RT s4mmy (@S4mmyEth)ICYMI: NVIDIA + OpenAI reveal $100bn strategic partnership to deploy 10GW of GPU infra.This is ∼2x Portugal’s entire annual electricity consumption.This builds on the existing $500bn Stargate agreement with SoftBank.That’s not all today:1) Oracle is in talks with Meta over $20b AI deal, and;2) xAI raises $10b at $200b valuation.GPUs and power are going to be the biggest bottlenecks for AI development.It’s no wonder that Big Tech is throwing eye watering sums of money at the problem.This ...

Market Overview - A-shares experienced a volatile upward trend on September 22, with all three major indices rising, led by technology stocks, particularly the Sci-Tech 50 Index which increased by over 3% [1][2] - The total market capitalization of A-shares reached 114.20 trillion yuan, with a rolling P/E ratio of 22.16 times for the entire A-share market and 13.97 times for the CSI 300 Index [5] Sector Performance - The electronic sector led the gains, with significant increases in stocks such as Hongfu Huan, Changying Precision, and Zhongke Lanyun, all hitting the 20% limit up [2][3] - The computer sector also saw strong performance, with stocks like Chuling Information and Zhongke Shuguang reaching the limit up [2][3] - Among the 31 sectors, electronics, non-bank financials, and power equipment saw the highest net inflows of financing, while sectors like non-ferrous metals and defense experienced net outflows [3][4] Trading Volume and Financing - The trading volume on September 22 was 2.14 trillion yuan, a decrease of 207 billion yuan from the previous trading day, but it marked the 29th consecutive trading day with volumes exceeding 2 trillion yuan [2][3] - The financing balance in the A-share market increased by over 46 billion yuan last week, with a total financing balance reported at 23,816.10 billion yuan as of September 19 [3][4] Capital Flow - On September 22, the net outflow of main funds from the Shanghai and Shenzhen markets was approximately 149.51 billion yuan, with 2,123 stocks experiencing net inflows and 3,021 stocks seeing net outflows [4] - The electronic, banking, and non-bank financial sectors had the highest net inflows, while sectors like power equipment and media saw significant outflows [4] Market Outlook - Analysts suggest that the current upward trend in the A-share market is likely to continue, driven by improving overseas liquidity and a strengthening domestic economic outlook [5][6] - There is a focus on sectors such as solid-state batteries, AI computing, humanoid robots, and commercial aerospace for potential investment opportunities [5][6]

Core Insights - The A-share market is experiencing a consolidation phase with a slight increase in major indices, reflecting a strong wait-and-see sentiment among investors. The technology sector continues to lead gains, while the consumer sector is undergoing a collective pullback [2][6][8]. Market Performance - On September 22, 2025, the Shanghai Composite Index rose by 0.22%, the Shenzhen Component increased by 0.67%, the ChiNext Index gained 0.55%, the SSE 50 rose by 0.43%, the CSI 300 increased by 0.46%, the STAR 50 surged by 3.38%, and the CSI 1000 rose by 0.69%. The total market turnover was 2.14 trillion yuan, with 2,175 stocks rising [2][8]. Sector Performance - In the A-share market on September 22, 2025, the electronic sector led with a gain of 3.55%, followed by the computer sector at 1.76%, and comprehensive finance at 1.10%. Conversely, the social services sector fell by 1.86%, food and beverage by 1.20%, and household decoration and leisure by 1.13% [8]. Conceptual Trends - Key concepts showing strong performance include Moore Threads (+6.10%), smart speakers (+6.07%), GPUs (+5.69%), and memory chips (+4.69%). In contrast, sectors such as travel, shipping, duty-free shops, and phosphorus chemicals experienced declines [8]. Market Drivers - The A-share market's slight rise is attributed to a consolidation phase, with technology stocks leading the way. The IPO news of Moore Threads significantly boosted related stocks, while the strong sales of the iPhone 17 positively impacted the consumer electronics supply chain. Additionally, the rise in gold and silver prices has supported the precious metals sector [8]. Future Outlook - The report maintains a bullish outlook on the Chinese stock market, emphasizing the importance of "liquidity" in 2025. It suggests that the market is expected to perform well, drawing parallels to previous bull markets in 1999, 2014, and 2019 [8]. Investment Strategy - Short-term focus should be on sectors with recent revenue growth and improving gross margins, such as fiberglass, cement, paper, fine chemicals, oil services, and medical services. For technology growth, attention should be on "double innovation" and the Hang Seng Technology Index, particularly in lithium batteries, military industry, and Hong Kong internet sectors [8]. Long-term Perspective - In the technology growth sector, continued optimism is expressed for AI computing, Hong Kong innovative pharmaceuticals, and military industries. There is also a focus on sectors benefiting from the "anti-involution" trend, including metals, transportation, chemicals, lithium batteries, photovoltaics, and pig farming [8].

港股市场整体承压，科技股显现韧性。恒生指数跌0.76%报26344.14点；恒生科技指数跌0.58%报 6257.91点；恒生中国企业指数跌1.07%报9370.73点。大市成交2905.37亿港元，南向资金净买入127.36 亿港元，半导体与消费电子板块获内资青睐。 9月22日，A股市场呈现结构性分化格局，三大指数小幅收涨，科技板块全线爆发推动科创50指数大涨 3.38%，消费电子与半导体产业链领涨市场，而旅游影视等消费板块集体回调；港股市场整体承压，三 大指数集体收跌，科技股相对抗跌，苹果概念与半导体板块逆势走强。 受国庆节前效应影响，两市成交额均较前一交易日显著萎缩，显示资金观望情绪浓厚，市场热点集中於 科技主线，结构性特征突出。 展望后市，A股科技主线有望延续强势，需关注量能变化与板块轮动节奏；港股或延续震荡整固，重点 跟踪外资流向与政策预期。操作上建议聚焦半导体、消费电子等景气方向，规避高位消费股回调风险， 把握科技成长与政策催化的结构性机会。 主要指数表现 A股市场缩量上涨，科技成长风格主导市场。上证指数全天震荡收涨0.22%报3828.58点；深证成指涨 0.67%报13157.97点；创业板指 ...

