文章目录AI科技热点日报 | 2026年6月1日 今日摘要一、大模型与基础技术 | Foundation Models Core Tech事件概要来源 / Sources二、国产大模型突破 | Domestic LLM Breakthroughs事件概要来源 / Sources三、AI企业融资与上市 | AI Funding IPO事件概要来源 / Sources四、AI算力与基础设施 | AI Computing Power Infrastructure事件概要来源 / Sources五、AI政策与监管 | AI Policy Regulation事件概要来源 / Sources六、AI应用与产业落地 | AI Applications Industry Deployment事件概要来源 / Sources行业数据速览媒体人解读参考来源英文来源中文来源 延伸阅读AI科技热点日报 | 2026年6月1日综合整理自The Verge、TechCrunch、Wired、MIT Technology Review、钛媒体、36氪、虎嗅网等国内外主流科技媒体当日报道聚焦人工智能领域核心动态附资深媒体人综合解读。 今日摘要今日AI产业迎来多项重磅事件英伟达GTC Taipei 2026大会开幕黄仁勋发表主题演讲并发布面向AI PC的Vera Rubin芯片国产大模型MiniMax M3正式发布编程能力超越GPT-5.5并启动A股IPO辅导Anthropic完成650亿美元融资估值达9650亿美元超越OpenAI。同时美国再加码芯片出口管制COMPUTEX 2026盛大开幕全球AI算力竞争进入新阶段。Today AI industry witnessed multiple major events: Nvidia’s GTC Taipei 2026 kicked off with Jensen Huang’s keynote announcing Vera Rubin chip for AI PCs; China’s MiniMax M3 officially launched with coding ability surpassing GPT-5.5 and started A-share IPO辅导; Anthropic raised $65 billion in funding, reaching $965 billion valuation to surpass OpenAI. Meanwhile, the US imposed new chip export restrictions, and COMPUTEX 2026 opened with the global AI computing power competition entering a new phase.一、大模型与基础技术 | Foundation Models Core Tech事件概要英伟达GTC Taipei 2026大会今日在中国台北国际会议中心正式开幕创始人兼CEO黄仁勋于上午11点发表主题演讲揭晓驱动新一代AI的突破性技术进展。会上正式发布Vera Rubin芯片这是代号Rubin的全新处理器系列专为代理式AIAgentic AI设计OpenAI和Anthropic已率先采用。同时英伟达推出NVIDIA DSX™平台提供AI工厂全栈基础设施方案。此外英伟达宣布与微软合作推出采用ARM架构的Windows PC芯片RTX SPARK采用台积电3nm工艺计划于2026年秋季上市标志着英伟达正式进军个人电脑芯片市场打破英特尔在该领域的垄断地位。Nvidia’s GTC Taipei 2026 conference officially opened at the Taipei International Convention Center today, with founder and CEO Jensen Huang delivering a keynote at 11 AM announcing breakthrough technologies powering next-generation AI. The company officially unveiled the Vera Rubin chip, a new processor series code-named Rubin designed specifically for Agentic AI, already adopted by OpenAI and Anthropic. Nvidia also launched the NVIDIA DSX™ platform providing full-stack AI factory infrastructure solutions. Additionally, Nvidia announced a partnership with Microsoft to launch RTX SPARK, an ARM-based Windows PC chip using TSMC’s 3nm process, scheduled for fall 2026 release, marking Nvidia’s official entry into the personal computer chip market to break Intel’s monopoly in this sector.来源 / Sources英伟达GTC Taipei 2026大会开幕新京报2026年6月1日英伟达发布Vera Rubin芯片CSDN2026年6月1日英伟达新PC芯片秋季发布和讯科技2026年6月1日二、国产大模型突破 | Domestic LLM Breakthroughs事件概要国产大模型公司MiniMax于6月1日正式发布新一代通用模型MiniMax M3这是国内首个同时具备前沿Coding能力、1M超长上下文、原生多模态三项核心能力的大模型。M3采用自研稀疏注意力架构MiniMax Sparse AttentionMSA在100万上下文规模下单token计算量仅为上一代模型的约1/20推理效率提升显著。在编程能力基准测试SWE-Bench Pro上MiniMax M3超过OpenAI的GPT-5.5和谷歌Gemini 3.1 Pro接近Claude的Opus 4.7。MiniMax同步推出AI编程产品MiniMax Code智能体集群可将大型任务拆解为多个可并发、可动态调整的阶段自主运行数天而无需人工干预。