2026 年初的资本市场，显得有些拥挤。 如果你最近盯着科创板和港股的公告看，会发现一个很有趣的现象，中国的芯片公司正在排队敲钟。摩 尔线程正式挂牌，燧原科技紧随其后，甚至连大厂内部孵化多年的阿里平头哥、 百度昆仑芯 ，也纷纷 传出了分拆独立上市的消息。 这种场面，像极了当年互联网最狂热的时候。 但是，如果你仔细观察这些公司的故事，会发现一个尤其荒诞的现象。在面对投资人时，每一家都恨不 得把"中国英伟达"的字样写到PPT上，但在实际的业务闭门会上，他们却在拼命撕掉这个标签，走上一 条和英伟达完全相反的路。 这就是 2026 年中国芯片界最大的公开秘密，没人想做"中国英伟达"，但在上市之前，人人都得穿上这 件衣服。 这是一场关于IPO估值的精巧表演，也是一场关于生存的战略撤退。 这波芯片上市潮背后的真实算盘，到底是什么？ 1、人人都想穿上英伟达的马甲 为什么这些明明走着不同路径的公司，在宣传时非要挂着英伟达这三个字。 原因非常简单：为了拉高估值。 在资本市场的估值逻辑里，英伟达是神。它的毛利、市场占有率和 CUDA 生态带来的高护城河，让它 拥有了接近 40 倍甚至更高的市盈率。 对于一个中国芯片创业公司来说， ...

Group 1 - Major manufacturers in East China have halted B200 leasing orders and shifted focus to B300 models, leading to a significant equipment iteration trend in the computing power leasing market [1] - The halt of B200 orders has not significantly impacted the flow of B200 units in the market, as existing inventory remains tight, with only a few units available in certain regions [1] Group 2 - The announcement allowing NVIDIA to export H200 chips to approved Chinese customers has led to a market stalemate, with many companies choosing to pause orders due to uncertainty in policy direction and government regulations [2] - The price of H200 modules has reportedly dropped from over 1.5 million yuan to 1.25 million yuan, although skepticism remains regarding the sustainability of this price drop due to rising memory costs and export fees [3][4] Group 3 - Domestic AI chip companies have turned to public listings after failed acquisition attempts by major industry players, with many now listed on the Sci-Tech Innovation Board or the Hong Kong Stock Exchange [6] - A state-owned computing power platform is assembling a high-profile executive team to reclaim technological sovereignty, leveraging its resources to access data from high-barrier sectors like finance and healthcare [7][8] Group 4 - A major internet company in North China has placed an order for over 30,000 NVIDIA L20 and L40 chips, indicating that older models still hold value in specific business scenarios despite claims of obsolescence [9] - The price of NVIDIA RTX 5090 graphics cards has surged significantly, with reports of price increases driven by rising demand and component costs, potentially as a strategy to shift demand towards the newly approved H200 chips [10] Group 5 - Zhonghao Xinying is reportedly implementing "minimum usage rate commitment" clauses in sales contracts to stabilize order expectations, raising concerns about the true market performance of its products [11] - The gross margin of Runze Technology reached 48.11% in the first three quarters of 2025, significantly higher than the industry average of 19%-25%, driven by early investments in computing power equipment [13] Group 6 - The domestic computing power project landscape is heating up, with major server manufacturers actively engaging in multiple projects, although challenges remain in service provision for smaller-scale clusters [14] - The separation of roles between funding and operational parties in new computing projects has led to a trend of "100% buyout" contracts becoming standard, with a common expectation of recouping investments within five years [15]

Core Insights - The current state of the computing power market in China is characterized by extreme supply-demand imbalance, with high-end GPU chips like the Nvidia B200 being nearly invisible in the domestic market, leading to significant price increases for both high-end and mid-range GPUs [1][2][7] - The surge in demand for computing power coincides with the listing of several domestic AI chip companies, indicating a hot market environment at the beginning of 2026 [1][11] - Speculative investments have exacerbated market volatility, with some suppliers shifting from graphics cards to memory products due to higher profit margins and lower risks [4][6] Supply and Demand Dynamics - The demand for computing power is currently high, but resources are scarce, leading to increased prices for GPUs and memory components [2][3] - Prices for DDR5 memory have surged over 300% since September 2025, with specific models now costing significantly more than they did just months prior [2][3] - The overall cost of building high-performance computing systems has escalated, with complete setups now reaching prices between 600,000 to 700,000 yuan [3][7] Market Trends - The influx of speculative capital has led to a chaotic market environment, with some channel suppliers engaging in practices that inflate prices artificially [4][5] - The emergence of counterfeit memory products has raised concerns about quality and reliability in the market [6] - Major cloud service providers are increasingly turning to domestic chip manufacturers as a response to supply shortages and rising costs of imported components [8][9] Future Outlook - The market is expected to evolve into a multi-layered structure, where domestic chips will gain traction in specific applications, particularly in sectors requiring data security and autonomy [8][9] - The ongoing development of AI computing architectures aims to unify various hardware standards, potentially reducing dependency on imported GPUs [12][13] - The next two years are critical for domestic chip manufacturers to address technical challenges and improve product usability in order to capitalize on the growing market demand [11][14]

