嗨~~大家晚上好呀！今天是2026年3月18日，星期三，欢迎来到《晚安咖啡》，我是你们的主播安安～✨ 今晚的咖啡时间，我将为大家带来新鲜热乎的商业科技动态，一起聊聊那些改变我们生活的创新故事。准备好了吗？让我们放松心情，开启今晚的温暖陪伴吧！
@@@
据美国消费者新闻与商业频道CNBC报道，中国科技巨头腾讯控股于2026年3月18日公布了其2025年度财务报告。数据显示，该公司在2025年全年营收实现了百分之十四的同比增长，这一表现超出了市场分析师的普遍预期。推动增长的主要动力来自于其核心业务板块，包括增值服务、网络广告以及金融科技与企业服务。与此同时，腾讯正在持续加大在人工智能领域的战略性投入，将人工智能技术深度整合至其社交媒体、云计算、广告系统及企业服务等多个核心产品线中，旨在提升用户体验并创造新的增长点。公司管理层在财报中强调，对人工智能等前沿技术的长期投资是构建未来竞争力的关键。尽管面临激烈的市场竞争和宏观经济环境的挑战，腾讯通过业务结构的优化和技术创新，依然展现了稳健的财务韧性和增长潜力。此次财报的亮眼表现，也反映了市场对其战略方向和技术布局的积极反馈。
@@@
据美国消费者新闻与商业频道CNBC报道，科技巨头Meta公司旗下的Manus部门近日正式推出了一款桌面应用程序，旨在将其人工智能助手直接部署到用户的个人电脑设备上。这一举措正值其竞争对手OpenClaw公司掀起的人工智能代理热潮之际，标志着个人设备端人工智能助手市场的竞争进一步加剧。该桌面应用允许用户无需通过网页浏览器即可直接调用Manus的人工智能代理，旨在提供更快速、更集成化的本地体验。公司表示，此举是为了回应用户对更便捷、更私密人工智能工具的需求，让AI能力更深入地融入日常数字工作流。目前，该应用已提供适用于主流操作系统的版本供用户下载。行业观察指出，随着OpenClaw等公司的产品引发广泛关注，各大科技企业正加速将生成式人工智能从云端推向终端设备，以争夺这一新兴市场的份额。Meta此次发布桌面客户端，是其将人工智能定位为核心战略、并试图在硬件生态之外拓宽软件服务入口的又一关键步骤。
@@@
据行业媒体报道，三星电子已与芯片设计公司AMD达成合作，将为后者计划于2026年推出的下一代人工智能加速器供应关键的HBM4高带宽内存。这项合作标志着两家科技巨头在快速增长的AI硬件领域进一步深化了伙伴关系。与此同时，三星也在积极探讨为其晶圆代工业务引入新的外部投资者，以增强在该领域的竞争力与资源。这一举措被视为三星应对当前激烈市场竞争、强化其半导体制造版图的重要战略步骤。HBM4作为新一代内存技术，对于处理人工智能和大数据工作负载至关重要，能够显著提升计算效率。三星作为全球领先的内存芯片制造商，此次与AMD的绑定，不仅巩固了其在高端内存市场的供应地位，也可能影响未来AI加速器市场的竞争格局。关于晶圆代工业务的引资讨论目前仍在进行中，具体细节和潜在合作伙伴尚未公开。
@@@
据新浪科技报道，腾讯控股于2026年3月18日发布了2025年全年财报。数据显示，其面向企业客户的业务，即To B业务，全年营收达到2294亿元人民币，创下历史新高。其中，腾讯云业务在2025年实现了全年规模的盈利，这标志着该业务在持续投入后进入了新的发展阶段。财报进一步指出，金融科技及企业服务板块作为公司的重要增长引擎，其收入占比持续提升，反映了公司在产业互联网领域的深耕取得了实质性进展。与此同时，腾讯在研发方面继续保持高投入，以巩固其在云计算、人工智能等核心技术领域的竞争力。此次财报的发布，向市场展示了腾讯在消费互联网之外，其企业服务业务的稳健财务表现和持续增长潜力。
@@@
据TechWeb报道，人工智能公司OpenAI于今年3月18日正式发布了新一代小型化模型GPT‑5.4 mini与GPT‑5.4 nano。这是该公司迄今为止推出的能力最强的小型模型系列，旨在以更低的成本和更高的效率为开发者提供强大的语言处理功能。新模型在多项核心基准测试中表现优异，特别是在代码生成、数学推理和语言理解任务上，其性能已接近甚至超越部分规模更大的前代模型。GPT‑5.4 mini在保持较高性能的同时显著优化了响应速度与计算资源消耗，而更为轻量级的nano版本则专门为资源极度受限的边缘设备与移动应用场景设计。OpenAI表示，此次发布是为了响应市场对高效、可负担人工智能工具的广泛需求，帮助开发者更便捷地将先进AI能力集成到各类应用程序中。两款新模型现已通过API向开发者开放，并采用了按使用量计费的灵活模式。该公司强调，新模型同样内置了完善的安全与对齐机制，以降低潜在风险。这一产品线的更新被视为OpenAI扩大其技术覆盖范围、服务更广泛开发者和企业用户的关键举措。
@@@
