文：刘涛 经历三年低迷期后，香港股票发行市场重现繁荣，短短一年间强势反弹，并预将重夺全球最大的股票首发地。 8月20日，香港交易所公布2025年中期业绩，上半年实现总收入140.76亿元（港币，下同），同比增长32.53%，净利润为85.19亿元，同比增长39%，均创 历史半年度新高。 这些龙头企业选择赴港IPO，一方面是国际化战略需要，另一方面港交所对新经济企业、科创企业的高包容度也极具吸引力。"A+H"模式的升温不仅拓宽 企业融资管道，更透过香港这个国际平台助力中国企业全球化布局。比如，宁德时代将90%的募资投入匈牙利项目，讯飞医疗借助香港资本市场加速AI技 术迭代。在龙头企业的示范作用下，越来越多的未盈利的生物医药公司、人工智能企业等科技企业加入赴港上市队列。 中国人工智能初创公司DeepSeek的崛起亦是香港IPO市场的另一个催化剂。今年1月，DeepSeek发布了低成本但功能强大的模型，由此推动了中国科技股 的上涨，投资者重新评估中国的创新能力，并引发了对中国股票的价值重估。 在中国科技股的助攻下，香港成为今年亚洲表现最好的股票市场之一，如今，在港上市的内地企业已占整个市场市值的81%。 利润和收入 ...

Group 1 - Antig's plan to build three natural gas power plants has been approved to supply power to Meta's largest data center in Louisiana, which covers 4 million square feet and is expected to consume up to 5 gigawatts of electricity annually when fully operational [2] - Global display equipment spending is projected to reach $75.8 billion from 2020 to 2027, driven primarily by OLED technology, with OLED-related equipment spending expected to grow by 31% year-on-year in 2025, while LCD spending is anticipated to decline by 45% [2] - Nikon announced the closure of its Yokohama factory by September 30, with production lines moving to other facilities in Kanagawa and Tokyo, primarily affecting microscope products and flat panel display manufacturing equipment [2] Group 2 - SpaceX's Starlink service has gained traction in the in-flight Wi-Fi market, signing contracts with major airlines such as Air France and Qatar Airways, with ongoing negotiations with Emirates and other regional airlines [3] - The Zhiyuan Lingxi X2 robot is set to complete mass production by the end of August, with unlimited shipments starting in the fourth quarter [3]

中文数据在国内AI大模型的训练性能提升方面发挥着重要作用。国家数据局近日发布的数据显示，目 前国内多数AI模型训练使用的中文数据占比已经超过60%，有的模型达到80%。中文高质量数据的开发 和供给能力持续增强，推动我国人工智能模型性能快速提升。 国家数据局局长刘烈宏表示，我国人工智能的快速发展，与我国高度重视数据工作是密不可分的。作为 人工智能发展的核心要素之一，数据在推动"人工智能+"过程中发挥着关键作用，高质量数据集的建设 至关重要。 下一步，国家数据局将通过体系化布局持续推进高质量数据集建设，加快打造具身智能、低空经济、生 物制造等重点领域数据高地，推动全社会强化数据要素价值认同，加快推进数据要素价值共创，培 育"为优质数据买单"的市场共识。 "在人工智能时代，Token，也就是大家通常所说的词元，是处理文本的最小数据单元，如同互联网时代 大家所说的'流量'。"刘烈宏介绍，2024年初，我国日均Token的消耗量为1000亿，到今年6月底，日均 Token消耗量已经突破30万亿，一年半时间增长了300多倍，反映了我国人工智能应用规模的快速增长。 （文章来源：新华社） 据介绍，截至今年6月底，我国已经建设 ...

Group 1 - The article discusses the concept of a "chip iron curtain," indicating a new form of isolation created by sanctions and bureaucratic measures rather than physical barriers [1][2] - Western capitals are implementing regulations against China, reflecting a fear of losing control over globalization [1][2] - The U.S. decision to pause the ban on NVIDIA's H20 chip exports to China signifies an acknowledgment of its own vulnerabilities [1][2] Group 2 - The speed of regulatory barriers being erected exceeds the understanding of their consequences, with U.S. lawmakers proposing sanctions while companies negotiate privately with China [2] - The impact of U.S. bans is harming Western companies, with NVIDIA facing losses in the billions and ASML seeking new clients [2] - China's response involves a proactive strategy focused on self-sufficiency, aiming for over 70% autonomy in its semiconductor and industrial software sectors by 2028 [2] Group 3 - Russia serves as a testing ground for Western pressure on China, with China learning from this experience to accelerate its internal integration [3] - The article suggests that the "chip iron curtain" symbolizes the decline of an empire fearful of its own sunset, as Asia builds its supply chains and new rules [3] - As the U.S. debates regulations, China is actively expanding its manufacturing capabilities, indicating a shift in the future economic landscape towards the East [3]

