourth quarter of last year. In the first quarter of this year, the shipment volume increased by 3% year-on-year, reaching 57.2 million units, which may indicate that a new wave of computer upgrades is imminent.

The traditional PC consumer market typically has a replacement cycle of 3 to 5 years, and the introduction of AI PC may shorten this cycle. Major manufacturers are making every effort to gain a higher market share.

Hardware readiness, application convergence, and price increases

Artificial intelligence is rapidly penetrating the PC hardware field. Upstream giants in the PC industry chain, such as NVIDIA (NVDA.NASDAQ) and Intel (INTC.NASDAQ), have already launched a new generation of processors with integrated AI acceleration units, taking a significant step in enhancing local AI processing capabilities.

The latest PC products from terminal manufacturers Huawei and Lenovo are both equipped with Intel's cutting-edge Core Ultra series processors, with a total computing power of up to 34 TOPs (Tera Operations Per Second, a unit of processor computing power equivalent to ten trillion operations per second), sufficient to smoothly run AI models with 20 billion parameters on the device. This is undoubtedly a leap forward in hardware.

However, their selling points at the application level are very similar. Huawei and Lenovo both emphasized the AI document summary function of their new PCs.

Huawei uses its PanGu large model to extract key information from audio and video materials or real-time meetings and achieve functions such as real-time subtitle translation. The Lenovo Xiaotian intelligent agent can summarize document content and generate PPT slides quickly through hand-drawn images. In addition, these new PCs also have AI video frame interpolation, intelligent noise reduction, dialect recognition, and other functions.

However, many online applications have long been able to provide similar functions and, in some aspects, perform even better. For example, Tencent Meeting launched an AI assistant function in February, which can real-time organize meeting minutes, intelligently sort out discussion points, and quickly locate key information. Feishu, a product under ByteDance, also provides similar meeting assistant functions, including real-time beautification, audio noise reduction, and dialect recognition.

An algorithm engineer at an internet company said, "These capabilities can be achieved through algorithm improvements or third-party applications. If edge-side large models cannot achieve better results, then the innovation of these so-called AI PCs is not significant."

According to Lenovo's demonstration at the Innovation and Technology Conference, even in the case of limited cloud-side bandwidth, PCs can process information locally to protect user privacy and security. However, the Lenovo PC Yuanqi Edition is currently in the pre-sale stage, and users cannot experience the functions of edge-side personal large models or intelligent agents on products being sold in direct stores.

Huawei directly provides a clever application called "AI Space," which integrates various software such as WPS AI, Wenxin Yiyuan, iFlytek Xinghuo, and Zhipu Qingyan, to achieve functions such as PPT production and intelligent document generation. These capabilities, which were previously scattered across various software applications, are now concentrated and presented as a quick entry point for users, seeming to be a distance away from a true AI PC.

With the introduction of AI functions, the pricing of PCs has also seen a significant increase.

For example, Lenovo will release the Yoga Book 9i Yuanqi version on May 20th, with a starting price of 17,999 yuan, a 6% increase compared to the same model released last year. Other PC products equipped with the AI smart assistant Xiaotian also generally exceed 10,000 yuan. The first launch price of Lenovo's flagship product, the Thinkpad X1 Carbon AI 2024, positioned as a premium flagship, has reached 14,999 yuan, a 15% increase compared to the same model released last year.

Huawei's MateBook X Pro has a starting price of 11,199 yuan, a price range that has been dominated by Apple in the past.

Currently, the AI functions added to these AI PCs are still in the free-to-use stage, and it is still unknown whether they will be charged in the future. Perhaps they will follow Microsoft's example and charge a subscription fee for Copilot annually, or customize fees based on differences in smart assistant configuration, storage requirements, and computing power needs.

At present, AI PC hardware already has AI acceleration computing power, but it is still unclear how terminal manufacturers will leverage the advantage of computing power and deploy local large models. With the continuous advancement of AI technology, the market's expectations for AI PCs continue to rise, but the maturity of software applications and price factors are still the focus of consumer attention.

Fang Jing, chief analyst of the electronic industry at Minsheng Securities, stated that for PCs, hardware is the foundation, and better processors and larger memory drives can drive AI. However, what ultimately determines the user experience is the capability of the models themselves.

On the road to AI PCs, Lenovo is outspoken, Huawei is more balanced, and Apple is cautious.

