Investment Rating - The report maintains a positive outlook on the computer industry, indicating a "Look Favorably" investment rating for the sector [6][7]. Core Insights - The computer industry is experiencing a low holding ratio, with public fund allocation at 2.6% in Q2 2025, down 0.6 percentage points from the previous quarter, ranking 13th among 30 primary industries [8][9]. - AI remains the main theme for the computer sector throughout 2025, supported by three key factors: the introduction of domestic super-node solutions improving cost-performance, the launch of several foundational large models driving AI applications into commercialization, and continuous innovations across various fields such as stablecoins and 3D printing [9][11]. - The report highlights significant company updates, particularly the official upgrade of iFLYTEK's reasoning large model X1, which enhances capabilities in multiple languages and applications [38][43]. Summary by Sections Investment Allocation - In Q2 2025, the computer industry's public fund allocation decreased to 2.6%, marking a historical low since 2010, with a configuration coefficient of 0.56, down from 0.67 in Q1 2025 [8][9]. - The report suggests increasing positions in Hong Kong-listed computer stocks such as Kingdee and Meitu [6][7]. AI Development - The report identifies three main drivers for the future performance of the computer industry: 1. The launch of domestic super-node solutions that enhance cost-performance and reduce the gap with overseas solutions [9][10]. 2. The introduction of multiple foundational large models that facilitate the commercialization of AI applications [10][11]. 3. Ongoing innovations in various sectors, including stablecoins and 3D printing, which are expected to gain traction [11][12]. Valuation Metrics - As of July 22, 2025, the computer industry’s PE (TTM) stands at 85.4x, placing it in the 93.40% historical percentile, while the PS (TTM) is at 3.4x, in the 48.90% historical percentile [24][25]. - The report notes that current valuation levels exceed those of 2020 and 2023, reflecting optimistic market expectations regarding potential profitability [24][25]. Company Updates - iFLYTEK's reasoning large model X1 has been officially upgraded, showcasing improvements in comprehensive capabilities and multi-language support, with applications in education, healthcare, and enterprise solutions [38][43]. - The report emphasizes the growth trend in AI revenue for iFLYTEK, with significant increases in both consumer and enterprise AI solutions [44]. Market Dynamics - The report discusses the varying rhythms of different technology sectors, influenced by the certainty and traceability of new technologies, with AI applications expected to follow a similar trajectory to cloud computing [36][37]. - The report anticipates a rapid increase in market capitalization for AI-related companies as performance metrics begin to materialize in the latter half of 2025 [37][38].

Core Viewpoint - The article discusses the advancements in AI data center architecture, particularly focusing on Huawei's CloudMatrix384, which aims to address the limitations of traditional AI clusters by providing a more efficient, flexible, and scalable solution for AI computing needs [5][12][49]. Group 1: AI Computing Demand and Challenges - Major tech companies are significantly increasing their investments in GPU resources to enhance AI capabilities, with examples like Elon Musk's plan to expand his supercomputer by tenfold and Meta's $10 billion investment in a new data center [1]. - Traditional AI clusters face challenges such as communication bottlenecks, memory fragmentation, and fluctuating resource utilization, which hinder the full potential of GPUs [3][4][10]. - The need for a new architecture arises from the inability of existing systems to meet the growing computational demands of large-scale AI models [10][11]. Group 2: Huawei's CloudMatrix384 Architecture - Huawei's CloudMatrix384 represents a shift from simply stacking GPUs to a more integrated architecture that allows for high-bandwidth, peer-to-peer communication and fine-grained resource decoupling [5][7][14]. - The architecture integrates 384 NPUs and 192 CPUs into a single super node, enabling unified resource management and efficient data transfer through a high-speed, low-latency network [14][24]. - CloudMatrix384 achieves impressive performance metrics, such as a throughput of 6688 tokens/s/NPU during pre-fill and 1943 tokens/s/NPU during decoding, surpassing NVIDIA's H100/H800 [7][28]. Group 3: Innovations and Technical Advantages - The architecture employs a peer-to-peer communication model that eliminates the need for a central CPU to manage data transfers, significantly reducing communication overhead [18][20]. - The UB network design ensures constant bandwidth between any two NPUs/CPUs, providing 392GB/s of unidirectional bandwidth, which enhances data transfer speed and stability [23][24]. - Software innovations, such as global memory pooling and automated resource management, further enhance the efficiency and flexibility of the CloudMatrix384 system [29][42]. Group 4: Cloud-Native Infrastructure - CloudMatrix384 is designed with a cloud-native approach, allowing users to deploy AI applications without needing to manage hardware intricacies, thus lowering the barrier to entry for AI adoption [30][31]. - The infrastructure software stack includes modules for resource allocation, network communication, and application deployment, streamlining the process for users [33][40]. - The system supports dynamic scaling of resources based on workload demands, enabling efficient utilization of computing power [45][51]. Group 5: Future Directions and Industry Impact - The architecture aims to redefine AI infrastructure by breaking the traditional constraints of power, latency, and cost, making high-performance AI solutions more accessible [47][49]. - Future developments may include expanding node sizes and further decoupling resources to enhance scalability and efficiency [60][64]. - CloudMatrix384 exemplifies a competitive edge for domestic cloud solutions in terms of performance and cost-effectiveness, providing a viable path for AI implementation in Chinese enterprises [56][53].

