occurrences
Microsoft held the Build 2024 developer preview conference in the early morning of May 21, Beijing time, and unveiled the Copilot+ PC at the conference and launched the new Surface Pro, GPT-4o model with OpenAI built in, using ARM architecture SOC, and optional Qualcomm Snapdragon X Elite and Snapdragon X Plus chip platforms.
On May 21, local time, Qualcomm announced the launch of the Snapdragon Dev Kit for Windows (SnapdragonDev Kit for Windows), which aims to provide hardware support for Windows developers and accelerate application construction for Copilot+ PCs.
Key points of investment
End-side AI breakthroughs are expected to become the driving force for a new round of smart terminal growth. Joint breakthroughs on the hardware and algorithm side have made end-side AI have a certain foundation. The addition of AI capabilities is expected to become an important upgrade for smart terminals in recent years, driving a new round of smart terminal shipments.
1) On the hardware side, chip manufacturers focus on strengthening AI processing capabilities and supporting the deployment of AI capabilities on the end side. For example, Qualcomm's Snapdragon 8 Gen3 and Tianji 9300+ series chips can support the deployment of large models on the mobile phone side. Qualcomm's Snapdragon X Elite/X Plus and Intel's Lunar Lake series chips to be released in 24Q3 can be used in AIPC scenarios;
2) On the software side, the lightweight AI model can maintain the model's performance while reducing the scale, and ensure the user experience of the end-side AI model. For example, OpenAI's newly released GPT-4O can be run on Apple computers; Microsoft's recently launched open source AI model PHI-3-mini has parameters of only 3.8B, but the test performance can be compared with MixTral8x7B and GPT-3.5; the test performance of the Gemma-7B, an open source lightweight model previously released by Google surpasses Llama 213B, and can be deployed on a PC.
Windows on Arm is ready, and the PC chip market is expected to reshape the PC chip market. Due to the streamlining of its instructions, it can reduce chip size, reduce power consumption, save costs, and improve efficiency, and is gradually breaking the “Wintel” pattern that has dominated the PC market for many years. Apple launched the first self-developed Mac chip M1 using the Arm architecture in 2020, beginning to leverage the x86 camp's dominant position in the PC market. We believe that Microsoft will place Copilot+ PC series products on ARM architecture chips, which is expected to accelerate the application of Arm architecture chips in the PC market.
Previously, Microsoft also tried putting Windows systems on Arm architecture chips, but no breakthrough was achieved due to application ecosystem compatibility issues. Now Microsoft has further simulated/translated x86 applications on Arm chips, created a new translation layer Prism, and added more optimizations to the bottom layer. This move can be compared to Apple's previous key transition from x86 to Arm using Rosetta 2. Based on the above, we believe that Microsoft has made thorough preparations for the launch of Windows on Arm and has fully taken into account previous failure experiences, and the PC chip market landscape is expected to usher in a turning point.
Deeply involved in the Arm architecture, Qualcomm is expected to advance rapidly in the PC market. Qualcomm's Arm architecture chips have previously been widely used in mobile phones, smart cockpits, and the Internet of Things, and the PC market is expected to become a new fertile ground for Qualcomm's Arm layout. With its first-mover advantage in AI chips, Qualcomm's AI PC chips have been used by many PC manufacturers such as Microsoft, HP, Samsung, Lenovo, Acer, Asus, and Dell. We are optimistic that Qualcomm will continue to improve in the PC field in the future.
The implementation of end-side AI is expected to bring new business growth impetus to Zhongke Chuangda. Since 2008, the company has established deep cooperation with Qualcomm on the Android system, accumulated rich cooperative development experience and capabilities, and formed a comprehensive independent intellectual property system from hardware drivers, operating system kernels, middleware to upper level applications. The continued implementation of end-side AI is expected to bring new growth drivers to the company's business:
1) On the chip manufacturer side, the company and chip manufacturers jointly establish and operate joint laboratories to provide services such as driver development, system optimization, testing and certification to help chip manufacturers establish relevant ecosystems; 2) On the terminal side, the deployment of large AI models on the end side is usually accompanied by next-generation hardware platforms, thus bringing about requirements for system development and hardware tuning.
At the end of 2023, the company achieved stable operation of the LLama-213 billion parameter model on edge devices equipped with the Qualcomm 8 series chip platform, becoming the first large-scale model enterprise in the industry to achieve this breakthrough; in March 2024, as an Intel ecosystem partner, the company was invited to attend the Intel 2024 new commercial client AIPC product conference and showed the latest results of its end-side model application on Intel Core Ultra processors, Rubik's Cube intelligent legal assistant; in May 2024, commercial AI at Lenovo At the PC Product Awards conference, the company implemented the application of AI PC Legal Assistant on the Lenovo ThinkPad X1 Carbon.
Profit forecasting and valuation
We maintain our previous profit forecast and forecast that the company's 2024-2026 revenue will be 60.17, 71.56, and 8.748 billion yuan, and net profit to mother of 587, 8.26, and 1,034 billion yuan, maintaining the “increase in holdings” rating.
Risk warning
The launch of new technologies and products fell short of expectations; increased market competition led to lower gross margins; downstream customer demand fell short of expectations; technology was compromised and core technical talent was lost.