文章

卫星化学净利翻倍 · 孙正义驳斥太空数据中心

#844 · 2026-06-23 · 21ZHAO Blog
Reading Path / ARTICLE 先抓主张，再转成行动 #844 · 21ZHAO Blog · 读完进入产品或下一篇

卫星化学净利翻倍 · 孙正义驳斥太空数据中心

一、 权威必看

EN: Satellite Chemical (002648) experienced a limit-up stock price movement on June 23, reaching 25.18 yuan per share with a total market capitalization of 84.8 billion yuan. The company announced its 2026 first-half performance forecast, projecting a net profit between 6 billion and 7 billion yuan, representing a year-on-year increase of 118.68% to 155.13%. Even at the lower bound of this forecast, both net profit and non-GAAP net profit exceed historical peaks for the same period. The company attributes this success to its integrated light hydrocarbon industry chain, stable production operations, product iteration, and cost optimization strategies that effectively buffered against international crude oil price volatility.

中： 卫星化学（002648）于6月23日股价一字涨停，报收25.18元/股，总市值攀升至848亿元。公司发布2026年上半年业绩预报，预计净利润60亿元至70亿元，同比增长118.68%至155.13%；扣非净利润56.87亿元至66.87亿元，同比增长96.37%至130.9%。即便按业绩预告下限计算，公司净利润与扣非净利润均超越历史同期最高水平。作为国内轻烃产业链一体化龙头企业，卫星化学通过上游原料自给与下游产品延伸的深度协同，形成了碳二、碳三双产业链布局。在国际原油价格受地缘因素推动持续高位震荡的背景下，油制烯烃路径成本骤升，而卫星化学采用的乙烷裂解路线优势凸显。公司充分利用产业链一体化优势，生产装置保持安全稳定运行，通过供应链多元化布局及审慎运用金融工具，有力缓冲了成本波动，确保了主营业务盈利能力的稳定性。

二、 深度与多元

EN: Masayoshi Son, founder of SoftBank Group, publicly challenged Elon Musk’s concept of “space data centers” during an annual shareholder meeting for SoftBank’s telecom subsidiary on June 23. Son argued that the winner of the AI race will be determined by computing power on Earth, stating that space data centers have little value. He pointed out that while reducing electricity costs is a primary advantage of space data centers, electricity accounts for only about 7% of the total operating cost of AI data centers compared to hardware costs like chips. The high costs of space transportation, maintenance, and communication latency outweigh the benefits. Son emphasized that SoftBank will focus on building “massive” ground-based data center computing power globally, noting that results in the space sector are still over a decade away while the AI competition will be decided within ten years.

中： 日本软银集团创始人孙正义在6月23日举行的软银集团旗下电信子公司年度股东大会上，对埃隆·马斯克倡导的“太空数据中心”理念公开提出异议。孙正义指出，人工智能竞赛的胜负将取决于地球上的算力，太空数据中心价值寥寥。他分析称，在太空建设数据中心的主要优势在于大幅降低电力成本，但与芯片等硬件成本相比，电力成本仅占人工智能数据中心运营总成本的7%左右。降低电力成本的代价是更高的太空运输费用、维护成本以及通信延迟，这导致其得不偿失。孙正义称赞马斯克是“杰出的变革者”，但表示软银将专注于在全球打造“规模庞大”的地面数据中心算力。他提到，在太空领域要看到成果还需要十多年时间，而人工智能领域的竞争将在未来十年左右见分晓。软银公司总裁宫川纯一也表示，近期内不会参与此类项目，并指出日本数据中心建设势头较弱，公司有责任为日本人工智能的发展创造有利环境。

三、 科技与财经

EN: The tech community is witnessing a significant shift in developer tooling preferences, with a notable trend of developers pivoting towards domestic AI models for coding agents. A prominent example is the emergence of “Orca,” a terminal coding agent written in Rust that is natively built around DeepSeek. Developers are highlighting DeepSeek’s cost-effectiveness and performance capabilities as key drivers for this migration away from expensive overseas models like Claude or GPT. This shift reflects a broader economic calculation within the developer community, where reducing API costs while maintaining high-quality code generation becomes a strategic priority. The open-source community is actively exploring these alternatives to build more affordable and efficient development workflows.