据科技媒体cnBeta报道，英伟达首席执行官黄仁勋近日亲口证实，公司已重启面向中国市场的H200人工智能芯片的生产，并已开始向多家中国客户供货。这一消息是在外界普遍关注美国出口管制政策对高端芯片在华销售影响的背景下传出的。黄仁勋表示，英伟达正在严格遵守相关贸易法规的前提下，为中国市场提供符合规定的产品。H200是英伟达推出的高性能GPU，广泛应用于人工智能训练与推理领域。此次恢复生产与供货，标志着英伟达在调整其产品策略以适配特定市场要求方面取得了进展。公司方面没有透露具体有哪些中国客户采购了这批芯片，也没有公布涉及的具体数量与金额。行业观察人士指出，这一动向对于依赖高端算力的中国人工智能产业而言是一个重要进展。目前，英伟达仍在持续与各方沟通，以确保其商业活动完全符合国际规则。
@@@
据TechWeb报道，阿里巴巴集团旗下的云计算部门阿里云近日正式宣布将对部分产品与服务进行价格调整。此次调整涉及的核心产品包括弹性计算实例、对象存储服务以及数据库产品等多个领域，部分产品价格将出现不同程度上浮。阿里云方面表示，此次价格调整是基于全球市场原材料与能源成本持续上升、核心技术研发投入加大以及为保障服务质量持续升级基础设施等多方面因素的综合考量。公司强调，将继续致力于为客户提供稳定高效的云服务，并计划在未来进一步增加在技术创新与全球数据中心网络扩展方面的投资。此次价格调整预计将从下个季度开始逐步生效，阿里云已开始通知可能受到影响的客户群体，并建议用户根据自身业务需求评估潜在影响。
@@@
据Readhub报道，宇树科技创始人王兴兴近日表示，预计在今年内，该公司研发的人形机器人将能够在奔跑速度上超越世界短跑冠军博尔特。博尔特保持的男子百米世界纪录为九秒五八，这要求机器人具备极高的动态平衡能力与爆发力。宇树科技作为专注于高性能仿生机器人的企业，其最新一代产品在关节电机、运动控制算法和整机稳定性方面持续取得突破。王兴兴指出，实现超越人类极限的奔跑速度是机器人运动能力发展的重要里程碑，这不仅展示了技术进步，也为未来机器人在复杂、非结构化环境中的实际应用奠定了关键基础。公司团队正致力于优化机器人的步态与能耗效率，以达成这一极具挑战性的目标。这一进展也反映出全球人形机器人行业在运动性能竞赛上的加速态势。
@@@
据美国科技资讯网站cnBeta报道，五角大楼正计划推出一项新举措，允许获得批准的商业人工智能公司在高度机密的军事数据上训练其模型。这项计划旨在利用私营部门在人工智能领域的创新能力，加速国防领域尖端人工智能工具的开发。根据该计划，符合安全标准的人工智能企业将能够在确保数据绝对安全的前提下，接触到此前仅限于政府内部使用的敏感信息。美国国防部官员表示，此举是为了弥合商业人工智能快速发展与国防应用需求之间的差距，确保美军能够获取并整合最先进的人工智能能力。该计划预计将建立严格的安全协议和审查流程，只有通过严格背景调查并满足一系列网络安全要求的公司才能参与。分析认为，这一举措可能为人工智能技术在军事指挥、情报分析、后勤规划等关键领域的应用开辟新的路径。目前，相关计划的详细实施框架和时间表仍在进一步制定中。
@@@
据科技媒体cnBeta报道，美国联邦航空管理局已正式批准电动垂直起降飞行器，也就是eVTOL空中出租车，于今年夏季在全美范围内展开商业运营。这一决定标志着城市空中交通迈入了全新的商业化阶段。获得运营许可的机型需完全符合该机构严苛的安全与适航标准，这些标准与传统民航飞机处于同一水平。首批投入服务的空中出租车预计将部署在特定城市区域，主要用于机场与市中心之间的接驳，以及城市内的短途出行。这种飞行器能够垂直起降，无需传统跑道，其设计目标是在降低噪音的同时实现零排放运营。行业分析指出，这一批准为相关制造商和运营商铺平了道路，有望缓解地面交通拥堵问题。不过，具体的票价、运营路线以及更广泛的城市空中交通管理细则，仍有待相关企业与地方管理部门进一步公布。
@@@
好啦，今天的节目到这里就结束啦。希望今天的分享能给你带来一点小小的灵感或温暖。感谢你的收听和陪伴！如果您喜欢我们的节目，也非常欢迎订阅和转发给你的小伙伴，这对我们非常有帮助。那我们下期再见吧！
Hi~~ Good evening, everyone! Today is Wednesday, March 18, 2026. Welcome to "Good Night Coffee," I'm your host, An An~ ✨ For tonight's coffee time, I'll be bringing you the freshest business and tech updates. Let's chat about those innovative stories that are changing our lives. Ready? Let's relax and begin tonight's warm companionship!