Summary of Huawei CloudMatrix384 Architecture and Performance Analysis Industry and Company - **Industry**: AI Infrastructure - **Company**: Huawei Core Points and Arguments 1. **Comparison with NVIDIA**: The report provides a comprehensive technical and strategic evaluation of Huawei's CloudMatrix384 AI cluster compared to NVIDIA's H100 cluster architecture, highlighting fundamental differences in design philosophy and system architecture [1][2][3] 2. **Architecture Philosophy**: Huawei's CloudMatrix384 adopts a radical, flat peer-to-peer architecture, utilizing a Unified Bus (UB) network that eliminates performance gaps between intra-node and inter-node communications, creating a tightly coupled computing entity [2][3] 3. **Performance Metrics**: The CloudMatrix-Infer service on Ascend 910C outperforms NVIDIA's H100 and H800 in terms of computational efficiency during the pre-fill and decode phases, showcasing Huawei's "system wins" strategy [3] 4. **Challenges**: Huawei faces significant challenges with its CANN software ecosystem, which lags behind NVIDIA's CUDA ecosystem in terms of maturity, developer base, and toolchain richness [3][4] 5. **Targeted Optimization**: CloudMatrix384 is not intended to be a universal replacement for NVIDIA H100 but is optimized for specific AI workloads, marking a potential bifurcation in the AI infrastructure market [4][5] Technical Insights 1. **Resource Decoupling**: The architecture is based on a disruptive design philosophy that aims to decouple key hardware resources from traditional server constraints, allowing for independent scaling of resources [6][7] 2. **Unified Bus Network**: The UB network serves as the central nervous system of CloudMatrix, providing high bandwidth and low latency, crucial for the performance of the entire system [8][10] 3. **Non-blocking Topology**: The UB network creates a non-blocking all-to-all topology, ensuring nearly consistent communication performance across nodes, which is vital for large-scale parallel computing [10][16] 4. **Core Hardware Components**: The Ascend 910C NPU is the flagship AI accelerator, designed to work closely with the CloudMatrix architecture, featuring advanced packaging technology and high memory bandwidth [12][14] 5. **Service Engine**: The CloudMatrix-Infer service engine is designed for large-scale MoE model inference, utilizing a series of optimizations that convert theoretical hardware potential into practical application performance [17][18] Optimization Techniques 1. **PDC Decoupled Architecture**: The architecture innovatively separates the inference process into three independent clusters, enhancing scheduling and load balancing [18][19] 2. **Large-scale Expert Parallelism (LEP)**: This strategy allows for extreme parallelism during the decoding phase, effectively managing communication overhead with the support of the UB network [22][23] 3. **Hybrid Parallelism for Prell**: This approach balances load during the pre-fill phase, significantly improving throughput and reducing idle NPU time [24] 4. **Caching Services**: The Elastic Memory Service (EMS) leverages all nodes' CPU memory to create a unified, decoupled memory pool, enhancing cache hit rates and overall performance [24][29] Quantization and Precision 1. **Huawei's INT8 Approach**: Huawei employs a complex, non-training-dependent INT8 quantization strategy that requires fine calibration, contrasting with NVIDIA's standardized FP8 approach [30][31] 2. **Performance Impact**: The report quantifies the contributions of various optimization techniques, highlighting the significant impact of context caching and multi-token prediction on overall performance [29][30] Conclusion - The analysis indicates that Huawei's CloudMatrix384 represents a significant shift in AI infrastructure design, focusing on specific workloads and leveraging a tightly integrated hardware-software ecosystem, while also facing challenges in software maturity and market penetration [4][5][30]