In the exploration of AI PCs, Lenovo took the lead in defining edge-side models and personal intelligent agents. Huawei continued its consistent strategy of building ecosystems, emphasizing the Internet of Everything and edge-cloud collaboration. Apple appeared relatively cautious, emphasizing a unified underlying architecture and self-developed chips.

Perhaps the changes in the main businesses of the three companies can hint at the formation of their different styles.

Since the 2018 fiscal year, Lenovo has steadily held the top spot in global PC market sales volume, but its PC business revenue has seen a significant decline. In the third quarter of the 2023 fiscal year, Lenovo's revenue was 111.83 billion yuan, with a year-on-year growth of only 3%, and a net profit of 2.394 billion yuan, a year-on-year decline of 24%. The Smart Devices Group, composed of personal computers, tablets, smartphones, and other smart devices, saw its revenue decrease by 13.7% year-on-year.

Lenovo's sales gross margin is also lower than that of industry competitors, at less than 15% before the 2020 fiscal year and currently only 17%. In contrast, other PC manufacturers such as HP and Dell Technologies have had sales gross margins of over 20% for many years, while Huawei and Apple have maintained sales gross margins of over 40%.

The reason behind this may lie in Lenovo's PC product strategy, which focuses on capturing a larger market share and offering more mid-to-low-priced products, which may lead to a lower overall gross margin. Companies like HP and Dell Technologies have focused on high-margin niche markets in recent years, such as high-end business laptops or gaming PCs.

In recent years, Lenovo has intensified its investment in server-related businesses such as AI servers, edge servers, and storage servers, as revenue from terminal businesses, led by PCs, has declined. It's evident that AI has become one of Lenovo's most prioritized sectors. In June 2023, Lenovo announced an additional investment of $1 billion over the next three years to accelerate global AI deployment in enterprises.

Lenovo continues to emphasize AI integration in its PC lineup, not only pushing the prices of its PCs above 15,000 yuan, thus targeting the high-end market, but also leveraging its AI servers to expand its B2B business.

Regarding the issue of raised prices, Yang Yuanqing, Chairman and CEO of Lenovo Group, explained in interviews with media outlets such as Economic Observer that running large models on PCs, even though compressed, still require high machine specifications, possibly influencing pricing.

In contrast to Lenovo's focus on the PC and server markets with an emphasis on edge AI capabilities, Huawei's AI core remains its smartphones, with PCs serving as just one entry point for multi-device collaboration.

Huawei believes that one of the key aspects of AI PCs is their ability to connect multiple devices, considering Huawei's diverse product range including smartphones, wearables, and cloud services. Interconnection and interoperability to create new scenarios are essential for driving overall revenue growth in Huawei's terminal business. This strategy led to a 17.3% year-on-year increase in Huawei's terminal sales revenue to 251.5 billion yuan in 2023.

According to market research firm Canalys, in 2023, mainland China's desktop and laptop sales saw a double-digit growth of 11% for Huawei, while other manufacturers experienced varying degrees of decline.

Unlike Lenovo and Huawei, Apple has never claimed that its Mac computer series is called AI PC. Instead, it emphasizes that the M-series chips and the Mac OS software ecosystem can provide powerful AI capabilities, showcasing the strength of native features such as real-time speech-to-text, translation, text prediction, and visual understanding.

The newly exposed M4 chip is planned to be equipped with a completely upgraded neural engine, with a significantly increased number of cores to meet the growing demand for artificial intelligence tasks. The computing power of the M4 chip is still unknown, but the investment in research and development indicates Apple's emphasis on this.

According to the financial report, Apple's research and development investment in 2023 amounted to $29.92 billion, accounting for 7.8% of its revenue, returning to the level before the launch of the iPhone in the early 2000s. During the earnings call, Apple CEO Tim Cook stated that Apple's research and development expenditure is "very competitive" and is investing a significant amount of time and effort to integrate artificial intelligence into its software platform. These features are expected to be available to customers later this year, and there may be a possibility of seeing "Apple GPT" in Apple's products.

The success or failure of technology companies is often determined by research and development investment. In 2023, among the top 100 global R&D investment companies, Apple ranked fourth, with Huawei closely following in fifth place. Huawei's R&D expenditure was 164.7 billion yuan, accounting for 23.4% of its revenue, with accumulated R&D expenditure exceeding 1.11 trillion yuan over the past decade.

