Core Insights - HIVE Digital Technologies has achieved a bitcoin mining capacity of 23 exahashes per second (EH/s) and aims to reach 25 EH/s by the end of the month, marking a 283% growth this year [1][2] Mining Capacity and Infrastructure - The increase in mining capacity is attributed to the full deployment of hardware at its 100 megawatt (MW) hydroelectric campus in Paraguay, powered by the Itaipú Dam [2] - HIVE reached a mining capacity of 10 EH/s in May [2] Transition to AI Computing - HIVE is converting parts of its bitcoin mining operations into liquid-cooled, AI-ready data centers, with a recent land acquisition in Grand Falls, New Brunswick, capable of supporting 25,000 GPUs [3] - The company is transitioning its facilities to high-performance computing (HPC) standards to cater to businesses requiring constant uptime, such as AI cloud platforms and financial services [4] Future Projections - Additional facilities in Toronto and Sweden are expected to increase the company's projected GPU total to 36,000 by the end of 2026, up from approximately 5,000 today [4] - The company is positioning itself to leverage its renewable-powered facilities for a cost and speed advantage as demand for compute-intensive AI workloads grows [4] Financial Position - HIVE currently holds around 2,201 BTC in its treasury, ranking as the 34th-largest digital asset treasury firm according to Bitcointreasuries data [5] Market Reaction - The company's shares fell by 2.9% in pre-market trading on Monday [6]

Core Viewpoint - Terra Wolf's stock reached an all-time high following the announcement of preliminary earnings and a significant joint venture with Fluid Stack, indicating a strategic shift towards AI computing alongside traditional Bitcoin mining [1][4]. Company Developments - Terra Wolf reported preliminary earnings numbers for the upcoming quarter, contributing to a surge in stock price [1]. - The company announced a 25-year AI compute deal with Fluid Stack, which could potentially increase its annual revenue to over $380 million, significantly more than its current trailing sales of approximately $144 million [4][5]. - The partnership with Google, which holds about 14% of Terra Wolf through warrants, adds credibility and support to the company's business model [3][6]. Industry Trends - There is a notable trend of companies transitioning from Bitcoin mining to AI computing, with Terra Wolf being part of this shift alongside others like Applied Digital and CleanSpark [2][7]. - The integration of AI computing capabilities into traditional Bitcoin mining facilities is becoming common, as companies seek to optimize idle resources for cloud computing and data center needs [4][8]. Financial Insights - Prior to the announcement, sales estimates for Terra Wolf were around $200 million for the current year and $434 million for the next year, which are expected to rise significantly due to the new deal [5]. - The market capitalization of Terra Wolf is approximately $5.6 billion, which is relatively small compared to larger tech companies, indicating potential for growth in the AI infrastructure space [8]. Market Dynamics - The stock experienced an 18% rally, reflecting strong market momentum, which may also be influenced by a high short interest of about 25% of the free float [10].

周二，TeraWulf(WULF.US)盘中大涨近24%，续创年内新高累计暴涨200%，现报16.97美元。消息面 上，TeraWulf表示，已与AI云平台运营商Fluidstack成立了一家合资企业，将在德克萨斯州阿比林建设并 运营一个168兆瓦的AI计算设施。该项目附带一份为期25年的托管协议，价值约95亿美元。TeraWulf 称，它将持有该合资企业51%的股份，计划于2026年下半年开始运营。 TeraWulf还表示，在这笔交易之后，其签约的高性能计算平台目前承载的关键IT负载已超过510兆瓦， 并正在追求每年新增250至500兆瓦签约容量的增长目标。 ...

市场观察——TACO一来，大盘拉升 10月20日，市场冲高回落，创业板指一度涨超3%。沪深两市成交1.74万亿，较上一个交易日缩量2005亿，成交额创8月8日以来新低。截至收盘，沪指涨 0.63%，深成指涨0.98%，创业板指涨1.98%。尤其是AI算力板块，今日迎来报复性反弹，"CPO"含量超50%的创业板人工智能ETF华夏(159381)、5G通信 ETF(515050)一度涨超6%，截至收盘分别涨3.89%和3.41%。 ...

2、见效更快：4G/5G"先覆盖后变现"的工程属性与ARPU增速受限导致回本周期达数 年；AI模型、算力投入后可即插即用，并不断迭代满足新需求、新场景，进行滚动式兑 现。 3、硬件迭代更快，优化价格、需求周期：AI算力"年更+软硬协同"的高频迭代，在12– 18个月内刷新单位算力成本并催生新需求，从而在价格下行之前就以"更高代际的价值锚"重 定价，打破了传统硬件的量价周期。当算力更便宜、更易得时，开发者会把更复杂的模型与 系统作为新基准，提升参数量、上下文与并行度。模型架构的迭代有机会减小单次推理、训 练所需算力，而AI产业发展过程中杰文斯悖论将演绎很多次，如Genie3等生成视频的世界模 型，或需要跨数量级提升的算力才可满足。投资建议：我们认为AI算力更高的天花板与更 好的竞争格局支撑起AI算力相对4G/5G的更高估值框架与更强Beta，相关标的——光互连： 中际旭创，新易盛，天孚通信，光库科技，长芯博创，仕佳光子，源杰科技，长光华芯，太 辰光；交换机：锐捷网络，盛科通信；铜互连：兆龙互连，中际旭创。 东吴证券近日发布AI算力跟踪深度：AI的产出是可乘数放大的"智能"，由"算力—模型 —应用—数据"飞轮驱动 ...

