Core Viewpoint - NVIDIA's entry into the CPU market with its DGX Spark has faced significant delays, raising concerns about its ability to compete against established players like Intel and AMD [1][10][16]. Group 1: Product Launch and Specifications - NVIDIA's DGX Spark, featuring the GB10 Grace Blackwell chip, was initially set to launch in July 2025 but has been delayed, with new expected shipping dates pushed to September 15 [3][4]. - The GB10 chip boasts a performance of approximately 1000 TOPS (FP4) and includes 128 GB of LPDDR5X unified memory, designed to meet the demands of AI model inference [4][9]. - The device is capable of running AI models with parameter scales of up to 200 billion at FP4 precision and 100 billion at FP8 precision, making it suitable for deploying specialized AI models [4][9]. Group 2: Production Challenges - High integration levels in the GB10 chip, which combines multiple cores and a GPU, have led to lower yield rates during mass production, complicating the manufacturing process [7][10]. - The production process involves complex steps, such as the CoWoS-L packaging by TSMC, which requires precise temperature control and can lead to delays if any step encounters issues [7][10]. Group 3: Market Competition and Pricing - The pricing for the base model of the GB10 is reported to be around £3600 (approximately 33,000 RMB), which, while lower than traditional NVIDIA DGX systems, may still be prohibitive for many developers [9][10]. - AMD has launched its Threadripper 9000 series processors, capturing a significant market share, with AMD's server CPU share reaching 39.4% in Q1 2025, indicating strong competition for NVIDIA [11][16]. Group 4: Software Ecosystem and Compatibility - NVIDIA faces significant challenges in building a software ecosystem for its Arm architecture CPUs, as many applications are not optimized for this architecture, leading to performance issues [14][20]. - The success of the GB10 will depend on NVIDIA's ability to ensure that essential software runs smoothly on its platform, as poor user experience could undermine its performance advantages [18][20]. Group 5: Strategic Partnerships and Future Outlook - NVIDIA's strategy includes collaborating with major PC manufacturers like ASUS and Dell to mitigate risks and share benefits, but delays have strained these partnerships [20][21]. - The company must consider forming closer alliances with software developers and operating system providers, such as Microsoft, to enhance the availability of native Arm applications and improve overall ecosystem compatibility [20][21].

Summary of Conference Call Notes Company and Industry Involved - The conference call primarily discusses developments in the **AI hardware sector**, particularly focusing on **NVIDIA** and its product offerings related to AI computing and data centers. Core Points and Arguments 1. **Blackwell Series Products**: The HGX series 8-card servers have been in production since last year, with deliveries starting in February. The GB200 cabinet is fully produced, and an upgrade to GB300 is expected in Q3 of this year [1][2] 2. **AI Factory Core Computing Unit**: The GB300 is positioned as a core computing unit for AI factories, supporting large-scale inference and training tasks. There have been significant upgrades compared to the GB200, although detailed specifics were not reiterated in this call [2] 3. **Production Challenges**: Q1 production rates were lower than expected due to assembly issues at ODM factories, leading to a downward revision of the annual cabinet shipment forecast [2][3] 4. **NVLink Fusion Technology**: This new technology allows customers to purchase only an NVLink Switch chip or NVLink Fusion IP, simplifying the procurement process for ASIC chips [3] 5. **DGX Spark and DGX Station**: The DGX Spark is aimed at personal supercomputer users, featuring NVIDIA's GB10 chip and supporting local model training. The DGX Station is a desktop-level AI supercomputer capable of running large models efficiently [4] 6. **AI Supercomputer in Taiwan**: NVIDIA plans to collaborate with TSMC and Foxconn to establish the first AI supercomputer in Taiwan, which is expected to be a cornerstone of the local AI ecosystem [5] 7. **RTX Pro Servers**: The RTX Pro servers, announced by ASUS, are designed to accelerate the transition of IT data centers to AI factories, boasting performance improvements over previous flagship systems [6] 8. **Software Ecosystem Expansion**: NVIDIA is also expanding its software ecosystem, launching various professional acceleration libraries aimed at standardizing AI acceleration capabilities across industries [7] 9. **Taiwan's Semiconductor Role**: Taiwan's advanced semiconductor manufacturing capabilities are crucial for NVIDIA's hardware deployment, fostering a deep collaboration in design, manufacturing, and application [8] 10. **Market Outlook**: The overseas computing power sector is gradually recovering, with companies in this space expected to release strong earnings this year. The computing PC sector is noted to be at a relatively low valuation [8] Other Important but Overlooked Content - The conference highlighted the ambition of NVIDIA to standardize and modularize AI acceleration capabilities across various industries, indicating a strategic direction towards broader applications of AI technology [7] - The establishment of a new NVIDIA office in Taiwan, named NVIDIA Constellation, signifies a commitment to local research and development, particularly in AI and semiconductor design [7][8]