China’s AI company MiniMax officially launched its new general-purpose model MiniMax M3 on June 1st, China’s first large model with three core capabilities: frontier coding ability, 1M long context, and native multimodal. M3 uses MiniMax’s self-developed Sparse Attention (MSA) architecture, reducing per-token computation to about 1/20 of the previous generation at 1M context scale, significantly improving inference efficiency. On the programming capability benchmark SWE-Bench Pro, MiniMax M3 surpassed OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro, approaching Claude Opus 4.7. MiniMax also launched AI programming product MiniMax Code, whose agent cluster can split large tasks into multiple concurrent and dynamically adjustable phases, running autonomously for days without human intervention.来源 / SourcesMiniMax推出MiniMax M3模型新京报2026年6月1日MiniMax发布新一代通用模型腾讯新闻2026年6月1日MiniMax新模型编程能力超越GPT-5.5腾讯新闻2026年6月1日三、AI企业融资与上市 | AI Funding IPO事件概要美国AI公司Anthropic于5月28日宣布完成650亿美元H轮融资投后估值达9650亿美元首次超越OpenAI成为全球估值最高的人工智能公司。本轮融资吸引美光、三星、SK海力士等存储芯片巨头联合参与。Anthropic CEO达里奥·阿莫迪披露公司年化收入已从2022年的1000万美元增长至2026年5月的超440亿美元四年间增长超过4000倍。同时MiniMax于6月1日启动A股IPO辅导目标科创板上市这是其港股上市仅半年后的又一重大资本决策公司全球用户约3亿年化经常性收入过去两个月翻倍至超3亿美元。US AI company Anthropic announced on May 28th the completion of a $65 billion Series H funding round, achieving a post-money valuation of $965 billion, surpassing OpenAI for the first time to become the world’s highest-valued AI company. The round attracted memory chip giants including Micron, Samsung, and SK Hynix as co-investors. Anthropic CEO Dario Amodei disclosed the company’s annualized revenue grew from $10 million in 2022 to over $44 billion in May 2026, a more than 4000-fold increase in four years. Meanwhile, MiniMax started A-share IPO tutoring on June 1st targeting the STAR Market, another major capital decision just six months after its HK listing, with approximately 300 million global users and annualized recurring revenue doubling to over $300 million in the past two months.来源 / SourcesAnthropic完成650亿美元融资新京报2026年5月29日反超OpenAI Anthropic估值逼近万亿美元腾讯新闻2026年5月31日MiniMax启动A股上市辅导腾讯新闻2026年6月1日四、AI算力与基础设施 | AI Computing Power Infrastructure事件概要COMPUTEX 2026今日在台北盛大开幕主题为AI Together英伟达、AMD、英特尔、高通、ARM、Marvell等芯片巨头齐聚预计数千家参展商参展创历史之最。与此同时中国电信研究院联合清华大学、无锡沐创集成电路设计有限公司成功完成基于国产抗量子芯片的AI多智能体可信通信创新试验标志着中国在抗量子密码与人工智能深度融合领域取得重要突破。深圳河套学院联合多家单位依托昇腾910C国产AI算力集群完成1.6万亿参数大模型DeepSeek-V4-Pro全流程训练印证国产AI芯片可支撑世界级超大参数模型训练工作。COMPUTEX 2026 opened grandly in Taipei today with the theme “AI Together”, gathering chip giants including Nvidia, AMD, Intel, Qualcomm, ARM, and Marvell, with thousands of exhibitors expected to set a historical record. Meanwhile, China Telecom Research Institute jointly with Tsinghua University and Wuxi Muchuang Integrated Circuit Design Co. successfully completed an AI multi-agent trusted communication innovation test based on domestically developed anti-quantum chips, marking a significant breakthrough in China’s deep integration of anti-quantum cryptography and artificial intelligence. Shenzhen Hetao College and multiple institutions completed full-process training of the 1.6 trillion parameter large model DeepSeek-V4-Pro using domestic Ascend 910C AI computing clusters, demonstrating that China’s AI chips can support world-class super-large parameter model training.