@@@
According to CNBC, Chinese tech giant Tencent Holdings announced its 2025 annual financial report on March 18, 2026. The data shows the company achieved a 14% year-over-year revenue growth for the full year of 2025, surpassing the general expectations of market analysts. The primary drivers of this growth were its core business segments, including Value-Added Services, Online Advertising, and FinTech and Business Services. Meanwhile, Tencent continues to increase its strategic investments in the field of artificial intelligence, deeply integrating AI technology into multiple core product lines such as social media, cloud computing, advertising systems, and enterprise services, aiming to enhance user experience and create new growth opportunities. The company's management emphasized in the earnings report that long-term investment in frontier technologies like AI is key to building future competitiveness. Despite facing intense market competition and macroeconomic challenges, Tencent has demonstrated robust financial resilience and growth potential through business structure optimization and technological innovation. The strong performance in this earnings report also reflects the market's positive feedback on its strategic direction and technological layout.
@@@
According to CNBC, Manus, a division of tech giant Meta, recently officially launched a desktop application designed to deploy its AI assistant directly onto users' personal computer devices. This move comes amid the AI agent craze stirred up by its competitor OpenClaw, marking a further intensification of competition in the personal device AI assistant market. The desktop app allows users to directly invoke Manus's AI agent without needing a web browser, aiming to provide a faster, more integrated local experience. The company stated this is in response to user demand for more convenient and private AI tools, integrating AI capabilities more deeply into daily digital workflows. Currently, versions of the app for mainstream operating systems are available for download. Industry observers note that as products from companies like OpenClaw gain widespread attention, major tech firms are accelerating the push of generative AI from the cloud to end-user devices to compete for share in this emerging market. Meta's release of this desktop client is another key step in its strategy to position AI as a core focus and broaden software service entry points beyond its hardware ecosystem.
@@@
According to industry media reports, Samsung Electronics has partnered with chip design company AMD to supply critical HBM4 high-bandwidth memory for the latter's next-generation AI accelerator planned for launch in 2026. This collaboration marks a further deepening of the partnership between the two tech giants in the rapidly growing AI hardware sector. Concurrently, Samsung is actively exploring bringing in new external investors for its foundry business to enhance competitiveness and resources in that field. This move is seen as an important strategic step for Samsung to address the current intense market competition and strengthen its semiconductor manufacturing footprint. HBM4, as a new generation of memory technology, is crucial for handling AI and big data workloads, significantly boosting computational efficiency. As a global leading memory chip manufacturer, Samsung's tie-up with AMD not only consolidates its supply position in the high-end memory market but could also influence the competitive landscape of the future AI accelerator market. Discussions regarding investment in the foundry business are still ongoing, with specific details and potential partners not yet disclosed.
@@@
According to Sina Tech, Tencent Holdings released its full-year 2025 financial report on March 18, 2026. The data shows its business-facing segment, the To B business, achieved annual revenue of 229.4 billion RMB, reaching a record high. Notably, Tencent Cloud achieved profitability for the full year of 2025, marking a new development stage for the business after sustained investment. The report further indicated that the FinTech and Business Services segment, as an important growth engine for the company, continues to increase its revenue proportion, reflecting substantial progress in the company's deep cultivation of the industrial internet sector. Meanwhile, Tencent maintained high investment in research and development to solidify its competitiveness in core technology areas like cloud computing and AI. The release of this earnings report demonstrates to the market the robust financial performance and sustained growth potential of Tencent's enterprise services business, beyond its consumer internet operations.
@@@
According to TechWeb, AI company OpenAI officially released its new generation of small-scale models, GPT‑5.4 mini and GPT‑5.4 nano, on March 18 this year. These are the most capable small models the company has released to date, designed to provide developers with powerful language processing capabilities at lower cost and higher efficiency. The new models performed excellently in multiple core benchmarks, particularly in code generation, mathematical reasoning, and language understanding tasks, with performance approaching or even surpassing some larger predecessor models. GPT‑5.4 mini significantly optimizes response speed and computational resource consumption while maintaining high performance, while the even more lightweight nano version is specifically designed for resource-constrained edge devices and mobile application scenarios. OpenAI stated this release responds to the broad market demand for efficient, affordable AI tools, helping developers more easily integrate advanced AI capabilities into various applications. Both new models are now available to developers via API, adopting a flexible pay-as-you-go pricing model. The company emphasized that the new models also incorporate robust safety and alignment mechanisms to mitigate potential risks. This product line update is seen as a key move by OpenAI to expand its technology reach and serve a broader base of developers and enterprise users.
