文章目录
- AI科技热点日报 | 2026年6月4日
- 📌 今日摘要
- 一、微软 Build 2026：七款自研 MAI 模型齐发，"Agent 优先"时代到来
- 事件概要
- 来源 / Sources
- 二、博通 2026 财年 AI 芯片销售额预测达 560 亿美元，已向 OpenAI 交付定制芯片
- 事件概要
- 来源 / Sources
- 三、比亚迪"尧舜禹"人形机器人项目首度曝光，2026 年内部部署 2 万台
- 事件概要
- 来源 / Sources
- 四、千寻智能 3 个月累计融资近 50 亿元，Spirit v1.6 击败英伟达登顶 RoboArena
- 事件概要
- 来源 / Sources
- 五、阿里双线推进 Agent 生态：千问向第三方开放 + QoderWork CN 推 AI 生产力计划
- 事件概要
- 来源 / Sources
- 六、AI 三巨头 IPO 超级周期：合计融资近 2000 亿美元，联合国报告警示 AI 能耗压力
- 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4日
综合整理自 The Verge、TechCrunch、新华网科技、财联社、IT 之家、上证报、亿邦动力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
微软 Build 2026 大会一次性发布 7 款自研 MAI 模型，"Agent 优先"时代正式宣告；博通预测 2026 财年 AI 芯片销售额达 560 亿美元，已向 OpenAI 交付定制芯片；比亚迪执行副总裁李柯首次证实代号"尧舜禹"的人形机器人项目，2026 年内部部署 2 万台；具身智能公司千寻智能完成 15 亿元 A+ 轮融资，三个月内累计融资近 50 亿元，旗下 Spirit v1.6 登顶 RoboArena 力压英伟达 Cosmos3；阿里旗下千问 APP 宣布向第三方 Agent、Skill 全面开放，肯德基、瑞幸、蜜雪冰城、东航首批接入；联合国大学发布报告警告，到 2030 年全球数据中心用电量将达 945 太瓦时，AI 扩张正加剧能源、水资源与土地压力。
At Microsoft Build 2026, the company unveiled seven self-developed MAI models in a single launch, formally ushering in the “Agent-first” era. Broadcom projected its fiscal year 2026 AI chip revenue at $56 billion, having already delivered custom chips to OpenAI. BYD Executive Vice President Li Ke confirmed for the first time the company’s “Yao Shun Yu” humanoid robot program, with an internal deployment plan of 20,000 units in 2026. Embodied AI startup Qianxun Intelligence closed a 1.5 billion yuan Series A+ round, bringing its three-month cumulative financing to nearly 5 billion yuan, while its Spirit v1.6 model topped the RoboArena leaderboard, outperforming NVIDIA Cosmos 3. Alibaba’s Qianwen App announced full opening to third-party Agents and Skills, with KFC, Luckin Coffee, Mixue Bingcheng, and China Eastern Airlines becoming the first batch of adopters. The United Nations University issued a report warning that global data center electricity demand will reach 945 TWh by 2030, with AI expansion intensifying pressure on energy, water, and land resources.
一、微软 Build 2026：七款自研 MAI 模型齐发，"Agent 优先"时代到来
事件概要
北京时间 6 月 3 日凌晨，微软 Build 2026 开发者大会在西雅图启幕。微软 AI 负责人苏莱曼一口气发布 7 款自研 MAI 模型，覆盖推理、图像、语音、转录、代码五大方向。旗舰推理模型 MAI-Thinking-1 拥有 35B 活跃参数与 128K 上下文窗口，最核心的差异化在于"无蒸馏训练"——未使用其他模型输出进行训练，在 SWE Bench Pro 软件工程基准测试中得分 51.2%，与 Claude Sonnet 4.6 持平；MAI-Code-1-Flash 拥有 50 亿参数，剑指 GitHub 生态；MAI-Image 2.5 表现据称已超越 Nano Banana Pro；MAI-Transcribe-1.5 实现高精度语音转文字。同步亮相的还有 Surface RTX Spark Dev Box：1 petaflop AI 算力、20 核 CPU、128GB 统一内存，热设计功耗仅 100W。微软还在大会上推出 Scout 智能体与系统级安全沙箱 MXC，并联合英伟达推出统一技术栈，覆盖从 Windows 设备到云端再到底的全场景 Agentic AI 部署。
At Microsoft Build 2026 held in Seattle in the early hours of June 3 Beijing time, Microsoft AI chief Suleyman unveiled seven self-developed MAI models in a single launch, covering reasoning, image, voice, transcription, and code capabilities. The flagship reasoning model MAI-Thinking-1 features 35B active parameters and a 128K context window, with its core differentiator being “no distillation training”—trained without outputs from other models—and scoring 51.2% on the SWE Bench Pro software engineering benchmark, on par with Claude Sonnet 4.6. MAI-Code-1-Flash comes with 5 billion parameters, targeting the GitHub ecosystem. MAI-Image 2.5 reportedly outperforms Nano Banana Pro. MAI-Transcribe-1.5 delivers high-precision speech-to-text. Also unveiled was the Surface RTX Spark Dev Box with 1 petaflop of AI computing power, 20-core CPU, 128GB unified memory, and only 100W TDP. Microsoft additionally introduced the Scout Agent and system-level security sandbox MXC, while partnering with NVIDIA on a unified technology stack covering Agentic AI deployment from Windows devices to cloud and edge.