来源 / SourcesComputex 2026重磅来袭腾讯新闻2026年6月1日中国电信研究院联合完成抗量子芯片试验和讯基金2026年6月1日深圳河套学院完成DeepSeek-V4-Pro训练深圳特区报2026年6月1日五、AI政策与监管 | AI Policy Regulation事件概要美国商务部于5月31日采取新行动封堵已存在一年的芯片出口监管漏洞防止多家企业向位于中国境外的中国实体出口包括英伟达最先进的Rubin和Blackwell处理器以及AMD的MI350x芯片在内的顶尖AI芯片。此举正值英伟达GTC Taipei大会召开之际显示美国持续加码对华科技遏制。与此同时发改委召开发布会指导国产大模型加大力度适配国产算力芯片阿里、腾讯等国内云服务商正加速承接这一政策导向推动国产算力生态建设。The US Department of Commerce took new action on May 31st to close a one-year-old chip export control “loophole” preventing companies from exporting top AI chips including Nvidia’s most advanced Rubin and Blackwell processors and AMD’s MI350x chips to Chinese entities located outside China. The move came as Nvidia’s GTC Taipei conference opened, showing the US continues to escalate tech containment against China. Meanwhile, China’s National Development and Reform Commission held a press conference guiding domestic large models to increase adaptation of domestic computing chips, as Alibaba, Tencent and other domestic cloud service providers accelerate implementing this policy direction to promote the domestic computing power ecosystem.来源 / Sources美方变本加厉堵漏洞新浪财经2026年6月1日[中国电信研究院抗量子芯片试验](https://www.toutiao.com/article/7646321742013350400/今日头条2026年6月1日六、AI应用与产业落地 | AI Applications Industry Deployment事件概要国产大模型双子星智谱与MiniMax继今年1月相继登陆港股后正加速推进A股上市进程形成AH双平台布局。MiniMax已于5月29日同中信证券签署辅导协议启动科创板IPO辅导智谱则已聘请国泰海通、中金公司完成第四期辅导。两家公司全球用户合计超6亿年化收入合计超百亿美元。在应用层面AI相关产品正加速出海上海临港的算力服务已覆盖柬埔寨吴哥古迹三维重建、全球用户专业大模型训练等跨境业务智能产品出口表现亮眼人工智能相关产品和服务正成为拉动外贸增长的新动能。China’s large model “twin stars” Zhipu and MiniMax, after listing on Hong Kong in January this year, are accelerating their A-share listing process, forming an “AH” dual-platform structure. MiniMax signed a tutoring agreement with CITIC Securities on May 29th to start STAR Market IPO tutoring, while Zhipu has completed the fourth tutoring phase with Guotai Haian and CICC. The combined global users of both companies exceed 600 million, with combined annualized revenue exceeding $10 billion. At the application level, AI-related products are accelerating overseas expansion, with Shanghai Lingang’s computing power services covering cross-border businesses including 3D reconstruction of Cambodia’s Angkor heritage and global user professional large model training, with smart product exports performing outstandingly and AI-related products and services becoming a new driving force for foreign trade growth.来源 / Sources国产大模型双子星携手回A新浪新闻2026年6月1日人工智能相关产品拉动外贸增长新浪财经2026年6月1日AI大模型双雄MiniMax、智谱均已开启A股IPO辅导新浪财经2026年6月1日行业数据速览指标数据备注Anthropic最新估值9650亿美元超越OpenAI成为全球最高Anthropic年化收入2026年5月超440亿美元4年增长超4000倍MiniMax全球用户约3亿ARR超3亿美元DeepSeek-V4-Pro参数量1.6万亿参数全部基于国产算力训练COMPUTEX 2026参展商数量数千家历史之最英伟达Vera Rubin芯片已发布OpenAI/Anthropic已采用媒体人解读今日AI产业呈现中美双线并进的格局在国际侧英伟达借GTC Taipei大会全面布局AI PC芯片市场Anthropic以650亿美元巨额融资刷新AI赛道融资纪录标志着全球AI竞争从模型能力向全栈生态转移的趋势进一步深化在国内侧MiniMax M3的发布证明国产大模型在编程与Agent能力上已具备国际竞争力且头部企业加速AH上市布局显示资本对AI落地价值的认可正在从估值泡沫回归基本面。