Group 1: Macro Insights - The Federal Reserve maintained the federal funds rate target range at 4.25%-4.5%, marking the fourth consecutive meeting without changes, aligning with market expectations. However, inflationary concerns have intensified, leading to downward revisions in economic growth forecasts for 2025 and 2026, alongside an increase in unemployment rate predictions and price index forecasts [2][10][11] - The impact of tariffs on inflation has not yet fully materialized, indicating significant uncertainty regarding future inflation trends. Tariff measures require time to affect consumer prices, and geopolitical issues in the Middle East may further exacerbate inflation [2][10][11] - The market is currently exhibiting signs of stagflation trading, with expectations of a potential recovery trading phase in the latter half of the year as tax reduction measures and debt ceiling increases are implemented [3][12] Group 2: Nuclear Fusion Industry - Shanghai Superconductor's IPO application has been accepted, signaling an acceleration in the industrialization of nuclear fusion. The company is a leading producer of high-temperature superconducting materials, holding over 80% of the domestic market share for second-generation high-temperature superconducting tapes [5][20][22] - The global market for high-temperature superconducting materials is projected to grow from 790 million yuan in 2024 to over 10.5 billion yuan by 2030, driven by applications in controllable nuclear fusion and other downstream industries [6][22][23] - Shanghai Superconductor's revenue is expected to grow significantly, with projections of 240 million yuan in 2024, representing a year-on-year increase of 187.4%. The company is anticipated to achieve profitability in 2024 after previous losses [6][22][23] Group 3: Automotive Industry - The heavy truck market in China is showing signs of recovery, with a projected 16% year-on-year increase in sales to 1.06 million units in 2025, driven by the implementation of the vehicle replacement policy [17][18] - In May 2025, domestic heavy truck sales reached 89,000 units, reflecting a year-on-year increase of 13.6%. The market is expected to benefit from the ongoing vehicle replacement initiatives [18][19] Group 4: Chemical Industry - The demand for photoinitiators is increasing due to their expanding application scenarios, leading to rising product prices. Key companies in this sector include Jiuri New Materials, Yangfan New Materials, and Qiangli New Materials [34][35] - The photoinitiator market is expected to grow rapidly, driven by environmental regulations and the emergence of new applications such as 3D printing [35]

Summary of Conference Call Notes Industry Overview - The conference call primarily discusses the **cloud service provider (CSP)** industry and advancements in **AI connectivity** technology. Key Points and Arguments CSP Capital Expenditure - **Overseas CSP capital expenditure** is better than market expectations, with the top four CSPs spending a total of **$71.1 billion** in Q1 2025, representing a **59% year-over-year increase** [1] - Individual expenditures include: - **Microsoft**: **$15.8 billion**, up **59%** - **Google**: **$17.2 billion**, up **43%** - **Amazon**: **$24.3 billion**, up **62%** - **Meta**: **$12.9 billion**, up **93%** [1] - Meta has raised its full-year capital expenditure plan for 2025 to a range of **$64 billion to $72 billion**, up from the previous estimate of **$60 billion to $65 billion** [1] - According to Bloomberg, the capital expenditure growth rate for these four CSPs is projected to reach **40%** in 2025 [1] AI Connectivity Breakthroughs - **Huawei** has launched the **CloudMatrix 384**, which consists of **384 Ascend 910C computing cards**, making it the largest single-node scale among currently commercialized supernodes [1] - The **DeepSeek-R1** service, based on the CloudMatrix 384 supernode, has been officially launched in collaboration with **Silicon-based Flow** and **Huawei Cloud**. It guarantees a single-user throughput of **20 TPS** while achieving a decoding throughput of **1920 Tokens/s**, comparable to the performance of **H100 deployments** [2] Valuation and Investment Opportunities - The **computing power industry chain** is viewed as having a favorable valuation with potential for recovery. Companies mentioned include: - **New Yisheng** - **Zhongji Xuchuang** - **Tianfu Communication** - **Taicheng Technology** - **Bochuang Technology** - **Yingweike** - **Chunzhong Technology** - **Huafeng Technology** - **Oulutong** - **Yihua Co.** - **Unisplendour** - **Shenling Environment** - **Gaolan Co.** - **Yingweike** - **Guanghuan New Network** - **Runze Technology** [3] Risk Factors - A key risk highlighted is the potential for **AI application development** to fall short of expectations [4]