Race to Hyper | AI PC: LENOVO GROUP takes the lead in seizing the first opportunity

Author: Zhou Yuan/Wall Street Journal

"GenAI+" (Generative Artificial Intelligence) is changing everything in the personal and enterprise terminals of the mobile internet era, giving rise to a whole new industry.

Yang Yuanqing, Chairman and CEO of Lenovo Group, said, "The future PC will be an AI PC, the future smartphone will be an AI smartphone, and the future workstation will be an AI workstation."

On December 15th, Lenovo Group launched commercial and consumer-grade laptops with AI PC features, marking the entry of the PC industry into the AI Ready stage.

Wall Street Journal has noticed that since 2023, especially in the fourth quarter, various players in the PC industry ecosystem, including chip manufacturers, terminal vendors, AI model vendors, and operating system software vendors, have been highly involved in the AI PC industry. Even the dominant player in the smartphone mobile SoC market, Qualcomm, couldn't resist the allure of this new industry.

On December 7th, Lenovo and IDC jointly released the industry's first "AI PC Industry (China) White Paper" (hereinafter referred to as the "White Paper"), which provides a clear definition of AI PC. As a new technological product, AI PC will become a personal AI assistant.

Amir Abulikemu, Vice President of Lenovo Group and Vice President of Strategy and Business Development in China, stated that the PC is undergoing its third major upgrade and is expected to once again become the preferred terminal for AI to benefit everyone. The PC will truly make AI a personal assistant for everyone (Personal AI Twin).

As a new technological product, the rise of AI PC has simultaneously given birth to new ecosystems and industries. Looking ahead to the future, which marks the 48th year since the birth of the PC, it can be seen that AI PC will gradually replace traditional PCs. In this process, the new technological dividends brought by GenAI+PC will also stimulate the commercial value increment of all participants in the ecosystem.

According to an IDC report, by 2027, AI PC will account for 85% of the overall PC market, becoming the absolute mainstream of the new PC market.

Thirsty Horse at the Spring: The Rapid Rise of the AI PC Market

The "White Paper" jointly released by Lenovo and IDC on December 7th has pointed out the industrial direction of the combination of AIGC and PC, as well as the commercial value increment of AI technologies represented by ChatGPT.

The "White Paper" provides a structural or modular definition of AI PC. AI PC is a hybrid that combines AI models, applications, and hardware devices.

In December 2023, ChatGPT celebrated its first anniversary. Prior to this, although the future vision of AI technology was promising, the reality was harsh. The industry had not found a technological foundation for the large-scale implementation of AI in the metaverse. The current hot chili chicken, NVIDIA, saw its stock price drop by over 50% in the year before the release of ChatGPT on December 22, 2022. However, in the year following the release of ChatGPT, the company's stock price soared 2.2 times (as of the close on December 12).

The emergence of the killer application ChatGPT not only solidified the position of the leather jacket master, Huang Renxun, but also pulled the industry out of the dilemma of lacking clear commercialization.

As a result, the direction of AI technology implementation has shifted from the super-concept metaverse to the concrete AIGC (Artificial Intelligence Generated Content), also known as GenAI. Both are sub-components of AGI (Artificial General Intelligence).

Once the industry determines its direction, the speed of advancement will astonish the world.

Just one month after the release of ChatGPT (December 22, 2022), AMD took the lead and released the Ryzen 7040 series platform (January 23, 2023), which integrated CPU, GPU, and NPU for the first time. The NPU integrated in this chip, using the Ryzen AI engine, directly created the concept and application of AI PC.

From early 2023 to the fourth quarter, AI large model companies, PC software developers, chip manufacturers, and terminal companies are all deeply involved in the AI PC industry on a large scale, aiming to seize the initiative in this emerging industry.

On December 15, Intel will launch the AI PC processor "Core Ultra" platform, which has 8 times the computing power of the previous generation and has already adapted to over 10 domestic AI large models. On December 14, Intel will release the next-generation Meteor Lake processor for AI PC.

On December 7, on the same day as the release of the "White Paper" by Lenovo and IDC in Beijing (Beijing time), AMD released the professional Ryzen 8040 series platform for AI PC laptops in San Jose, California.

If Intel and AMD were already deeply involved in the PC field and built their businesses on it, the smartphone SoC leader has also joined the competition, indicating the attractiveness of the AI PC industry.

On October 24, Qualcomm released the AI PC dedicated computing platform "Snapdragon X Elite".

