At the 2024 International Consumer Electronics Show (CES), Intel and AMD, the leaders in PC CPU chips, announced new desktop or laptop processors with integrated AI inference modules. NVIDIA, the &#34;shovel seller&#34; in the AI training field, also sought to gain market share in the AI PC market by launching the new GPU product GeForce RTX 4080 SUPER at CES, which covers both high-quality gaming and AI application software. Compared to AI training, the field of AI inference is closely linked to the demand from large-scale consumer electronics and other application terminals, so the development focus of the AI industry is expected to shift from &#34;training&#34; to &#34;inference&#34; comprehensively.
The debut of ChatGPT at the end of 2022 marked the accelerated entry of human society into the AI era. In 2023, there was an explosive growth in GPU demand focused on AI training. Driven by this trend, NVIDIA became the world's first chip company with a market value exceeding one trillion US dollars in 2023. In contrast, the field of AI inference, which has lower requirements for TFLOPS (i.e., computing power) and lacks heavyweight products other than ChatGPT in the AI application terminal, appears dim.
However, with the emergence of new consumer electronics in 2024, such as AI-integrated PCs and AI smartphones, the trend of &#34;AI + everything&#34; is unstoppable. South Korean tech giant Samsung held a major press conference at CES 2024 with the theme &#34;All for AI: Interconnected in the AI Era.&#34; Samsung's latest products are all AI-related, including AI TVs, AI refrigerators, AI washing machines, AI vacuum cleaners, and AI laptops. Volkswagen announced at CES that its voice assistant will be embedded with ChatGPT AI technology and used in all models. The automotive giant showcased the first car equipped with ChatGPT, allowing users to interact with the voice assistant through a conversation interface similar to ChatGPT to resolve queries. This AI technology is planned to be promoted in North America and Europe in the early second quarter of this year.
In a recent research report, UBS, the international bank, pointed out that the global technology industry is just beginning a large-scale growth cycle. UBS predicts that by 2027, AI technology will achieve extremely extensive application scale in various industries of major economies worldwide, driving AI large models and AI software applications to become a sub-market worth as much as $225 billion. Compared to the scale of only $2.2 billion in 2022, this can be described as an epic leap, with an expected compound annual growth rate of 152% during this period.
The trend of large-scale application of edge AI models and AI software based on the background of &#34;AI integration with everything&#34; will inevitably be presented in a wide range of consumer electronics application terminals, such as PCs, smartphones, and smartwatches. This also means that the importance of AI inference is increasingly prominent, as it is the core technology behind the efficient operation of edge large models and AI software.
Qualcomm CEO Cristiano Amon recently emphasized that the main market for global chip manufacturers will soon shift to the AI inference field.
Compared to AI training, the demand for GPU parallel computing power in the AI inference field is much lower in the context of &#34;massive data bombardment&#34; applications. The inference process involves using pre-trained models to make decisions or identify objects, and CPU-based central processing units, which excel at complex logic processing and control flow tasks, are efficient enough to handle many inference scenarios.
Therefore, as AI continues to integrate into all aspects of life, CPU-based central processing units are about to experience a new wave of explosive market expansion opportunities. The long-neglected consumer electronics CPU giants are returning to the global investor's attention. With the AI inference market becoming increasingly significant, major players in the consumer electronics field, such as AMD (AMD.US), Intel (INTC.US), and Qualcomm (QCOM.US), are experiencing their &#34;NVIDIA moment,&#34; where stock prices and performance are synchronously skyrocketing.
The era of AI PCs and AI smartphones begins! CPU giants return to the spotlight
2024 can be considered the year of AI PCs and AI smartphones. Well-known PC brands such as HP, Dell, Acer, Asus, MSI, and Gigabyte will launch their first wave of AI PCs based on Intel or AMD processors in 2024. Samsung is inclined to make AI technology the key factor in achieving larger smartphone sales this year, and new AI smartphones from Chinese smartphone manufacturers such as Xiaomi, Vivo, Honor, and Oppo are planned for release this year. With PC and smartphones as the two core consumer electronic carriers for AI applications, a large number of emerging trends are bound to emerge.
Intel's new Core Ultra processor integrates AI-specific Neural Processing Units (NPUs) and Arc GPUs into the CPU. The NPU is specifically designed to accelerate AI inference tasks, and this integrated CPU+NPU+GPU central processing unit is designed to be the company's &#34;most efficient processor,&#34; marking the official arrival of the AI PC era. Intel's Lunar Lake processor for laptops will be launched in the second half of 2024. This chip features a &#34;new low-power architecture and significant IPC improvements,&#34; with AI data processing performance of the GPU and NPU modules three times higher than that of Meteor Lake. Intel has stated that these chips are already available to its partners.
At CES 2024, AMD introduced the Ryzen 8000G series desktop processors, which are AMD's first desktop-level central processing units with integrated AI capabilities. They integrate an NPU for AI acceleration tasks and feature a CPU based on the Zen4 architecture. The Ryzen 8000G also integrates an NPU developed by the Ryzen AI engine and an RDNA3 integrated GPU.
