Summary of Huawei's AI Roadmap and Implications for the Semiconductor Industry Industry Overview - The focus is on the **semiconductor industry** in China, particularly regarding **Huawei's AI ambitions** and its impact on the local semiconductor ecosystem [1][2][3]. Key Points and Arguments Huawei's AI Roadmap - Huawei unveiled its AI roadmap on **September 18, 2025**, which includes: - Next-generation **Ascend AI chip roadmap** - Open-source **UnifiedBus (UB) protocol** for connecting numerous chips in a **SuperPoD** - Development of the most powerful **SuperPoDs and SuperClusters** globally [1]. - The roadmap indicates a need for **16 times more chips**, which is expected to be inflationary for **Foundry/WFE** and benefits the entire semiconductor value chain [1]. Confidence in Local Foundry Supply - Huawei's public articulation of its AI roadmap signals confidence in the resilience of its local foundry supply, contrasting its previous cautious approach post-U.S. sanctions [2]. - The company appears to have secured reliable manufacturing capabilities, marking a significant milestone in building a robust local semiconductor ecosystem [2]. Demand for Advanced Logic Capacity - The ambitious scale of Huawei's AI SuperPoDs and SuperClusters indicates a **surge in demand** for local advanced logic capacity [3]. - Huawei's **Ascend 950 chip** has only **6%** of the performance of Nvidia's **VR200**, but through innovative networking protocols, Huawei can deploy **114 times more chips** than Nvidia in a single SuperPoD, achieving **6.8 times higher total computing power** [3]. Positive Impacts on Local Semiconductor Companies - The developments are positive for the entire local AI supply chain, benefiting: - **Foundries** like **SMIC**, which is expanding its **7nm capacity** to support Huawei's production [4]. - **Semicap vendors** such as **NAURA**, **AMEC**, and **Piotech**, which are expected to see substantial benefits from capacity expansion plans [4]. - **Hygon**, a leading domestic x86 server CPU provider, may face increased competition from Huawei but remains a necessary alternative [4]. Investment Implications - **SMIC** is rated **Outperform** with a price target of **HKD 30** for H-shares and **CNY 110** for A-shares, based on a valuation of **1.5x NTM P/B multiple** [6]. - **Hua Hong** is also rated **Outperform** with a price target of **HKD 60** for H-shares and **CNY 85** for A-shares, driven by its acquisition of Fab 5 [7]. - **NAURA**, **AMEC**, and **Piotech** are rated **Outperform**, benefiting from domestic WFE substitution in China [8][9][10]. - **Hygon** is rated **Outperform** with a price target of **CNY 220**, leveraging its AI accelerator for growth [11]. - **Cambricon**, while a leading AI accelerator ASIC chip provider, is rated **Market-Perform** due to high valuation concerns [12]. Additional Important Insights - The local AI production capacity is projected to **triple by 2026**, indicating strong capital expenditure trends in China [3]. - The developments create a **virtuous cycle of innovation and investment** that will strengthen China's AI ecosystem for years to come [4]. This summary encapsulates the critical insights from Huawei's AI roadmap and its implications for the semiconductor industry, highlighting the potential for growth and investment opportunities within the local ecosystem.

Core Viewpoint - The U.S. government is taking measures to limit China's advancements in AI chip production, with specific emphasis on Huawei's limited capabilities in this area [1][2]. Group 1: U.S. Government Actions - Jeffrey Kessler, Deputy Assistant Secretary for Export Control, testified that Huawei's ability to produce AI chips is extremely limited, alleviating concerns about China's rapid advancements in the semiconductor field [1]. - U.S. Commerce Secretary Gina Raimondo stated that the U.S. has never considered providing China with the most advanced chips, emphasizing that high-end chips from companies like NVIDIA will not be exported to China [1]. - Kessler highlighted that by 2025, Huawei may only produce up to 200,000 Ascend AI chips, primarily for the Chinese market, while the demand for AI accelerators in China is estimated to be around 1.5 million units in 2024 [1]. Group 2: Budget and Resource Allocation - Kessler called for an increase in the budget for the Bureau of Industry and Security (BIS) to enhance export control enforcement, suggesting that the enforcement budget should be doubled [1]. - If the budget is approved, BIS plans to hire 200 new export enforcement agents and increase the number of overseas export control officials from 12 to over 30 [1][2].

Core Viewpoint - Nvidia plans to invest up to $500 billion in purchasing chips and electronic products manufactured in the United States over the next four years, in response to trade policies and supply chain concerns [1][2]. Group 1: Investment and Supply Chain Strategy - Nvidia's CEO, Jensen Huang, announced a significant investment strategy to shift the supply chain from Asia back to the U.S., emphasizing the ability to produce billions of products domestically [1]. - The company is currently producing its latest Blackwell systems in the U.S., bolstered by TSMC's $100 billion investment in Arizona, which enhances supply chain resilience [2]. - Huang expressed confidence in the U.S. government's support for the AI industry, which he believes will accelerate its development [2]. Group 2: Competitive Landscape - Nvidia faces increasing competition from Huawei, particularly in the AI chip sector, despite generating billions in revenue from China [3][4]. - Huang acknowledged Huawei as a formidable competitor, noting their success in various markets and the ineffectiveness of U.S. efforts to constrain the company [4]. - Intel is identified as the only U.S. company theoretically capable of producing chips similar to Nvidia's, but it faces significant challenges in its foundry business [3][5]. Group 3: Future Outlook - Huang emphasized the importance of Intel's success for the overall health of the semiconductor industry, while also indicating that establishing new supply chains will take time [5]. - The ongoing geopolitical tensions and trade policies are likely to shape the future landscape of the semiconductor industry, influencing investment decisions and competitive dynamics [1][2].