Wall Street News noticed that whether it is Intel, AMD, or Qualcomm, their AI PC computing platforms adopt the XPU structure, which includes CPU, GPU, and NPU. The "White Paper" refers to this as heterogeneous hybrid computing, with the main purpose of providing powerful parallel computing capabilities for PCs. Heterogeneous hybrid computing utilizes computing units with different instruction sets and architectures to form a local computing system, matching the corresponding computing units to different AI workloads. As a result, the leap in parallel computing power allows PCs to easily perform complex AI model inference tasks, thereby demonstrating higher levels of intelligence.

This is the fundamental power feature of AI PCs. Unlike traditional PCs with separate CPUs and GPUs, they lack the computational structure of NPUs.

In addition to chip manufacturers actively participating, PC operating system developer Microsoft and terminal representative Lenovo have also shown high levels of attention and enthusiasm for this new track.

On November 1st, Microsoft launched the GenAI-driven Copilot PC application software targeting 365 enterprise customers. On October 24th, on the same day as the release of the AI PC dedicated chip Snapdragon X Elite by Qualcomm, Yang Yuanqing announced the world's first AI PC.

Acer and HP plan to launch AI PCs in 2024. In May of this year, Dell launched the full-stack GenAI solution and announced the PC product Dell Generative AI Solutions in collaboration with NVIDIA.

First Entry Point: Personal Intelligent Agents

As one of the modules that make up the AI PC hybrid body, AI large model businesses have different actions when participating in the AI PC industry.

The AI large model ecosystem pioneered by OpenAI has made slow progress in promoting inclusive applications of AIGC.

The reason for this is twofold. On the one hand, the growth of large model capabilities has encountered bottlenecks. More importantly, users have concerns about privacy and security when their personalized data and behavioral data are stored in large models hosted on public clouds. Among them, C-end users are concerned about privacy leaks, while B-end users have an additional concern - insufficient security for trade secrets.

Therefore, AI large model businesses (such as Baidu, Alibaba, Tencent, Huawei, and iFlytek) are advancing on two parallel paths: adjusting the parameter values and sizes of large models to meet the localized deployment needs of the edge, solving response speed and security and privacy protection issues; and building large models with different scales of parameter levels to support cloud collaboration in different scenarios and achieve experience optimization.

Among them, Baidu and iFlytek mainly promote edge-side large models, while Tencent and Alibaba focus on cloud-side large models. Huawei, on the other hand, relies on cloud-side large models for edge-side invocation.

Lenovo is relatively unique. On the one hand, Lenovo has launched AI PCs and personal intelligent agents. On the other hand, Lenovo's 3S strategy takes enterprise B-end services as its transformation direction, so it has also launched private large model services targeting the B-end.

Lenovo's edge-side large models have a special concept - personal large models, which is a rare differentiated product in the industry.

On November 21st, at a forum with the theme of "Prediction and Strategy," Yang Yuanqing said in his speech, "The computational load of large models will gradually shift from the cloud to the edge and end devices. More and more AI inference tasks will be performed at the edge and device ends, making personal large models necessary and possible. The logic behind Yang Yuanqing's conclusion is that the current user base for large models is still relatively small, and most large models are trained on powerful public cloud platforms. In the future, as the user base expands, considerations such as data security, privacy protection, higher efficiency, and lower costs in response to user needs will become increasingly important, which will require the migration of computational workloads to the edge and endpoint.

It is necessary to point out that Lenovo's personal large model is actually a combination of personal knowledge base and local large models.

The white paper mentions that the computational workload of large models is continuously migrating from the cloud to the end devices, and a combination of public and local large models is used to form personal large models.

When users have demands, they can access various capabilities of the personal large model through Lenovo's "personal intelligent agent" interface. Users can interact with the personal intelligent agent using natural language: relying on the AI computing power unit embedded with XPU architecture, the AI PC provides mixed AI computing power to fulfill the user's intentions and meets the user's application needs in different scenarios through an open ecosystem.

In summary, the role of the personal intelligent agent is to "understand user intentions and synchronize distribution." It also plays a role in protecting user privacy and security. Therefore, the personal intelligent agent becomes the first entry point for communication between the two parties when realizing user intentions through mixed large models (local/cloud) and mixed computing power (devices) and implementing AI applications.

The AI PC itself can effectively identify and manage user privacy data through the personal intelligent agent, and it can process and model most of the data locally through localized models and sufficient AI computing power. Only when it does not involve user privacy, cloud services need to be called.

It is necessary to describe the path map of users calling various AI large model capabilities through the personal intelligent agent.

Based on the personal intelligent agent built on the local AI large model embedded in the terminal, after receiving a user request, the local large model can instantly and accurately understand the user's intentions, convert them into corresponding tasks, and then decompose and identify the tasks. It categorizes and queries the local knowledge base, and then uses API interfaces to call suitable models or applications in the ecosystem to execute the corresponding tasks.

Afterwards, when the device, model, or application completes the task, the corresponding results are returned to the personal intelligent agent, which then organizes and integrates the results before providing feedback to the user.