In contrast, Lenovo's R&D investment is less than one-tenth of Apple and Huawei's. Its R&D expenditure last year was 15.3 billion yuan, accounting for only 3.4% of its revenue.

Luca Rossi, President of Lenovo Group's Intelligent Devices Group (IDG), stated in an interview with Economic Observer that it is expected to see more advanced chip solutions with lower power consumption to enhance AI processing capabilities from the second half of this year to 2025.

Where will the more drastic transformation of AI PC occur first? Perhaps the latest developments from companies such as NVIDIA, Intel, AMD, and Qualcomm may reveal a trend of PCs shifting towards the use of Arm architecture.

The battle between X86 and Arm architectures is reigniting, potentially reshaping the PC ecosystem.

AI PCs emphasize the AI computing capabilities of their processors.

According to Microsoft's definition, AI PCs running the Windows operating system must have the ability to run Copilot locally and be equipped with an NPU capable of delivering 40 TOPs of performance.

Currently, the Intel Ultra processor based on the X86 architecture has a maximum computing power of 34 TOPs, while the Qualcomm Snapdragon X Elite platform based on the Arm architecture can achieve 45 TOPs, providing AI PCs with greater power.

X86 and Arm are the two core architectures of PC processors, which can be likened to the internal combustion engine and battery pack in a car.

Since its inception in 1978, the X86 architecture has been the preferred choice for PCs and servers due to its high performance, with the "Wintel" alliance (Windows+Intel/AMD) dominating the PC market. On the other hand, the ARM architecture has taken a leading position in mobile, VR/MR, and IoT devices due to its energy efficiency, low power consumption, and convenient cross-device transfer capabilities.

The only PC manufacturer that adopts this architecture is Apple.

In 2020, Apple removed the last Intel chip, breaking free from its reliance on Intel for the past decade. Since November 2020, all Mac computer products released by Apple have been equipped with its self-developed M-series chips, with the third generation already released. Apple's computer products not only demonstrate powerful performance but also showcase the advantages of its application ecosystem.

Taking the MacBook Air released by Apple in March this year as an example, equipped with the latest M3 chip, its performance has significantly improved, reaching up to 60% faster than models with the M1 chip. Compared to the old models equipped with top-tier Intel chips, the speed has increased by up to 13 times, with a battery life of up to 18 hours.

As the demand for AI computing power in PC chips increases, power consumption issues are also becoming more prominent. The X86 architecture, dominated by Intel, in the personal PC market is facing a new wave of competition, which may completely disrupt the hardware and application ecosystem of PCs.

Since last year, companies such as Nvidia, Intel, Qualcomm, and AMD have successively announced their investment in Arm PC processors, with the earliest expected launch in 2025. Microsoft, in collaboration with Qualcomm, is also actively embracing Windows on Arm, planning to release Windows 12 with integrated built-in Arm chips to better support various AI functions.

As one of the largest chip suppliers for the mobile sector (based on the Arm architecture), Qualcomm took the lead in October 2023 in launching the Arm architecture PC chip X Elite, which supports running generative AI models with over 13 billion parameters at the terminal side. PC manufacturers such as Dell Technologies, Acer, and Lenovo are set to release a large number of new products equipped with the Qualcomm Snapdragon X Elite in the summer of 2024.

According to data from market research firm Counterpoint, in 2022, Intel and AMD collectively held an 80% share of the global PC market, while Arm accounted for 12.8%. It is projected that by 2027, the Arm architecture's share will double to 25.3%, while the X86 share will decrease to 74.4%, with Intel's share dropping to 60%, although its leading advantage remains significant.

According to the first "AI PC Industry (China) White Paper" jointly released by IDC and Lenovo, the development of AI PC is a dynamic concept and is currently in the AI Ready stage, mainly manifested in the upgrade of chip computing architecture, possessing basic local hybrid AI computing power.

In the AI On stage, AI PC will have complete core features, providing a revolutionary AI innovation experience. It will be able to provide general personal AI assistant services based on a richer AI application ecosystem, as well as personal large model fine-tuning services in edge private domain environments.

Windows on Arm can enjoy the advantages of low power consumption and long battery life brought by Arm architecture processors, while also continuing the software usage habits of X86 PC users. Perhaps, the future development of AI PC will follow this path.