Group 1 - China is intensifying efforts to build a domestic AI ecosystem that does not rely on Western technology, preparing for long-term technological competition with the U.S. [1] - The 2025 World Artificial Intelligence Conference showcased many products that do not contain U.S. technology, highlighting China's advancements in AI [1][2] - Chinese companies, such as Huawei, are improving chip performance by aggregating more chips together, aiming to close the performance gap with U.S. products [1] Group 2 - The release of the "Global Governance Action Plan for Artificial Intelligence" at the conference emphasizes China's ambition to set global governance standards in AI and potentially weaken U.S. influence [2] - A nationwide initiative led by the Chinese government is underway to enhance AI capabilities, involving significant investments from state-owned enterprises, private companies, and local governments [2] - Winning the global AI race is deemed crucial for economic and military impact, with the U.S. currently holding a first-mover advantage [2] Group 3 - China's best open-source AI models have reportedly surpassed the performance of the best U.S. models since November of the previous year [3] - China plans to invest $564 billion in building its power grid over the next five years, a 40% increase compared to the previous five years, to support AI development [3] - The number of universities in China offering AI undergraduate programs has surged from 35 in 2019 to 626, indicating a strong pipeline of local talent for the AI industry [3]

Group 1 - Palantir's stock price experienced a short-term decline, with an overall drop of 3.3% during the day [1] Group 2 - The U.S. Department of Defense is testing AI models provided by Microsoft and OpenAI [2]

Group 1 - China proposed the establishment of a new global AI cooperation organization to enhance international collaboration in the AI sector, with potential headquarters in Shanghai [1][2] - The initiative reflects China's commitment to multilateralism and aims to address the needs of the Global South, promoting equitable and beneficial AI development [1][2] - The U.S. is responding to China's rise in AI by expanding its own AI-related exports to allies, indicating a competitive landscape between the two nations [1][2] Group 2 - The "Global Governance Action Plan for AI" was released at the conference, inviting international cooperation based on principles of safety, fairness, and open collaboration [2] - Despite U.S. export restrictions on advanced AI technologies to China, the latter continues to make significant advancements in the field, attracting attention from U.S. officials [2] - China's generative AI technology is rapidly catching up to the U.S., with a 40% increase in registered models over the past six months, indicating a narrowing performance gap [3]

Group 1 - The California state legislature has initiated a new AI regulatory framework following Governor Gavin Newsom's veto of a previous bill that held tech companies accountable for technological damages [1] - The SB 53 amendment, introduced by Senator Scott Wiener, aims to establish the first national AI transparency regulatory framework, requiring companies developing AI models exceeding a specific computational threshold to publicly disclose safety assessment protocols [1] - Companies such as OpenAI, Alphabet (GOOGL.US), and Anthropic will be required to submit special reports to the state attorney general in the event of significant safety incidents, such as sensitive technology leaks [1] Group 2 - The SB 53 amendment retains whistleblower protection provisions and a public cloud service platform proposal to support startups and academic institutions with low-cost computational resources, while removing controversial liability clauses [2] - The amendment aligns with recommendations from the AI Frontier Model Working Group, which includes prominent experts like Fei-Fei Li, providing professional support for the current legislative proposal [2] - The SB 53 bill is expected to enter the committee review stage later this month, with ongoing consultations with tech companies and academic institutions to refine the proposal [2]

Core Viewpoint - The proposal by the Trump administration to prohibit states from regulating artificial intelligence (AI) for the next decade is seen as a hindrance to the development of this technology, according to Microsoft's Chief Scientist Eric Horvitz [1][2]. Group 1: Regulatory Concerns - Horvitz warns that banning state-level regulations will lead to setbacks in scientific progress and the practical application of AI [1][2]. - He expresses concerns about the misuse of AI for spreading misinformation and malicious activities, particularly in sensitive areas like biohazards [1]. Group 2: Industry Position - Despite reports of Microsoft collaborating with other tech giants like Google, Meta, and Amazon to lobby for the proposed ban, Horvitz maintains a stance in favor of regulation [1]. - Microsoft is actively lobbying the U.S. Senate to pass a ten-year ban on state-level AI regulations, which is included in Trump's "Big Beautiful Bill" aimed for approval by July 4 [1].

Core Viewpoint - The Chinese government emphasizes the importance of compliance with laws and regulations by Chinese enterprises while opposing coercive actions that restrict cooperation with China [1] Group 1: Government Response - The Chinese government has consistently required Chinese enterprises to operate in accordance with laws and regulations [1] - The Chinese government opposes any coercive measures that limit cooperation with China from other countries [1] - The Chinese government expresses willingness to work with regional countries, including Malaysia, to maintain a free and open international trade order [1]