值得警惕的是美国芯片出口管制持续加码国产算力替代虽已证明技术可行性但在高端芯片制造环节仍受制于人。总体而言2026年AI产业的竞争主线已从拼模型转向拼落地、拼效率、拼生态能真正解决B端问题的企业将笑到最后。The AI industry today presents a “China-US dual-track parallel” pattern: internationally, Nvidia comprehensively deployed AI PC chip market via GTC Taipei, and Anthropic’s $65 billion mega-funding set a new AI funding record, marking the deepening trend of global AI competition shifting from “model capability” to “full-stack ecosystem”; domestically, MiniMax M3’s launch proves Chinese LLMs have international competitiveness in coding and agent capabilities, and leading companies accelerating “AH” listing shows capital’s recognition of AI implementation value is returning from valuation bubbles to fundamentals. Notably, US chip export controls continue escalating, and while domestic computing power substitution has proven technical feasibility, high-end chip manufacturing remains constrained. Overall, the 2026 AI industry’s main competition line has shifted from “competing on models” to “competing on implementation, efficiency, and ecosystem,” and companies that can truly solve B-end problems will ultimately prevail.参考来源英文来源NVIDIA GTC Taipei at COMPUTEX 2026Nvidia官网中文来源英伟达GTC Taipei 2026大会开幕新京报MiniMax推出MiniMax M3模型新京报Anthropic反超OpenAI估值逼近万亿美元腾讯新闻MiniMax启动A股上市辅导腾讯新闻中国电信研究院联合完成抗量子芯片试验和讯基金深圳河套学院完成DeepSeek-V4-Pro训练深圳特区报美方变本加厉堵漏洞新浪财经人工智能相关产品拉动外贸增长新浪财经本文内容综合整理自上述公开信息来源媒体人解读仅代表个人观点仅供参考。发布日期2026年6月1日 延伸阅读一人公司OPC典型案例与商业模式研究报告主流 Claw 类产品技术与商业差异全景报告常见误区管理者最容易踩的 5 个坑团队 AI 成熟度自评模型
AI科技热点日报 | 2026年6月1日
文章目录AI科技热点日报 | 2026年6月1日 今日摘要一、大模型与基础技术 | Foundation Models Core Tech事件概要来源 / Sources二、国产大模型突破 | Domestic LLM Breakthroughs事件概要来源 / Sources三、AI企业融资与上市 | AI Funding IPO事件概要来源 / Sources四、AI算力与基础设施 | AI Computing Power Infrastructure事件概要来源 / Sources五、AI政策与监管 | AI Policy Regulation事件概要来源 / Sources六、AI应用与产业落地 | AI Applications Industry Deployment事件概要来源 / Sources行业数据速览媒体人解读参考来源英文来源中文来源 延伸阅读AI科技热点日报 | 2026年6月1日综合整理自The Verge、TechCrunch、Wired、MIT Technology Review、钛媒体、36氪、虎嗅网等国内外主流科技媒体当日报道聚焦人工智能领域核心动态附资深媒体人综合解读。 今日摘要今日AI产业迎来多项重磅事件英伟达GTC Taipei 2026大会开幕黄仁勋发表主题演讲并发布面向AI PC的Vera Rubin芯片国产大模型MiniMax M3正式发布编程能力超越GPT-5.5并启动A股IPO辅导Anthropic完成650亿美元融资估值达9650亿美元超越OpenAI。同时美国再加码芯片出口管制COMPUTEX 2026盛大开幕全球AI算力竞争进入新阶段。Today AI industry witnessed multiple major events: Nvidia’s GTC Taipei 2026 kicked off with Jensen Huang’s keynote announcing Vera Rubin chip for AI PCs; China’s MiniMax M3 officially launched with coding ability surpassing GPT-5.5 and started A-share IPO辅导; Anthropic raised $65 billion in funding, reaching $965 billion valuation to surpass OpenAI. Meanwhile, the US imposed new chip export restrictions, and COMPUTEX 2026 opened with the global AI computing power competition entering a new phase.一、大模型与基础技术 | Foundation Models Core Tech事件概要英伟达GTC Taipei 2026大会今日在中国台北国际会议中心正式开幕创始人兼CEO黄仁勋于上午11点发表主题演讲揭晓驱动新一代AI的突破性技术进展。会上正式发布Vera Rubin芯片这是代号Rubin的全新处理器系列专为代理式AIAgentic AI设计OpenAI和Anthropic已率先采用。同时英伟达推出NVIDIA DSX™平台提供AI工厂全栈基础设施方案。