Financial Services News, December 1 (Reporter Fu Jing) “The rapid deployment of AI applications requires semiconductors in terms of performance, power consumption, integration, and reliability. This trend has spawned demand for system-level semiconductor manufacturing. To this end, Intel is actively investing, and the Intel 18A will be mass-produced in 2025, and the next-generation AI PC processor Panther Lake based on Intel 18A and the next-generation data center processor Clearwater Forest will also be released next year.” Recently, Wang Rui, senior vice president of Intel and chairman of Intel China, spoke at the Intel New Quality Productivity Technology Ecosystem Conference.

Chips play a central role in the big model era. According to the Financial Services Association reporter's observation, on the application side, chip manufacturers supported the launch of AI PC processors running local AI models, and terminal manufacturers such as HP, Lenovo, and Dell responded positively. The AI PC ecosystem is becoming more and more vibrant. Global AI PC shipments in Q3 this year already accounted for 20% of total PC shipments, and many new AI PC products will be launched within the year. In the data center sector, while GPUs dominate, CPUs are also targeting the growing big model inference market, and manufacturers are focusing on increasing server chip production capacity. On the edge side, through the configuration of CPU, GPU, and NPU, edge AI processing capabilities have also promoted the actual deployment of AI applications in retail, medicine, logistics, agriculture, transportation, manufacturing, etc.

“To meet the challenges in computing power, systems, and deployment costs brought about by the surge in AI workloads, Intel has launched a full-stack AI product portfolio including Core Ultra processors, Xeon scalable processors, Gaudi deep learning accelerators, and Intel GPUs.” Speaking of products, Wang Rui said so.

Application side: AI PC shipments will exceed 0.1 billion units by the end of next year

A Financial Services News reporter recently learned that Intel has set a goal of cumulative AI PC shipments exceeding 0.1 billion units by the end of next year. By the end of October, 20 million AI PCs based on Core Ultra processors had been shipped, and it is expected to reach 40 million units by the end of 2024.

Since Intel pioneered the AI PC concept in the industry in September of last year, AI PCs have continued to receive market attention and have stimulated a wave of PC switching. Looking at specific global shipments at present, the latest data released by research firm Canalys shows that in the third quarter of 2024, global AI PC shipments reached 13.3 million units, accounting for 20% of total PC shipments this quarter.

According to media reports, from an industrial chain perspective, the most valuable investment areas for AI PCs are upstream chip manufacturing and midstream machine manufacturing, especially chip manufacturing such as CPU, GPU, NPU, and memory. Currently, manufacturers such as Intel, AMD, and Qualcomm have launched AI PC processors that support the operation of large local AI models.

In September of this year, Intel released the Core Ultra 200V series processor codenamed Lunar Lake, increasing computing power to 120 TOPS. The reporter learned that many new AI PC products will be launched during the year. For example, HP will launch the NextGen AI PC equipped with the Lunar Lake platform at the end of the year, the AI PC Yuanqi series and Aura series AI PCs previously launched by Lenovo are equipped with Intel chips, and the ThinkPad X1 Carbon Aura AI Yuanqi Edition is scheduled to be launched in early December.

Photo by a reporter from the Financial Federation

Also, according to a Financial Services Association reporter, Intel brought AI features to the desktop field for the first time by launching the Core Ultra 200S processor this year, while also exploring opportunities on the all-in-one computer and mini PC circuit. It is reported that based on AI PCs, the efficiency, processing power, and level of intelligence of traditional applications will be improved, while in the field of gaming, it will enable new gameplay and stimulate the replacement of game hardware. According to reports, in order to create more AI PC application scenarios, Intel has partnered with more than 40 local ISVs (software solution providers) to expand the software ecosystem.

During the conference, a Microsoft source told the Financial Federation reporter that if the hardware specifications reach the peak NPU performance of at least 40 TOPS, a minimum of 16 GB of memory, and a minimum of 256 GB of storage space, a Windows 11 laptop can obtain Windows 11 AI+PC certification. In December of this year, features such as increasing the resolution of low-resolution images with one click, erasing excess elements in images with one click, and Bing screenshot search will be pushed to Windows Insider for users to use.

Server-side: Adding production capacity to explore “oil cooling” innovation

Currently, Intel is also adding server chips, and has introduced Xeon 6 performance core and energy efficiency core processors for data centers.

On October 28, Intel announced that it will expand its packaging and testing base in Chengdu. The Finance Association reporter recently learned that the focus of this expansion is not only setting up a one-stop customer solution center, but also increasing server chip production capacity. Wang Rui stressed that in the future, the Chengdu base will cover a wide range of products, from clients to server chips, to meet the needs of the Chinese market and greatly shorten the response time to customers.

A CIFA reporter learned that at present, more than 70% of GPU-accelerated servers have chosen Intel Xeon CPUs as their “head”. Based on Xeon 6 performance cores, new server products from manufacturers such as Xinhua 3, ZTE (000063.SZ), and Lenovo under Inspur Information (000977.SZ), and Ziguang (000938.SZ) will be launched one after another from September this year to the first quarter of next year.

Data center computing power demand and energy consumption challenges continue to rise, so liquid cooling has become the focus of attention in the industry. Previously, Intel had partnered with many OEMs, server vendors, cloud vendors, software vendors, operating system vendors, etc. to accelerate the green transformation of data centers, and a financial news agency reporter recently learned that in terms of immersion liquid cooling, “oil cooling” innovation has also begun: Xinhua 3 and Intel released an oil immersion solution, using environmentally friendly coolant, based on Intel's G-Flow immersion liquid center solution to provide cooling capacity for kilowatt chips. On the G-Flow side, Intel collaborated with Green Cloud Map and Lixun Technology during the design phase to optimize the flow channel and radiator design in the system and cabinet, and use the liquid height difference to generate gravity to achieve high-performance heat dissipation.

The G-Flow-based immersion liquid cooling solution uses liquid-cooled oil Photo by a reporter from the Financial Federation

“We used synthetic liquid-cooled oil for data centers in our solution. The combustion point is above 200 degrees. In fact, the CPU with the highest temperature in the components generally does not exceed 115 degrees, so it is relatively safe. Liquid cooling used to be fluorinated liquid. With the emphasis on environmental protection, more and more manufacturers may explore the use of 'oil cooling'. Xinhua 3 has already done a lot of research on 'oil cooling'. Currently, this plan is still in the early stages of promotion, and it is hoped that it can be promoted into a mass production plan in the future.” An Intel source told the Financial Services Association.