中： 科技财经领域正见证开发者工具偏好的显著转变，大量开发者开始将重心转向国内AI模型以构建编码Agent。一个典型代表是使用Rust编写的、原生围绕DeepSeek打造的终端Coding Agent “Orca”。开发者们强调DeepSeek在成本效益和性能表现上的优势，推动其从Claude或GPT等昂贵海外模型中迁移。这种转变反映了开发者社区内部更广泛的经济计算，即在保持高质量代码生成的同时，降低API调用成本成为战略重点。开源社区正积极探索这些替代方案，以构建更具性价比且高效的开发工作流。与此同时，微博热搜话题“叫全名是一件很微妙的事”也引发了网民关于职场沟通与社交边界的广泛讨论，反映出当代社会对人际互动细节的高度敏感。

四、 国际视野

EN: The global tech landscape is seeing increased competition in the agent framework space, with developers evaluating various open-source options. On platforms like V2EX and Linux.do, discussions highlight the disparity in popularity between different coding agents. While tools like Codex and Claude Code dominate the conversation, newer entrants like OpenCode are facing questions about their adoption rates despite offering free model access. Developers are scrutinizing the user experience, integration capabilities, and community support of these tools to determine which platform offers the best long-term value. This international discourse underscores the rapid evolution of AI-assisted development tools and the critical role of community feedback in shaping market leaders.

中： 国际视野下，全球科技圈在Agent框架领域正经历激烈的竞争，开发者们正在评估各种开源选项。在V2EX和Linux.do等平台上，讨论突显了不同编码Agent之间受欢迎程度的差异。虽然Codex和Claude Code主导了对话，但尽管OpenCode提供免费模型访问，其采用率仍受到质疑。开发者们仔细审视这些工具的用户体验、集成能力以及社区支持，以确定哪个平台能提供最佳的长期价值。这种国际 discourse 强调了AI辅助开发工具的快速演变，以及社区反馈在塑造市场领导者中的关键作用。与此同时，关于DeepSeek的押注也显示出全球开发者对多元化技术栈的兴趣，不再单一依赖某一家海外巨头，而是根据性能和成本灵活选择。

五、 青年与生活

EN: Young developers and tech enthusiasts are actively engaging in discussions about the practical applications of AI agents in daily coding workflows. The topic “Why does OpenCode not seem to have such high popularity?” on Linux.do reflects the community’s critical eye towards new tools. Users are comparing the ease of use, feature set, and ecosystem integration of OpenCode against established competitors like Codex and Claude Code. This scrutiny is not merely about hype but about genuine utility and reliability in professional settings. The conversation reveals a mature approach to technology adoption, where developers prioritize stability, cost-efficiency, and community trust over initial marketing buzz.

中： 青年开发者与技术爱好者正积极参与关于AI Agent在日常编码工作流中实际应用的讨论。Linux.do上的话题“为什么感觉opencode的热度不是很高呢”反映了社区对新工具的批判性眼光。用户将OpenCode的易用性、功能集和生态系统集成与Codex和Claude Code等成熟竞品进行比较。这种审视并非仅仅出于炒作，而是关乎在专业环境中的真正效用和可靠性。对话揭示了一种成熟的技术采用态度，开发者优先考虑稳定性、成本效益和社区信任，而非初期的营销热度。这种理性选择也体现在对DeepSeek等国产模型的支持上，显示出青年群体在技术选型上的独立思考和务实精神。

【21ZHAO 综合判断】

EN:

  • Analyze the cost-benefit ratio of infrastructure choices: Just as SoftBank highlights that electricity is only 7% of data center costs, developers should evaluate the total cost of ownership for AI tools, including API fees, latency, and integration effort, rather than just the headline price.
  • Diversify your tech stack strategically: The shift towards models like DeepSeek demonstrates the value of not relying on a single provider. Developers should build agents that are modular and capable of switching underlying models to mitigate vendor lock-in risks and optimize for performance-to-cost ratios.

中：

  • 深入分析基础设施选择的成本效益比：正如孙正义指出电力仅占数据中心成本的7%，开发者在评估AI工具时，应全面考量总拥有成本，包括API费用、延迟和集成工作量，而不仅仅是表面价格。
  • 战略性地多样化技术栈：向DeepSeek等模型的趋势转变证明了不依赖单一供应商的价值。开发者应构建模块化且能切换底层模型的Agent，以规避供应商锁定风险并优化性能与成本比率。

参考来源