This is a complete cycle path of interactive closed-loop, from the user's intention to the resolution.

2024: The Arrival of the AI PC Industry's First Year

The development of any industry ecosystem is a dynamic concept. The new PC industry built on the AI PC ecosystem is supported and promoted by various values and core features, gradually driving the industry towards maturity. However, this speed may far exceed the historical development record of the PC industry.

The core features of AI PCs include five aspects: personal intelligent agents that interact with natural language (the first entry point), embedded personal large models, standard local hybrid AI computing power (XPU structure SoC chips), an open AI application ecosystem (public large models), and device-level personal data and privacy protection.

The "White Paper" believes that the development of AI PCs will be divided into two stages: the AI Ready stage and the AI On stage. This is also the process of gradually implementing, maturing, and iterating the five major features of AI PCs.

In the AI Ready stage, it should have the basic hardware of local hybrid AI computing power (XPU structure) that is more targeted towards AI tasks. It can provide basic computing power guarantee for software and service innovation of AI PCs and enable experience innovation.

PC terminals that meet the characteristics of this stage have already appeared. On December 15th, Lenovo Group launched new commercial and consumer notebooks, aiming to enable new AI experiences while enhancing user productivity, creativity, and work efficiency. The new ThinkPad X1 Carbon, ThinkPad X1 Fold, as well as Xiaoxin Pro16 2024 and IdeaPad Pro 5i, are all equipped with Intel Core Ultra processors.

In this stage, the main performance is the chip upgrade computing architecture, and the integrated XPU structure hybrid computing processor is gradually being introduced to the market. At this time, the protagonist is the chip manufacturer. The hybrid XPU structure computing platform they develop will improve computing speed with higher energy efficiency ratio, and have higher stability and reliability during operation.

When the accumulation of software and hardware brings about an industry transformation, AI PCs will enter the AI On stage. At this time, AI PCs will have the complete five core features. Through in-depth communication with the industry, Wall Street News has found that in 2024, the industry will enter the AI On stage.

Therefore, in 2024, AI PCs that meet the standard definition will be able to provide revolutionary AI innovation experiences in core scenarios, and AI PCs will soon become everyone's personal AI assistant.

With core technological innovation, product experience optimization, and the prosperity of the AI application ecosystem, AI PCs will serve a wider range of general scenarios, thereby realizing edge-to-end collaborative computing and cross-device interconnection. They can even fine-tune and train personal large models based on personal data and behavioral characteristics in edge private domain environments.

Starting from 2024, AI PCs that meet the standards of the AI On stage, that is, AI PCs with complete core features, will gradually enter the market. This will drive rapid growth in the AI PC market and inject new momentum into the development of the entire industry.

Therefore, the "White Paper" believes that 2024 will be the year of AI PCs, and traditional PCs will accelerate their upgrade to AI PCs. Currently, upstream chip designers and manufacturers have already taken action. Intel plans to equip AI features to over 100 million PCs by 2025. The chip giant of the PC era, CEO Pat Gelsinger, stated during the company's earnings call in early November, "We believe that AI PCs are the key turning point in the PC market in the coming years, and every one of our products will integrate AI."

In the financial quarter ending in September this year, Intel's PC division saw the smallest decline in revenue in the past eight quarters. AMD, their long-time rival, achieved the strongest growth rate in their PC business in two years. AMD CEO Lisa Su expressed optimistic expectations, saying, "Considering the AI PC cycle and the update cycle of Microsoft Windows, we expect growth (trends) to continue until 2024."

Professional market research institutions also hold optimistic expectations for the development of the AI PC industry.

According to a forecast report released by Canalys, the shipment of AI PCs will exceed 50 million units in 2024. Canalys predicts that the adoption rate of AI PCs will accelerate from 2025, and by 2027, AI PCs will account for approximately 60% of the total PC shipments. The market research institution also suggests paying attention to the investment opportunities brought by the launch of transformative PC products.

According to an IDC report, the PC market's growth comes from both consumer and commercial sectors: in the personal consumer market, AI PCs will shorten the replacement cycle for users, accelerate the arrival of the replacement wave, and change the user demographic structure of the PC market. Small and medium-sized enterprises will leverage AI PCs to accelerate intelligent transformation, optimize customer experience, and improve operational efficiency. The changes in large enterprises will be reflected over a longer time span, as AI PCs will be combined with long-term intelligent transformation in large enterprises, fully unleashing internal vitality.

According to IDC's report, the assembly ratio of new AI PCs in the Chinese PC market will rapidly increase in the coming years. By 2027, the proportion of AI PCs will reach as high as 85%, becoming the absolute mainstream of the new PC market.