Core Insights - Bitdeer Technologies Group reported significant operational updates for December 2025, highlighting a substantial increase in Bitcoin production and self-mining capacity [1][15]. Operational Update - Bitdeer sold approximately 1.4 EH/s of SEALMINER A2s to external parties in December 2025 [3]. - The total proprietary hash rate deployed increased to 58.0 EH/s, up from 47.3 EH/s in November 2025 [4]. - Self-mining production reached 636 Bitcoins, marking a 339% year-over-year increase and a 21% increase from November 2025 [7][11]. SEALMINER R&D - The company is advancing two chip designs, SEAL04-1 and SEAL04-2, with SEAL04-1 achieving 6-7 J/TH power efficiency under low-voltage mode, targeting mass production in Q1 2026 [8]. - Eight units of GB200 systems were deployed and tested in Malaysia, with a cloud service launch expected in January 2026 [6]. Infrastructure Summary - Total electrical capacity across various sites reached 1,658 MW, with several sites online and others in progress [10]. - The company is actively evaluating U.S. data center leasing opportunities to deploy GPUs and enhance AI cloud services [12]. Production and Operations Summary - Total hash rate under management increased to 71.0 EH/s in December 2025, up from 60.3 EH/s in November 2025 [11]. - The number of mining rigs under management rose to 293,000, with self-owned rigs increasing to 211,000 [13]. Management Commentary - The Chief Business Officer emphasized the acceleration of self-mining operations and the planned growth through 2026, driven by the deployment of additional proprietary rigs [15].

Summary of Key Points from the Conference Call Industry Overview - The report focuses on the **GPU cloud pricing and availability** within the **semiconductor industry**, particularly regarding AI demand and cloud service providers like AWS, Google Cloud, and Azure [2][4]. Core Insights and Arguments - **AI Demand Environment**: There are ongoing investor concerns about the durability of AI demand, prompting a revisit of GPU cloud pricing and availability [2]. - **Availability of Accelerators**: - The **B200** accelerator is now more widely available, with spot instances appearing at AWS and GCP for the first time in November 2025 [4]. - The **B300** has also been spotted at AWS, indicating faster market penetration compared to the B200 [4]. - **Pricing Trends**: - Pricing for older NVDA generation GPUs has seen a **1.8% month-over-month decline**, while prices for newer models like the **H100** and **H200** have increased by **3.3%** and **1.2%** respectively [4]. - The pricing for older accelerators remains stable, suggesting that cloud vendors still find economic value in these legacy chips [2][4]. - **AMD's Market Position**: There is limited traction for AMD's offerings, with no instances available across the covered clouds, although some manual checks indicate availability at Oracle [4]. Additional Important Information - **Legacy GPU Availability**: Older generation GPUs, including Ampere and Hopper, continue to be widely available, with no significant decline in their location counts [4]. - **Google and Amazon ASICs**: Google’s TPU and Amazon’s Trainium are available at stable prices, although Trainium2 pricing is noted to be volatile [4]. - **Competitive Landscape**: The report highlights the competitive dynamics between NVIDIA and AMD, with NVIDIA maintaining a strong position in the market despite AMD's potential for growth in cloud and AI sectors [54][55]. Data Highlights - **Spot and On-Demand Pricing**: The report provides detailed pricing comparisons for various accelerators, indicating significant premiums for on-demand pricing over spot pricing, with some instances showing premiums as high as **6.84x** for H100 [7][11][32]. - **Performance Metrics**: The theoretical performance and price/performance ratios for key accelerators are analyzed, showing NVIDIA's GPUs generally outperforming AMD's offerings in terms of price efficiency [37][44]. This summary encapsulates the critical insights from the conference call, focusing on the semiconductor industry's current state, particularly in the context of AI demand and cloud services.