Core Viewpoint - MediaTek is entering a harvest period with its personal AI supercomputer strategy, collaborating with NVIDIA on the GB10 super chip, which is gaining traction among major PC brands and is set for mass shipment this month [1][2] Group 1: Product Development and Market Impact - The GB10 super chip is expected to significantly contribute to MediaTek's non-mobile business, enhancing its performance in the AI sector [1] - NVIDIA's CEO Jensen Huang announced the collaboration at CES, highlighting the GB10 chip as a key component of the world's smallest AI supercomputer, Project DIGITS [1] - The DGX Spark, developed in partnership with NVIDIA, offers unprecedented AI performance of 1,000 TOPS and supports AI models with up to 200 billion parameters [1][2] Group 2: Strategic Partnerships and Future Outlook - Major PC brands including Acer, ASUS, Dell, Gigabyte, HP, Lenovo, and MSI are set to supply DGX Spark products starting in July, which is seen as beneficial for both NVIDIA's AI market position and MediaTek's expansion into AI [2] - MediaTek acknowledges the uncertainties brought by tariff changes and is focused on strengthening collaborations with global supply chain partners to enhance adaptability [2] - The company maintains a positive long-term outlook on the pervasive trend of AI, supported by its solid financial position and ongoing investments in key technologies [2]

Core Viewpoint - The article discusses the upcoming launch of ASUS's Ascend GX10 mini computer based on NVIDIA's GB10 Grace Blackwell platform, highlighting its potential in AI development and workstation applications [1][2]. Group 1: Product Launch and Features - ASUS is set to launch the Ascend GX10 mini computer on July 22, which aims to provide powerful capabilities for AI development [1]. - The GB10 Superchip system integrates a Grace CPU with 10 high-performance Arm Cortex-X925 cores and 10 low-power Cortex-A725 cores, along with a Blackwell GPU, delivering 1 PetaFLOPS of FP4 computing throughput [2]. - The platform supports 128GB of unified LPDDR5X memory with a bandwidth of 273 GB/s, comparable to Apple's M4 Pro memory subsystem [2]. Group 2: Market Positioning and Competitors - NVIDIA positions the GB10 platform as an AI solution with data center-level performance suitable for workstations and edge deployments, although the pricing remains undisclosed [2][3]. - ASUS's Ascend GX10 is expected to be similar in pricing to NVIDIA's DGX Spark system, which is priced at $3,000, with other manufacturers like Dell, HP, and Lenovo also preparing their versions [3]. Group 3: Performance Insights - NVIDIA emphasizes the unified memory architecture and high FP4 throughput as key advantages over traditional CPU-GPU configurations, making GB10 ideal for running LLM and generative AI applications [3]. - However, leaked Geekbench performance data suggests that GB10's general computing performance is comparable to Qualcomm's Snapdragon X Elite and Apple's M3 processor, raising concerns about its single-threaded performance for AI workloads [3]. Group 4: Future Prospects - NVIDIA has not confirmed whether it will offer the GB10 to other PC manufacturers, which could significantly impact the market [6]. - The GB10 is seen as a stepping stone for adapting to NVIDIA's more powerful Grace-Blackwell superchip, with potential future applications in gaming and graphics core products [6].