Market Overview - The A-share market showed a mixed performance today, with the Shanghai Composite Index rising by 0.22% to close at 3828.58 points, while the Shenzhen Component increased by 0.67% to 13157.97 points. The Sci-Tech 50 index saw a significant rise of 3.38%, closing at 1408.64 points, and the ChiNext index rose by 0.55% to 3107.89 points. Overall, there were 3150 stocks that declined, while 2175 stocks advanced, with 102 stocks remaining flat. The total trading volume in the Shanghai and Shenzhen markets was 21215 billion yuan, a decrease of 2023.47 million yuan compared to the previous trading day [1][2]. Industry Performance - Among the primary industries, electronics, computers, non-ferrous metals, machinery equipment, non-bank financials, and automobiles led the gains, while sectors such as social services, beauty care, retail, food and beverage, construction decoration, and oil and petrochemicals experienced declines [1][2].

Core Viewpoint - The stock of Huadong Heavy Machinery has shown a significant increase in price and trading activity, indicating potential investor interest and market dynamics [1][2]. Group 1: Stock Performance - On September 22, Huadong Heavy Machinery's stock rose by 2.06%, reaching 8.93 CNY per share, with a trading volume of 184 million CNY and a turnover rate of 2.08%, resulting in a total market capitalization of 8.999 billion CNY [1]. - Year-to-date, the stock price has increased by 32.69%, with a 9.71% rise over the last five trading days, a 0.56% increase over the last 20 days, and an 18.59% increase over the last 60 days [1]. - The company has appeared on the "Dragon and Tiger List" twice this year, with the most recent occurrence on August 26 [1]. Group 2: Financial Performance - For the first half of 2025, Huadong Heavy Machinery reported a revenue of 364 million CNY, a year-on-year decrease of 34.69%, and a net profit attributable to shareholders of 26.1557 million CNY, down 3.72% year-on-year [2]. - The company has distributed a total of 185 million CNY in dividends since its A-share listing, with no dividends distributed in the last three years [2]. Group 3: Business Overview - Huadong Heavy Machinery, established on January 9, 2004, and listed on June 12, 2012, focuses on high-end equipment manufacturing, primarily in "container handling equipment" and "intelligent CNC machine tools" [1]. - The company is also expanding into the solar energy sector, leveraging opportunities in the photovoltaic industry to enhance its business structure [1]. - The revenue composition of the company is as follows: container handling equipment accounts for 99.40%, battery cells and others for 0.56%, technical services and others for 0.02%, and chip-related products for 0.02% [1].

Group 1 - Dazhong Public (600635) shares increased by over 14%, reaching a price of 3.25 HKD with a trading volume of 123 million HKD [1] - The company announced its stake in Shenzhen Capital Group, which holds shares in Yushu Technology [1] - Yushu Technology is expected to submit its listing application to the stock exchange between October and December 2025 [1] Group 2 - Market attention is on Moer Thread, which is scheduled for a meeting on the Sci-Tech Innovation Board this week [1] - Moer Thread, established in 2020, focuses on developing a full-featured GPU aimed at providing computing acceleration platforms for AI, digital twins, and high-performance computing [1] - The actual controller, Zhang Jianzhong, holds 44.07% of the company's shares, and Moer Thread has received investments from multiple institutions, including Shenzhen Capital Group [1]

消息面上，大众公用在互动平台表示，公司参股深创投，深创投及其基金持有宇树科技股份。据悉，宇 树科技预计将在2025年10月至12月期间向证券交易所提交上市申请文件。另外，市场关注摩尔线程本周 科创板上会，资料显示，摩尔线程成立于2020年，以自主研发的全功能GPU为核心，致力于为AI、数 字孪生、科学计算等高性能计算领域提供计算加速平台。实际控制人张建中控制公司44.07%股份。摩 尔线程成立后曾获得包括深创投在内的多家机构投资。 智通财经APP获悉，大众公用(01635)涨超14%，截至发稿，涨14.04%，报3.25港元，成交额1.23亿港 元。 ...

RT Avi Chawla (@_avichawla)PyTorch dataloader has 2 terrible default settings.Fixing them gave me ~5x speedup.When you train a PyTorch model on a GPU:- .to(device) transfers the data to the GPU.- Everything after this executes on the GPU.This means when the GPU is working, the CPU is idle, and when the CPU is working, the GPU is idle.Memory pinning optimizes this as follows:- When the model is trained on the 1st mini-batch, the CPU can transfer the 2nd mini-batch to the GPU.- This ensures that the GPU does ...

PyTorch dataloader has 2 terrible default settings.Fixing them gave me ~5x speedup.When you train a PyTorch model on a GPU:- .to(device) transfers the data to the GPU.- Everything after this executes on the GPU.This means when the GPU is working, the CPU is idle, and when the CPU is working, the GPU is idle.Memory pinning optimizes this as follows:- When the model is trained on the 1st mini-batch, the CPU can transfer the 2nd mini-batch to the GPU.- This ensures that the GPU does not have to wait for the ne ...