Group 1 - The article emphasizes the importance of identifying valuable research reports and understanding when to utilize them to avoid missing investment opportunities [1] - It highlights a core segment of AI computing power that accounts for 20% of total investment, which is currently experiencing a surge in North American orders, with two companies receiving significant attention from institutions [1] - Major players like Nvidia, Meta, Microsoft, and Google are actively engaging in the burgeoning SST market, which is projected to explode, with four key companies expected to benefit from this growth [1]

Group 1 - The overall manufacturing PMI is still affected by external unfavorable conditions, but policies promoting "anti-involution" and the service sector have shown positive effects, leading to improved corporate confidence [1] - As external trade shocks gradually dissipate, the policy tone of expanding domestic demand and its actual effectiveness will be further reflected in the data [1] - The market expects the central bank to introduce further easing measures, which may boost market risk appetite [1] Group 2 - Recent market performance shows strong structural characteristics, with hotspots concentrated in growth sectors [2] - The strong demand for overseas AI computing power continues to be validated, and China's clear energy transition goals position it as a manufacturing powerhouse participating in technological innovation [2] - Recommendations for October include focusing on AI computing and robotics-related industries, innovative pharmaceuticals, consumer electronics, batteries, and non-ferrous metals [2]

Core Insights - Hitachi Vantara and Supermicro are forming a strategic partnership to integrate Supermicro's AI compute capabilities with Hitachi's VSP One unified storage platform, enhancing AI infrastructure and data management for enterprises [1][4][5] Partnership Details - The collaboration will allow Supermicro to sell VSP One to its customers while Hitachi Vantara will sell Supermicro's servers and hardware, expanding market reach [1][6] - The agreement is in the finalization stage, aiming to streamline infrastructure for AI-driven applications and data-intensive workloads [1][2] Market Context - A survey indicates that 85% of organizations are using data lakehouses for AI model development, with 67% expecting to run most analytics on lakehouses in the next three years, up from 55% [2] - Enterprises face challenges such as fragmented systems and rising costs, which hinder AI scalability [2] VSP One Features - VSP One integrates block, file, object, and software-defined storage into a single architecture, providing high throughput and low latency for mission-critical applications [3][4] - It includes features like native support for Amazon S3 Tables and advanced data intelligence services, facilitating easier data management and analytics [3][4] Strategic Importance - The partnership is seen as a significant step in aligning compute and data capabilities, which is essential for scalable AI workloads [4] - Hitachi iQ, the AI and data orchestration portfolio, will leverage VSP One to optimize data workflows for advanced AI applications [4][5] Technology Availability - The technologies from this partnership will be made available through global channels, enhancing the offerings for both companies' partners [6]

Core Viewpoint - Jim Chanos has criticized CoreWeave Inc.'s economic model, arguing that the company's significant capital investments do not yield meaningful returns based on the CEO's statements [1][2]. Financial Analysis - Chanos highlighted that CoreWeave's second-quarter financials indicated an average of $19.1 billion in capital employed, leading to an annual "economic depreciation" of $2.729 billion when amortized over the seven-year useful life of the equipment [2][3]. - This depreciation figure surpasses the company's annualized second-quarter EBITDA of $2.624 billion, resulting in a 0% return on invested capital (ROIC) according to Chanos [3]. Company Response - CoreWeave CEO Michael Intrator defended the company's position during an interview, emphasizing the importance of their sophisticated software stack and dismissing the idea that the company is merely a real estate play [4][5]. - Intrator aimed to bolster confidence following the all-stock acquisition of Core Scientific Inc., asserting that the software is crucial to the AI ecosystem [4]. Market Performance - CoreWeave's shares increased by 8.65% to $139.98 on Wednesday, with a further rise of 0.71% in after-hours trading, marking a total increase of 249.95% since its listing in March 2025 [6].

Core Viewpoint - CIMG Inc. has established a subsidiary, Braincoin Limited, in Hong Kong to enhance its strategic expansion into the global computing power industry and advance its AI initiatives [1][2][4] Group 1: Company Expansion - The new subsidiary will act as a hub connecting mainland China with international markets, facilitating collaboration with global partners [2] - This move emphasizes the company's commitment to building a global computing power ecosystem and delivering cutting-edge AI solutions [4] Group 2: Business Models - Braincoin will operate under several business models starting in October, including the development and operation of AI computing centers for enterprises with high computing power demands [3][8] - The company will provide customized multi-modal enterprise AI assistants by integrating large AI models with clients' internal and industry-specific datasets [8] - Managed services will be offered to existing AI computing center clients, focusing on maintenance and operational support to enhance efficiency and reduce costs [8]