Investment Rating - The industry investment rating is "Recommended," indicating an expected increase in the industry index by more than 5% over the next 3-6 months compared to the benchmark index [37]. Core Insights - NVIDIA reported FY26Q1 revenue of $44.1 billion, a year-over-year increase of 69% and a quarter-over-quarter increase of 12%, significantly exceeding market expectations of $43.3 billion and company guidance of $43.0±2 billion. This growth was primarily driven by the data center business, which generated $39.1 billion in revenue, up 73% year-over-year and 10% quarter-over-quarter [3][7]. - The Blackwell architecture contributed approximately 70% of the data center computing revenue, marking the fastest ramp-up in GPU production in the company's history [4]. - The company expects FY26Q2 revenue to be $45.0 billion, with a potential loss of $8.0 billion in revenue due to recent export control restrictions affecting the H20 product line [5][8]. Summary by Sections 1. Performance Overview - FY26Q1 revenue reached $44.1 billion, with data center revenue at $39.1 billion, reflecting a 73% year-over-year growth. The GAAP and non-GAAP gross margins were 60.5% and 61.0%, respectively. Excluding a $4.5 billion expense, the non-GAAP gross margin would have been 71.3% [3][7]. - The diluted earnings per share were $0.76 (GAAP) and $0.81 (non-GAAP), with a potential adjusted non-GAAP EPS of $0.96 when excluding the aforementioned expense [3][7]. 2. Business Segment Performance - **Data Center**: Revenue reached a record high of $39.1 billion, with computing revenue at $34.2 billion (up 76% YoY) and networking revenue at $4.957 billion (up 56% YoY) [4]. - **Gaming**: Revenue was $3.763 billion, showing a 42% year-over-year increase, driven by strong adoption of Blackwell architecture GPUs [4]. - **Professional Visualization**: Revenue was $509 million, with a 19% year-over-year increase, although it remained flat quarter-over-quarter due to tariff-related uncertainties [4]. - **Automotive and Robotics**: Revenue was $567 million, reflecting a 72% year-over-year increase, driven by strong demand for autonomous driving and electric vehicles [4]. 3. Future Guidance - The company anticipates FY26Q2 revenue of $45.0 billion, accounting for an estimated $8.0 billion loss in H20 revenue due to export restrictions. Expected gross margins are projected at 71.8% (GAAP) and 72.0% (non-GAAP) [5][8].

Core Viewpoint - NVIDIA's CEO Jensen Huang expressed concern over the H20 export restrictions impacting the company's access to the Chinese AI market, which is valued at $50 billion, while highlighting the robust demand for AI processing capabilities driven by the Blackwell chip production [1][8][45]. Group 1: Financial Performance and Market Impact - NVIDIA's Q1 revenue reached $44 billion, a 69% year-over-year increase, despite the challenges posed by export restrictions [25]. - The company anticipates a loss of $8 billion in H20 revenue due to new export limitations, significantly affecting future business prospects in the Chinese market [8][43]. - The data center revenue grew by 73% year-over-year, driven by the rapid ramp-up of the Blackwell product line [5][27]. Group 2: AI Demand and Technological Advancements - There is an explosive growth in AI inference demand, with token generation increasing by 500% year-over-year, particularly in complex AI workloads [12][29]. - The Blackwell architecture is designed to support this demand, offering a throughput that is 40 times higher than the previous Hopper architecture [12][10]. - The average deployment rate for major hyperscale customers is nearly 1,000 NVL72 racks per week, indicating strong market adoption [10][28]. Group 3: Strategic Insights on AI Market - Huang emphasized that winning the Chinese AI market is crucial for global leadership, as it houses half of the world's AI researchers [3][45]. - The company is exploring options to create attractive solutions for the Chinese market in light of the export restrictions [8][46]. - The rise of open-source AI models like DeepSeek and Qwen is seen as a strategic advantage for the U.S. in maintaining its leadership in AI technology [13][46]. Group 4: Future Outlook and Growth Engines - NVIDIA is optimistic about future growth, citing multiple key growth engines including surging inference demand, sovereign AI initiatives, and enterprise AI [19][49]. - The company plans to achieve $45 billion in revenue for Q2, with expected gross margins of 71.8% [20][43]. - The establishment of AI factories globally is seen as a foundational step in building the necessary infrastructure for AI deployment across industries [15][62].