Positioned as a desktop-level platform, Qualcomm Snapdragon 8 Gen3 is Qualcomm's first mobile chip specifically designed for AI acceleration. It supports the generation of multiple models, including Meta Llama 2, and can handle AI large models with up to 10 billion parameters on the edge, executing up to 20 Tokens per second.
According to Zacks Investment Research, 2023 is a crucial year for the AI industry, with the introduction of training-end GPU products from NVIDIA and AMD, as well as various investments and strategic acquisitions. Looking ahead to 2024, Zacks suggests that with technology companies having the foundation of chips, they will continuously update AI large models and build AI applications. This is expected to drive consumer hardware upgrades, such as AI PCs and AI smartphones, as well as new AI-based software services, including edge AI large models and embedded chatbots.
Efficiently running edge AI large models and AI software is based on the core technology of AI inference, with the hardware foundation of CPU as the central processor. The architecture of the CPU determines its ability to perform general-purpose computing tasks, focusing on control flow and scheduling characteristics for handling complex sequential computing tasks and logical decisions, making CPUs shine in the field of AI inference.
In the field of AI inference, such as applications of edge AI large models in consumer electronics like AI PCs, AI smartphones, and smartwatches, as well as running various AI software, a CPU as the central processor focusing on complex logical decisions, integrated with NPUs and GPUs as auxiliary computing support, can achieve efficient operation of streamlined edge AI large models and multiple AI software. In these AI inference scenarios, the demand for GPU computing power is far less than in training-end scenarios. After all, AI training involves massive parallel computing tasks, processing large amounts of data and performing complex mathematical operations, which are suitable for accelerating processing through parallel computing. To efficiently execute these parallel tasks, GPUs or other AI-specific hardware (such as Google TPUs and ASIC chips) are the main force.
In most AI inference tasks, using a CPU as the core, supplemented by NPUs and GPUs, can effectively execute AI tasks that do not require massive parallel processing, such as processing small to medium-sized datasets or performing routine model inference tasks.
In this combination, the CPU is considered the core part, equivalent to the &#34;human brain&#34;, responsible for processing complex logical decisions and controlling task flows. It is the main computing unit of the entire chip system, responsible for executing program instructions, handling daily computing tasks, and coordinating and managing other parts of the system. For AI inference tasks that do not require massive parallel processing, the CPU can achieve efficient scheduling and processing. The NPU (Neural Processing Unit) is optimized for AI inference acceleration, providing fast AI inference performance at lower power consumption, especially suitable for handling neural network-related tasks. The GPU excels in parallel computing and is suitable for executing a large number of matrix and vector operations. It can shoulder heavy responsibilities in data-intensive AI inference tasks such as image and video analysis.
Therefore, Intel, AMD, and Qualcomm, the three long-forgotten old CPU giants, have recently returned to the spotlight of the capital market. The upward trend in their stock prices since November last year is the best proof that global funds favor these companies. These three giants were even mistakenly believed by some analysts to have missed the global AI boom.
Looking ahead, the field of AI inference is expected to surpass the training field. These &#34;shovel sellers&#34; are about to usher in their &#34;NVIDIA moment&#34;.
From the perspective of industry development trends, the AI computing load is likely to gradually shift from training to the inference end. This means that the threshold for AI chip companies may significantly decrease, and chip companies covering wearable devices, electric vehicles, and the Internet of Things are expected to fully penetrate the AI inference chip field in the future. Intel's CEO recently stated that AI inference technology will become more important than AI training technology, emphasizing that Intel will not solely rely on AI training and that Intel values the AI inference field more.
Qualcomm's CEO, Amon, pointed out that the main battlefield for chip manufacturers will soon shift from &#34;training&#34; to &#34;inference&#34;. Amon said in a recent interview, &#34;Currently, the AI market is mainly focused on the 'training' phase of using big data to train large language models, and NVIDIA is the main beneficiary in this field. But as AI large models become more streamlined, capable of running on devices, and focused on inference tasks, the main market for chip manufacturers will shift to 'inference', that is, model applications. It is expected that data centers will also be interested in processors specifically designed for inference tasks of trained models, and everything will contribute to the inference market surpassing the training market.&#34;
Morgan Stanley, a major Wall Street bank, pointed out in its top 10 investment themes for 2024 that with significant improvements in data processing, storage, and battery life in consumer edge devices, there will be more catalysts to drive the catch-up of the edge AI segment. The development focus of the AI industry will also shift from &#34;training&#34; to &#34;inference&#34; in 2024.
A Shift Towards &#34;Inference&#34; on the Edge**