Core Insights - Super Micro Computer (SMCI) is focusing on AI servers optimized for NVIDIA's latest GPU systems, which are in high demand for AI applications and large-scale deployments [1][4] Group 1: Company Performance - AI GPU platforms contributed over 75% of SMCI's revenues in the first quarter of fiscal 2026 [2][11] - SMCI launched new liquid-cooled 4U and 2-OU systems, which are now ready for volume shipments, enhancing its product offerings [2][11] - Despite a revenue decline in the first quarter of fiscal 2026, this was attributed to a shift in revenue timing rather than demand weakness, with expectations for strong revenue growth in the upcoming quarter [5][11] Group 2: Competitive Landscape - SMCI competes with Dell Technologies and Hewlett Packard Enterprise in the AI and data center market [6] - Dell Technologies has a strong market presence but has not matched SMCI's growth in AI-specific systems, leveraging its ability to bundle hardware with services [7] - Hewlett Packard Enterprise is aggressively expanding into AI and high-performance computing, positioning itself as a direct competitor to SMCI [8] Group 3: Valuation and Estimates - SMCI's shares have decreased by 26.7% over the past six months, contrasting with the 68.1% growth of the Zacks Computer – Storage Devices industry [9] - The company trades at a forward price-to-sales ratio of 0.47, significantly lower than the industry average of 1.75 [12] - Earnings estimates for fiscal 2026 and 2027 indicate a year-over-year increase of approximately 4.85% and 37.9%, respectively, although estimates have been revised downward in the past 60 days [15]

President Trump has lifted a US ban on exports to China of Nvidia's H200 chips. The H200 is more powerful than the H20, designed for export to China, but less powerful than the cutting-edge B200. Here's what to know about Nvidia's AI chips https://t.co/JuMPUUmJS8 ...

President Trump has lifted a US ban on exports to China of Nvidia's H200 chips. The H200 is more powerful than the H20, designed for export to China, but less powerful than the cutting-edge B200. Here's what to know about Nvidia's AI chips. https://t.co/Soz6ISRuOX ...

Core Viewpoint - The article discusses the potential release of NVIDIA's H200 in China, analyzing the implications from both the U.S. and Chinese perspectives, focusing on inventory clearance and market dynamics. Group 1: Reasons for U.S. Release - NVIDIA's CEO is advocating for the release of H200 to clear inventory, as the current market is dominated by the B series products, making it difficult to sell H200 in the U.S. [2] - The U.S. data centers are facing power supply issues, and the newer Blackwell architecture is more energy-efficient, leading to a gradual phase-out of older models like H100/H200. [2] - The ideal solution for NVIDIA is to legally sell H200 to China if it cannot be absorbed in the U.S. market. [2] Group 2: China's Attitude - There is a divided opinion in China regarding the release of H200; some believe that domestic AI chips are not yet competitive, while others fear that agreeing to the release could hinder local chip development and give the U.S. leverage. [3][11] - Economically, there seems to be no strong reason for China to ban the import of H200. [4] Group 3: Performance and Market Impact - The performance of H200, particularly in terms of computing power and memory bandwidth, currently exceeds that of domestic AI chips. [5] - Many existing codes are based on the Hopper architecture, making H200 easy to integrate for large companies. [8] - The domestic production capacity for high-end GPUs is not expected to significantly increase until 2027, indicating a continued reliance on foreign technology. [8] Group 4: Implications for Domestic Market - H200 has practical applications for Chinese customers, primarily in training scenarios, while domestic chips are more suited for inference tasks. [12] - The economic benefits of H200 may be limited due to rising memory prices, which could offset any price reductions. [13] - The overall impact of H200 on domestic GPU cards is expected to be minimal, as it does not directly compete with them. [13] Group 5: Market Reactions - The news about H200's potential release has caused market fluctuations, but the actual impact is likely to be limited, with key factors being policy direction, market demand, and funding conditions rather than just technical availability. [14]

Regulatory Landscape - Chinese regulators are preventing ByteDance from using Nvidia chips in new data centers [1] - China has issued guidelines requiring new, state-funded data center projects to use domestic AI chips exclusively [1] - Data centers less than 30% complete must remove foreign chips or cancel purchase plans; more complete projects will be assessed individually [1] - The new guidelines cover Nvidia's H20 chip, as well as more powerful processors like the B200 and H200 [1]