Core Insights - Nvidia reported better-than-expected revenue and profit, driven by strong growth in its data center business, which increased by 73% year-over-year [1][2] - The company’s adjusted earnings per share were $0.96, exceeding the expected $0.93, while revenue reached $44.06 billion, surpassing the forecast of $43.31 billion [1] - Nvidia's net profit grew by 26% year-over-year, rising from $14.9 billion ($0.60 per share) to $18.8 billion ($0.76 per share) [1] Financial Performance - Q1 revenue increased by 69% year-over-year, from $26 billion to $44.06 billion [1] - The data center segment accounted for 88% of total revenue, with sales reaching $39.1 billion [1] - The company spent $14.1 billion on stock buybacks and distributed $244 million in dividends during the quarter [1] Market Response - Following the earnings report, Nvidia's stock price rose approximately 6% in after-hours trading, nearing its historical high set in January [1] Future Guidance - Nvidia expects revenue for the upcoming quarter to be around $45 billion, slightly below the LSEG forecast of $45.9 billion [1] - The company indicated that without recent export restrictions on its H20 chip, guidance could have been higher by approximately $8 billion [1][2] Export Restrictions Impact - The U.S. government informed Nvidia that it now requires export licenses for the previously approved H20 processors to China, leading to a $4.5 billion charge due to excess inventory [2] - If not for the export restrictions, Nvidia could have achieved an additional $2.5 billion in sales [2] Profitability Metrics - The gross margin for the quarter was 61%, which would have been 71.3% without the costs associated with the China-related issues [2] AI Demand and Business Segments - Nvidia's CEO highlighted strong global demand for AI infrastructure, driven by applications like OpenAI's ChatGPT [3] - Major cloud service providers contributed nearly half of the data center revenue, with network product sales reaching $5 billion [3] - The gaming segment saw a 42% year-over-year revenue increase, totaling $3.8 billion, while the automotive and robotics segment grew by 72% to $567 million [4]

在之前的文章中，几次都写到了HBM相关的内容，我们后面也将持续跟踪存储器市场，包括NAND 和DRAM。在这些文章中，我们会提供最近的存储器市场的数据，以及大机构的预测趋势，包括各 个厂商（包括长鑫和长江存储）的市场占有率、出货量、成品晶圆和晶粒的库存、晶圆出货量等。 这篇文章参考的内容主要来自UBS的研报和数据。我们参考的内容都会放到星球中。 三星与SK海力士对第二季度的指引反映出一定程度的需求前置，尤其是来自智能手机和PC客户的需 求。两家公司均预计2Q25 DRAM出货量将环比增长小于10%。对于NAND，三星预计Q2出货量环比 增长大概5%左右，而SK海力士则超过20%。 UBS维持对DDR定价在2Q25环比增长5%的预测，其中LPDDR5的势头持续强于DDR5和 DDR4/LPDDR4。对于NAND均价，UBS将Q2预测从+5%下调至+3%。尽管嵌入式NAND闪存需求仍 相对积极，但SSD价格上涨正面临客户更强烈的抵制。 展望2025年下半年，由于关税不确定性，存储厂商的能见度低于正常水平。UBS继续认为NAND闪 存需求更可能面临规格降级（或客户通过套件调整压低内容成本，本质相同）。因此，UBS预 ...