来源 / Sources
- The Verge - Microsoft Build 2026 Coverage
- 微软 Build 2026 发布 7 款自研 MAI 模型 - 界面新闻
- 一文读懂微软 Build 2026 开发者大会 - 腾讯新闻
二、博通 2026 财年 AI 芯片销售额预测达 560 亿美元，已向 OpenAI 交付定制芯片
事件概要
财联社 6 月 4 日消息，博通（Broadcom）公布 2026 财年 AI 芯片业务指引，预计销售额将达 560 亿美元（约合人民币 4000 亿元）。博通确认已向 OpenAI 交付定制 AI 芯片，并将于 2027 年完成 1.3 吉瓦（GW）算力基础设施部署。博通 CEO Hock Tan 在投资者电话会上宣布，与阿波罗、黑石及其他战略投资者共同创建 “AI XPV” 平台，目标是到 2028 年在全球部署超过 20 吉瓦的计算能力。这一规模相当于目前全球大型数据中心总能耗的近 5 倍，是迄今为止科技行业最激进的算力扩张计划之一。在业绩驱动下，博通股价近期持续走强，AI 业务收入占公司总营收比重已升至约 45%。
Broadcom announced on June 4 its fiscal year 2026 AI chip business guidance, projecting sales of $56 billion (approximately RMB 400 billion). The company confirmed delivery of custom AI chips to OpenAI and plans to complete deployment of 1.3 gigawatts of computing infrastructure by 2027. CEO Hock Tan announced on the investor call that Broadcom is establishing an “AI XPV” platform with Apollo, Blackstone, and other strategic investors, targeting deployment of more than 20 gigawatts of computing capacity globally by 2028. This scale is nearly 5 times the current total energy consumption of the world’s large data centers, representing one of the most aggressive computing capacity expansion plans the tech industry has ever seen. Driven by this performance, Broadcom’s stock has continued to strengthen recently, with AI business revenue rising to approximately 45% of the company’s total revenue.
来源 / Sources
- Broadcom Investor Relations
- 博通 2026 财年 AI 芯片销售额预测达 560 亿美元 - 财联社
- 新浪财经 - 博通 560 亿美元 AI 芯片业务
三、比亚迪"尧舜禹"人形机器人项目首度曝光，2026 年内部部署 2 万台
事件概要
6 月 3 日晚间，比亚迪集团执行副总裁李柯在投资者平台"股东星球"《有请》栏目中首次官方证实自研人形机器人项目，代号"尧舜禹"。该项目由 4000 人专项研发团队推进，已迭代至第七代原型机，2026 年计划内部部署 2 万台，未来通过比亚迪全球 4S 经销商网络向消费者销售。李柯表示，机器人竞争的核心在于制造能力、软件能力与硬件能力的综合实力，而汽车相关的 AI 技术与机器人技术具有同源性，刀片电池、伺服电机、域控制器等核心部件与汽车通用率超过 60%，实现"技术复用、成本共担"。比亚迪还与香港科技大学共建具身智能实验室，并投资智元机器人等产业链企业，构建"自研+生态"协同格局。比亚迪采取"工业先行、消费跟进"路径：现阶段重点布局工业机器人（比亚迪自身即最大客户），技术成熟后再向家庭服务场景延伸，并计划开放平台支持自研与合作模式并行。
On the evening of June 3, BYD Executive Vice President Li Ke officially confirmed for the first time on the investor platform “Shareholder Planet” the company’s self-developed humanoid robot program codenamed “Yao Shun Yu.” The project is being advanced by a 4,000-person dedicated R&D team, with the seventh-generation prototype already completed, planning internal deployment of 20,000 units in 2026, and will be sold to consumers through BYD’s global 4S dealer network in the future. Li Ke stated that the core of robot competition lies in the comprehensive strength of manufacturing, software, and hardware capabilities, while automotive AI technology and robot technology share the same origin. Core components such as blade batteries, servo motors, and domain controllers share over 60% commonality with automobiles, achieving “technology reuse and cost sharing.” BYD has also co-established an embodied AI laboratory with HKUST and invested in companies like AgiBot to build a “self-development + ecosystem” collaborative landscape. BYD adopts an “industry-first, consumer-follow” path: focusing on industrial robots at the current stage (with BYD itself as the largest customer), then extending to home service scenarios as the technology matures, with plans to open the platform to support both self-development and partnership models in parallel.