此外英伟达宣布与微软合作推出采用ARM架构的Windows PC芯片RTX SPARK采用台积电3nm工艺计划于2026年秋季上市标志着英伟达正式进军个人电脑芯片市场打破英特尔在该领域的垄断地位。Nvidia’s GTC Taipei 2026 conference officially opened at the Taipei International Convention Center today, with founder and CEO Jensen Huang delivering a keynote at 11 AM announcing breakthrough technologies powering next-generation AI. The company officially unveiled the Vera Rubin chip, a new processor series code-named Rubin designed specifically for Agentic AI, already adopted by OpenAI and Anthropic. Nvidia also launched the NVIDIA DSX™ platform providing full-stack AI factory infrastructure solutions. Additionally, Nvidia announced a partnership with Microsoft to launch RTX SPARK, an ARM-based Windows PC chip using TSMC’s 3nm process, scheduled for fall 2026 release, marking Nvidia’s official entry into the personal computer chip market to break Intel’s monopoly in this sector.来源 / Sources英伟达GTC Taipei 2026大会开幕新京报2026年6月1日英伟达发布Vera Rubin芯片CSDN2026年6月1日英伟达新PC芯片秋季发布和讯科技2026年6月1日二、国产大模型突破 | Domestic LLM Breakthroughs事件概要国产大模型公司MiniMax于6月1日正式发布新一代通用模型MiniMax M3这是国内首个同时具备前沿Coding能力、1M超长上下文、原生多模态三项核心能力的大模型。M3采用自研稀疏注意力架构MiniMax Sparse AttentionMSA在100万上下文规模下单token计算量仅为上一代模型的约1/20推理效率提升显著。在编程能力基准测试SWE-Bench Pro上MiniMax M3超过OpenAI的GPT-5.5和谷歌Gemini 3.1 Pro接近Claude的Opus 4.7。MiniMax同步推出AI编程产品MiniMax Code智能体集群可将大型任务拆解为多个可并发、可动态调整的阶段自主运行数天而无需人工干预。China’s AI company MiniMax officially launched its new general-purpose model MiniMax M3 on June 1st, China’s first large model with three core capabilities: frontier coding ability, 1M long context, and native multimodal. M3 uses MiniMax’s self-developed Sparse Attention (MSA) architecture, reducing per-token computation to about 1/20 of the previous generation at 1M context scale, significantly improving inference efficiency. On the programming capability benchmark SWE-Bench Pro, MiniMax M3 surpassed OpenAI’s GPT-5.5 and Google’s Gemini 3.1 Pro, approaching Claude Opus 4.7. MiniMax also launched AI programming product MiniMax Code, whose agent cluster can split large tasks into multiple concurrent and dynamically adjustable phases, running autonomously for days without human intervention.来源 / SourcesMiniMax推出MiniMax M3模型新京报2026年6月1日MiniMax发布新一代通用模型腾讯新闻2026年6月1日MiniMax新模型编程能力超越GPT-5.5腾讯新闻2026年6月1日三、AI企业融资与上市 | AI Funding IPO事件概要美国AI公司Anthropic于5月28日宣布完成650亿美元H轮融资投后估值达9650亿美元首次超越OpenAI成为全球估值最高的人工智能公司。本轮融资吸引美光、三星、SK海力士等存储芯片巨头联合参与。Anthropic CEO达里奥·阿莫迪披露公司年化收入已从2022年的1000万美元增长至2026年5月的超440亿美元四年间增长超过4000倍。同时MiniMax于6月1日启动A股IPO辅导目标科创板上市这是其港股上市仅半年后的又一重大资本决策公司全球用户约3亿年化经常性收入过去两个月翻倍至超3亿美元。US AI company Anthropic announced on May 28th the completion of a $65 billion Series H funding round, achieving a post-money valuation of $965 billion, surpassing OpenAI for the first time to become the world’s highest-valued AI company. The round attracted memory chip giants including Micron, Samsung, and SK Hynix as co-investors. Anthropic CEO Dario Amodei disclosed the company’s annualized revenue grew from $10 million in 2022 to over $44 billion in May 2026, a more than 4000-fold increase in four years. Meanwhile, MiniMax started A-share IPO tutoring on June 1st targeting the STAR Market, another major capital decision just six months after its HK listing, with approximately 300 million global users and annualized recurring revenue doubling to over $300 million in the past two months.