Open source securities: PCs, with high-performance hardware and productivity properties, are expected to become the first stop for AI edge model landing.

According to the research report released by Open Source Securities, the DeepSeek R1 model has received widespread attention recently. It significantly reduces development costs through architectural and algorithmic innovation, achieving ideal model performance. It also supports local deployment of different levels of distilled models (currently supporting 1.5/7/14/32/70/671B parameters), aiming to retain the capabilities of large models as much as possible while "shrinking" them. With the promotion of models like DeepSeek for local deployment on PC, it is expected to drive users to experience more efficient and personalized AI services. Against this background, it is emphasized that model capabilities determine product experience, and the qualitative change in PC product experience is expected to drive a wave of device upgrades. The main points of Open Source Securities are as follows: Why is the first stop for local deployment on PC: Strong hardware capabilities and productivity attributes 1. PC: The terminal with the most abundant hardware performance. PCs have strong scalability and can be equipped with higher performance GPUs, NPUs, etc., to efficiently process complex calculations. They also have more space to allocate more storage and memory to ensure smooth operation of the model, while devices like phones and smart glasses are more limited. 2. PC: AI + productivity directly address consumer pain points. PCs contain productivity tools and personal assistants, enabling AIPC to quickly process documents and analyze data during work, plan schedules and manage tasks in life, and provide inspiration and efficiently complete designs during creation. For example, the Douyin 2024 text-life video model was released in September, with devices like phones being relatively limited in the above scenarios. Why do consumers need AI PCs: Demand for personalized and complex production The core of AIPC is to solve personalized problems more customized, efficiently, and securely by running large models locally. Previously, AIPC could already complete tasks such as text-life text, text-life image, automatic report generation, and AI local knowledge base, but usually adopted a cloud + end hybrid model approach due to model limitations. However, the performance of models distilled from DeepSeek R1 has been further improved. For example, DeepSeek-R1-Distill-Qwen-32B achieved 72.6% and 94.3% scores on AIME2024 and MATH-500, respectively, surpassing OpenAI-o1-mini significantly on most benchmarks. Furthermore, brands like Lenovo and Huawei have integrated Deepseek into their systems to provide users with more intelligent and convenient AI interactive experiences. Industry chain impact speculation: 1. Integrated Machines: Integrated machine manufacturers play a decisive role in the landing of AIPC, completing technical convergence and industrial integration to deliver a complete AIPC experience. They are expected to benefit from the machine replacement demand brought about by the penetration of AIPC. Storage: The penetration of AIPC is expected to drive growth in storage demand and performance upgrades. According to Micron, the Llama2 model with 70B parameters and 4-bit precision requires approximately 42GB of memory to produce 1.4 Tokens/s output. Models with less than 10B parameters also require about 2GB of incremental memory. Observing the Lenovo Meteq series AIPC, 32GB of memory + 1TB flash memory is standard. Heat dissipation: The increase in AI power and power consumption drives the demand for heat dissipation, with high-value solutions like heat boards/graphite heat sinks expected. PCB: Low loss + HDI are two major trends in PCB. Structural components: The trend of lightweight high-endization is evident, and AI also promotes the definition of PCs, such as the foldable screen PC expected to be released by Huawei in March. Beneficiary targets 1. Brand machines and ODM: LENOVO GROUP (00992), Huaqin Technology (603296.SH), Luxshare Precision Industry (002475.SZ), Emdoor Information (001314.SZ), etc. 2. Heat dissipation: Shenzhen FRD Science & Technology (300602.SZ), Jones Tech Plc (300684.SZ), Guangdong Suqun New Material (301489.SZ), Long Young Electronic (301389.SZ), etc.; PCB: Aoshikang Technology (002913.SZ), Victory Giant Technology (300476.SZ), etc. 3. Structural components: Suzhou Chunqiu Electronic Technology (603890.SH), Lingyi Itech (002600.SZ), Lens Technology (300433.SZ), Shenzhen Bromake New Material (301387.SZ), etc. 4. Parts: Shenzhen Laibao Hi-Tech (002106.SZ), Highbroad Advanced Material (301321.SZ), Shenzhen Hui Chuang Da Technology (300909.SZ), etc. 5. Chips: Hygon Information Technology (688041.SH), Loongson Technology Corporation (688047.SH), Chipsea Technologies (Shenzhen) Corp., Ltd (688595.SH), Dosilicon Co., Ltd. (688110.SH), Giantec Semiconductor Corporation (688123.SH), Montage Technology (688008.SH). Risk Warning:AIPC penetration fell below expectations, AIPC product experience fell below expectations, downstream demand fell below expectations.