Edge AI refers to the technology that directly processes AI data streams on edge devices such as PCs, smartphones, IoT devices, and cars. Research firm Gartner predicts that by 2025, 50% of enterprise data will be generated at the edge, spanning billions of devices. This means that the inference of AI models (the process of applying models for decision-making or recognition) is expected to be performed in bulk on edge devices rather than on remote servers or in the cloud.
According to a research report by InvestorPlace, 2024 will be the year of a major explosion in AI application software, with hundreds of new AI-based applications expected to emerge. These AI applications, covering various industries, will be widely adopted in society. InvestorPlace predicts that by 2025, we will be in the midst of a wave of AI software that covers all industries. Therefore, the trend of a major explosion in AI software implies a new wave of replacement for AI PCs and AI smartphones that support these software applications. In addition, the increasing performance requirements of software on hardware are expected to stimulate the demand for chips used in inference tasks for consumer electronics.
Renowned research firm Counterpoint Research predicts that the global PC market's shipment volume will return to pre-pandemic levels by 2024, thanks to the replacement of Windows 11, the next wave of Arm PCs, and AI PCs. It is expected that AI PCs will grow at a compound annual growth rate of 50% starting from 2020 and will dominate the PC market after 2026, with a penetration rate exceeding 50%.
<img src="https://imageproxy.pbkrs.com/http://img.zhitongcaijing.com/images/contentformat/2dfdb59e61b9f5ef1efb49a4c1fd04a4.jpg?x-oss-process=image/auto-orient,1/interlace,1/resize,w_1440,h_1440/quality,q_95/format,jpg" alt="" original-src="https://imageproxy.pbkrs.com/http://img.zhitongcaijing.com/images/contentformat/2dfdb59e61b9f5ef1efb49a4c1fd04a4.jpg"/>
Another research firm, Canalys, predicts that global PC shipments in 2023 will be approximately 249 million units, a YoY decrease of 12.4%. It is projected that PC shipments in 2024 will reach 267 million units, a YoY growth of 7.6%. In 2024, the shipment volume of AI PCs will be around 20 million units, and by 2027, 60% of PCs will have the latest version of AI capabilities. PC shipments in 2027 will exceed 175 million units. According to Qunzhi Consulting, 2024 is expected to be the first year of AI PC development, with a penetration rate in the PC market expected to reach 7%. The penetration rate is expected to approach 30% in 2025 and exceed 50% in 2026.
According to the forecast data from renowned market research firm Counterpoint Research, the shipment volume of AI smartphones with built-in generative AI capabilities is expected to exceed 500 million units by the end of 2027. Counterpoint predicts that 2024 will be a crucial year for global AI smartphones, but the estimated shipment volume will only reach 100 million units. Therefore, the compound annual growth rate of the AI smartphone market between 2023 and 2027 is estimated to be as high as 83%.
According to a report released by Counterpoint in December, &#34;Samsung and Qualcomm are the most direct leaders in this segment of the market, as their current product supply system and production capacity make them pioneers.&#34; &#34;Just like what it has done in foldable phones, Samsung is likely to occupy nearly 50% of the market share in the next two years, followed by major Chinese OEM manufacturers such as Xiaomi, Vivo, Honor, and Oppo.&#34;
In terms of stock price expectations, the stock price of CPU and GPU industry giants AMD has been soaring recently, and Wall Street analysts are becoming more optimistic about AMD's target stock price. Wall Street investment firm Melius Research upgraded AMD's rating from &#34;hold&#34; to &#34;buy&#34; and significantly raised its target price for the next 12 months from $125 to $188 (AMD's latest closing price is $146.18). The firm believes that from 2024, generative AI is expected to have a &#34;halo effect&#34; on enterprise IT spending, and chip companies with excellent performance can continue to maintain the momentum of stock price growth from 2023 onwards.
Analyst John Vinh from KeyBanc continues to maintain an &#34;outperform&#34; rating on AMD and significantly raised the target price from $140 to $170. Vinh stated that AMD is highly likely to have a market share of about 15% in the AI chip market in the coming years, which means a massive revenue opportunity of approximately $20 billion.
Analyst Ivan Feinseth from Tigress Financial significantly raised the target stock price for Intel from $46 to $66 (Intel's latest closing price is $48.45) and maintained a &#34;buy&#34; rating. Feinseth's raised price target is based on two major catalysts: Intel's newly launched PC processors combined with AI acceleration and the continued expansion of IFS (Intel Foundry Services), which will further accelerate revenue and cash flow growth trends.
Ruihua analyst Vijay Rakesh has raised the target price of Qualcomm from $140 to $155 (Qualcomm's latest closing price is $139.03), and maintains a &#34;buy&#34; rating on the stock. Another well-known investment institution, Bernstein, has also raised the target price of Qualcomm from $145 to $160, and maintains an &#34;outperform&#34; rating.

NVDA

SSNGY

VWAGY

The focus of the AI industry's development is expected to shift from "training" to "inference" comprehensively. Intel, AMD, and NVIDIA have successively released new products that integrate AI inference modules, and new consumer electronics such as AI PCs and AI smartphones that integrate AI are about to emerge. Samsung Electronics has launched a full line of AI-related products at CES, and Volkswagen AG will embed ChatGPT artificial intelligence technology. The global technology industry is just getting started, and the prospects for the AI industry are broad.

The unstoppable momentum of AI is sweeping across all industries! In the field of reasoning, "Selling Shovels" welcomes its "Nvidia Moment".