来源 / SourcesAnthropic完成650亿美元融资新京报2026年5月29日反超OpenAI Anthropic估值逼近万亿美元腾讯新闻2026年5月31日MiniMax启动A股上市辅导腾讯新闻2026年6月1日四、AI算力与基础设施 | AI Computing Power Infrastructure事件概要COMPUTEX 2026今日在台北盛大开幕主题为AI Together英伟达、AMD、英特尔、高通、ARM、Marvell等芯片巨头齐聚预计数千家参展商参展创历史之最。与此同时中国电信研究院联合清华大学、无锡沐创集成电路设计有限公司成功完成基于国产抗量子芯片的AI多智能体可信通信创新试验标志着中国在抗量子密码与人工智能深度融合领域取得重要突破。深圳河套学院联合多家单位依托昇腾910C国产AI算力集群完成1.6万亿参数大模型DeepSeek-V4-Pro全流程训练印证国产AI芯片可支撑世界级超大参数模型训练工作。COMPUTEX 2026 opened grandly in Taipei today with the theme “AI Together”, gathering chip giants including Nvidia, AMD, Intel, Qualcomm, ARM, and Marvell, with thousands of exhibitors expected to set a historical record. Meanwhile, China Telecom Research Institute jointly with Tsinghua University and Wuxi Muchuang Integrated Circuit Design Co. successfully completed an AI multi-agent trusted communication innovation test based on domestically developed anti-quantum chips, marking a significant breakthrough in China’s deep integration of anti-quantum cryptography and artificial intelligence. Shenzhen Hetao College and multiple institutions completed full-process training of the 1.6 trillion parameter large model DeepSeek-V4-Pro using domestic Ascend 910C AI computing clusters, demonstrating that China’s AI chips can support world-class super-large parameter model training.来源 / SourcesComputex 2026重磅来袭腾讯新闻2026年6月1日中国电信研究院联合完成抗量子芯片试验和讯基金2026年6月1日深圳河套学院完成DeepSeek-V4-Pro训练深圳特区报2026年6月1日五、AI政策与监管 | AI Policy Regulation事件概要美国商务部于5月31日采取新行动封堵已存在一年的芯片出口监管漏洞防止多家企业向位于中国境外的中国实体出口包括英伟达最先进的Rubin和Blackwell处理器以及AMD的MI350x芯片在内的顶尖AI芯片。此举正值英伟达GTC Taipei大会召开之际显示美国持续加码对华科技遏制。与此同时发改委召开发布会指导国产大模型加大力度适配国产算力芯片阿里、腾讯等国内云服务商正加速承接这一政策导向推动国产算力生态建设。The US Department of Commerce took new action on May 31st to close a one-year-old chip export control “loophole” preventing companies from exporting top AI chips including Nvidia’s most advanced Rubin and Blackwell processors and AMD’s MI350x chips to Chinese entities located outside China. The move came as Nvidia’s GTC Taipei conference opened, showing the US continues to escalate tech containment against China. Meanwhile, China’s National Development and Reform Commission held a press conference guiding domestic large models to increase adaptation of domestic computing chips, as Alibaba, Tencent and other domestic cloud service providers accelerate implementing this policy direction to promote the domestic computing power ecosystem.来源 / Sources美方变本加厉堵漏洞新浪财经2026年6月1日[中国电信研究院抗量子芯片试验](https://www.toutiao.com/article/7646321742013350400/今日头条2026年6月1日六、AI应用与产业落地 | AI Applications Industry Deployment事件概要国产大模型双子星智谱与MiniMax继今年1月相继登陆港股后正加速推进A股上市进程形成AH双平台布局。MiniMax已于5月29日同中信证券签署辅导协议启动科创板IPO辅导智谱则已聘请国泰海通、中金公司完成第四期辅导。两家公司全球用户合计超6亿年化收入合计超百亿美元。