Investment Rating - The industry investment rating is "Positive" [2][29] Core Insights - NVLink Fusion builds a semi-custom AI infrastructure, enhancing the ecosystem's moat. Nvidia's CEO Jensen Huang announced NVLink Fusion at COMPUTEX 2025, marking the opening of Nvidia's proprietary high-performance interconnect technology NVLink to partners for integrating third-party CPUs and AI accelerators, thus creating a semi-custom AI infrastructure. This aims to overcome traditional data center bottlenecks in scale and performance, providing more flexible and optimized system design solutions for cloud service providers and large enterprises [6][11] - The GB300 is expected to launch in Q3 2025, with multiple personal and enterprise products announced. The new AI computing platform Grace Blackwell and its upgraded version GB300 were introduced, with GB300 offering 1.7 times the inference performance of the previous H100, equipped with 1.5 times HBM memory and 2 times network bandwidth, achieving up to 40 petaflops per node [13][23] - The data center market is transitioning to a nearly trillion-dollar market driven by AI factories and infrastructure. Nvidia's CEO stated that the data center is on the verge of becoming a trillion-dollar market, driven by AI factories and infrastructure. The expansion of AI infrastructure investment is expected to increase orders for quality companies in the domestic AI industry chain, with fundamentals likely to continue to deliver [23] Summary by Sections NVLink Fusion - NVLink Fusion provides two main configurations: one connects third-party custom CPUs with NVIDIA GPUs via NVLink, and the other connects NVIDIA's Grace series CPUs with non-NVIDIA custom accelerators (GPU, ASIC, FPGA) to meet various computational needs. Initial adopters of NVLink Fusion include MediaTek, Marvell, Alchip, Astera Labs, Synopsys, and Cadence [7][11] GB300 Launch - The GB300 is set to launch in Q3 2025, with significant performance improvements over its predecessor. The Blackwell system is expected to start mass production by the end of 2024 and has already been deployed on platforms like CoreWeave [13][23] Market Transformation - The report emphasizes the transformation of the data center market into a trillion-dollar industry, highlighting the impact of AI infrastructure and factory investments on the domestic AI industry chain [23]

Group 1 - ByteDance issued a compliance notice urging business partners not to give gifts or cash to employees, emphasizing a zero-tolerance policy towards corruption and bribery [2] - Kuaishou faced allegations of requiring employees to use its app for one hour daily, which was later denied by internal sources, stating that while usage is encouraged, it is not mandatory [3] - Kunlun Wanwei's newly launched AI product experienced high user traffic leading to service limitations, indicating strong initial demand [4] Group 2 - The co-founder of Zero One Everything, Gu Xuemai, has left the company to pursue new entrepreneurial ventures, as the company shifts its focus towards lightweight model training and application [5] - A paper published in a top journal was found to contain inappropriate language, prompting an investigation by the journal [6][7] - Elon Musk announced his return to a 24/7 work schedule, emphasizing the need for operational improvements at X and Tesla [9][10] Group 3 - NVIDIA's Blackwell GPU set a new record for AI inference speed, achieving 1000 tokens per second per user, showcasing advancements in AI processing capabilities [11] - Apple plans to open its AI models to third-party developers to stimulate new application development, aiming to enhance its competitive position in the AI market [12] - OpenAI is acquiring AI device company io for $6.5 billion, marking its largest acquisition to date and expanding its hardware capabilities [13] Group 4 - JD.com is investing in ZhiYuan Robotics, indicating strong interest in the embodied intelligence sector, with the company positioned among the top players in this field [14] - Google announced the launch of Google AI Ultra, a comprehensive AI suite aimed at enhancing productivity across various industries [18][19] - Tencent introduced a smart agent development platform and plans to open-source multiple models, reflecting its commitment to advancing AI technology [21][22]