@@@
According to tech media cnBeta, NVIDIA CEO Jensen Huang recently personally confirmed that the company has resumed production of its H200 AI chips for the Chinese market and has begun supplying several Chinese customers. This news emerged against the backdrop of widespread attention on the impact of U.S. export control policies on high-end chip sales in China. Huang stated that NVIDIA is providing compliant products for the Chinese market while strictly adhering to relevant trade regulations. The H200 is a high-performance GPU from NVIDIA widely used in AI training and inference. The resumption of production and supply marks progress for NVIDIA in adjusting its product strategy to meet specific market requirements. The company did not disclose which specific Chinese customers purchased these chips, nor the specific quantities or amounts involved. Industry observers note this development is significant for China's AI industry, which relies on high-end computing power. Currently, NVIDIA continues to communicate with all parties to ensure its commercial activities fully comply with international rules.
@@@
According to TechWeb, Alibaba Group's cloud computing division, Alibaba Cloud, recently officially announced price adjustments for some of its products and services. This adjustment involves core products across multiple areas including Elastic Compute instances, Object Storage Service, and database products, with prices for some products increasing to varying degrees. Alibaba Cloud stated that this price adjustment is based on a comprehensive consideration of factors including continuously rising global market costs for raw materials and energy, increased investment in core technology R&D, and the need to upgrade infrastructure to ensure service quality. The company emphasized its continued commitment to providing stable and efficient cloud services and plans to further increase investment in technological innovation and global data center network expansion in the future. The price adjustments are expected to take effect gradually starting next quarter. Alibaba Cloud has begun notifying potentially affected customer groups and advises users to assess potential impacts based on their own business needs.
@@@
According to Readhub, Wang Xingxing, founder of Unitree Robotics, recently stated that within this year, the humanoid robot developed by the company is expected to surpass the world sprint champion Usain Bolt in running speed. Bolt's men's 100-meter world record stands at 9.58 seconds, which requires the robot to possess extremely high dynamic balance capabilities and explosive power. As a company focused on high-performance bionic robots, Unitree's latest generation products have continued to make breakthroughs in joint motors, motion control algorithms, and overall machine stability. Wang Xingxing pointed out that achieving a running speed beyond human limits is a significant milestone in the development of robotic mobility, showcasing not only technological progress but also laying a key foundation for the future practical application of robots in complex, unstructured environments. The company's team is dedicated to optimizing the robot's gait and energy efficiency to achieve this highly challenging goal. This progress also reflects the accelerating trend in the global humanoid robot industry's competition on movement performance.
@@@
According to the U.S. tech news website cnBeta, the Pentagon is planning to launch a new initiative that would allow approved commercial AI companies to train their models on highly classified military data. This plan aims to leverage the private sector's innovation capabilities in AI to accelerate the development of cutting-edge AI tools for defense. Under this plan, AI enterprises meeting security standards will be able to access sensitive information previously restricted to internal government use, under conditions ensuring absolute data security. U.S. Department of Defense officials stated this move is to bridge the gap between the rapid development of commercial AI and defense application needs, ensuring the U.S. military can access and integrate the most advanced AI capabilities. The plan is expected to establish strict security protocols and review processes, with only companies passing rigorous background checks and meeting a series of cybersecurity requirements able to participate. Analysis suggests this initiative could open new paths for the application of AI technology in critical areas such as military command, intelligence analysis, and logistics planning. Currently, detailed implementation frameworks and timelines for the related plan are still being developed.
@@@
According to tech media cnBeta, the U.S. Federal Aviation Administration has officially approved electric vertical take-off and landing aircraft, or eVTOL air taxis, to commence commercial operations across the United States this summer. This decision marks the entry of urban air mobility into a new commercial phase. Aircraft models receiving operational approval must fully comply with the agency's stringent safety and airworthiness standards, which are on par with those for traditional civil aviation aircraft. The first air taxis to enter service are expected to be deployed in specific urban areas, primarily for transfers between airports and city centers, as well as short-distance travel within cities. These aircraft can take off and land vertically without needing traditional runways and are designed for zero-emission operation while reducing noise. Industry analysis indicates this approval paves the way for related manufacturers and operators, potentially alleviating ground traffic congestion. However, specific fares, operational routes, and broader urban air traffic management details still await further announcements from relevant companies and local authorities.
@@@
Alright, that's all for today's episode. I hope today's sharing brought you a little inspiration or warmth. Thank you for listening and for your company! If you enjoyed our show, we'd love for you to subscribe and share it with your friends—it helps us a lot. See you next time!