在应用层面AI相关产品正加速出海上海临港的算力服务已覆盖柬埔寨吴哥古迹三维重建、全球用户专业大模型训练等跨境业务智能产品出口表现亮眼人工智能相关产品和服务正成为拉动外贸增长的新动能。China’s large model “twin stars” Zhipu and MiniMax, after listing on Hong Kong in January this year, are accelerating their A-share listing process, forming an “AH” dual-platform structure. MiniMax signed a tutoring agreement with CITIC Securities on May 29th to start STAR Market IPO tutoring, while Zhipu has completed the fourth tutoring phase with Guotai Haian and CICC. The combined global users of both companies exceed 600 million, with combined annualized revenue exceeding $10 billion. At the application level, AI-related products are accelerating overseas expansion, with Shanghai Lingang’s computing power services covering cross-border businesses including 3D reconstruction of Cambodia’s Angkor heritage and global user professional large model training, with smart product exports performing outstandingly and AI-related products and services becoming a new driving force for foreign trade growth.来源 / Sources国产大模型双子星携手回A新浪新闻2026年6月1日人工智能相关产品拉动外贸增长新浪财经2026年6月1日AI大模型双雄MiniMax、智谱均已开启A股IPO辅导新浪财经2026年6月1日行业数据速览指标数据备注Anthropic最新估值9650亿美元超越OpenAI成为全球最高Anthropic年化收入2026年5月超440亿美元4年增长超4000倍MiniMax全球用户约3亿ARR超3亿美元DeepSeek-V4-Pro参数量1.6万亿参数全部基于国产算力训练COMPUTEX 2026参展商数量数千家历史之最英伟达Vera Rubin芯片已发布OpenAI/Anthropic已采用媒体人解读今日AI产业呈现中美双线并进的格局在国际侧英伟达借GTC Taipei大会全面布局AI PC芯片市场Anthropic以650亿美元巨额融资刷新AI赛道融资纪录标志着全球AI竞争从模型能力向全栈生态转移的趋势进一步深化在国内侧MiniMax M3的发布证明国产大模型在编程与Agent能力上已具备国际竞争力且头部企业加速AH上市布局显示资本对AI落地价值的认可正在从估值泡沫回归基本面。值得警惕的是美国芯片出口管制持续加码国产算力替代虽已证明技术可行性但在高端芯片制造环节仍受制于人。总体而言2026年AI产业的竞争主线已从拼模型转向拼落地、拼效率、拼生态能真正解决B端问题的企业将笑到最后。The AI industry today presents a “China-US dual-track parallel” pattern: internationally, Nvidia comprehensively deployed AI PC chip market via GTC Taipei, and Anthropic’s $65 billion mega-funding set a new AI funding record, marking the deepening trend of global AI competition shifting from “model capability” to “full-stack ecosystem”; domestically, MiniMax M3’s launch proves Chinese LLMs have international competitiveness in coding and agent capabilities, and leading companies accelerating “AH” listing shows capital’s recognition of AI implementation value is returning from valuation bubbles to fundamentals. Notably, US chip export controls continue escalating, and while domestic computing power substitution has proven technical feasibility, high-end chip manufacturing remains constrained. Overall, the 2026 AI industry’s main competition line has shifted from “competing on models” to “competing on implementation, efficiency, and ecosystem,” and companies that can truly solve B-end problems will ultimately prevail.参考来源英文来源NVIDIA GTC Taipei at COMPUTEX 2026Nvidia官网中文来源英伟达GTC Taipei 2026大会开幕新京报MiniMax推出MiniMax M3模型新京报Anthropic反超OpenAI估值逼近万亿美元腾讯新闻MiniMax启动A股上市辅导腾讯新闻中国电信研究院联合完成抗量子芯片试验和讯基金深圳河套学院完成DeepSeek-V4-Pro训练深圳特区报美方变本加厉堵漏洞新浪财经人工智能相关产品拉动外贸增长新浪财经本文内容综合整理自上述公开信息来源媒体人解读仅代表个人观点仅供参考。发布日期2026年6月1日 延伸阅读一人公司OPC典型案例与商业模式研究报告主流 Claw 类产品技术与商业差异全景报告常见误区管理者最容易踩的 5 个坑团队 AI 成熟度自评模型