来源 / Sources
- 第一财经 - 比亚迪确认自研人形机器人
- 东方财富 - 李柯透露比亚迪正在开发人形机器人
- 腾讯科技 - 比亚迪人形机器人战略
四、千寻智能 3 个月累计融资近 50 亿元，Spirit v1.6 击败英伟达登顶 RoboArena
事件概要
6 月 3 日，具身智能公司千寻智能（Spirit AI）宣布完成 15 亿元 A+ 轮融资。本轮由一线美元基金、大型产业投资方及国家级国资基金联合注资，老股东持续超额加码。自 2024 年 1 月成立以来，千寻智能 3 个月内密集完成 4 轮融资，累计金额近 50 亿元（2 月连续两轮近 20 亿元、4 月 A 轮 10 亿元、本轮 A+ 15 亿元），刷新国内具身智能赛道融资速度纪录。本轮资金将聚焦三大方向：新一代具身智能基座模型迭代、全球化真实数据基础设施体系、多行业场景规模化商业落地。千寻自研的具身基座模型 Spirit v1.6 同日在被称为北美具身智能"奥林匹克"的 RoboArena 榜单登顶，力压英伟达 Cosmos3 与 Physical Intelligence Pi0.5，成为首个在该榜单夺冠的中国具身模型。RoboArena 由加州大学伯克利、斯坦福与英伟达联合发起，被视为具身智能领域"世界级权威主榜单"。商业化方面，千寻已与博世、京东、宁德时代等头部企业达成战略合作，其人形机器人 Moz 已批量部署于宁德时代中州基地电池 PACK 产线，并进入京东 MALL 零售场景。
On June 3, embodied AI company Qianxun Intelligence (Spirit AI) announced the completion of a 1.5 billion yuan Series A+ round. The round was jointly funded by top-tier USD funds, large industrial investors, and national-level state-owned capital funds, with existing shareholders continuing to over-subscribe. Since its founding in January 2024, Qianxun has intensively completed 4 funding rounds within 3 months, accumulating nearly 5 billion yuan (2 consecutive rounds of nearly 2 billion yuan in February, 1 billion yuan Series A in April, and 1.5 billion yuan in the current A+ round), setting a new record for financing speed in China’s embodied AI track. This round’s funds will focus on three directions: iteration of a new generation of embodied AI base models, a global real-world data infrastructure system, and scaled commercial deployment in multi-industry scenarios. On the same day, Qianxun’s self-developed embodied base model Spirit v1.6 topped the RoboArena leaderboard, dubbed the “Olympics” of North American embodied AI, outperforming NVIDIA Cosmos 3 and Physical Intelligence Pi0.5, becoming the first Chinese embodied model to win the championship. RoboArena is jointly initiated by UC Berkeley, Stanford, and NVIDIA, and is considered a “world-class authoritative main leaderboard” in the embodied AI field. In terms of commercialization, Qianxun has reached strategic cooperation with leading companies such as Bosch, JD.com, and CATL, with its humanoid robot Moz already batch-deployed in CATL’s Zhongzhou base battery PACK production line and entering JD MALL retail scenarios.
来源 / Sources
- 上证报 - 千寻智能 3 个月近 50 亿元融资
- 新浪科技 - 千寻打破具身融资纪录
- 亿邦动力 - 千寻智能 A+ 轮融资与 RoboArena 登顶
五、阿里双线推进 Agent 生态：千问向第三方开放 + QoderWork CN 推 AI 生产力计划
事件概要
6 月 3 日，阿里在 Agent 生态上连落两子。第一子：千问 APP 宣布向第三方 Agent、Skill 全面开放，所有企业均可在千问运营自己的品牌 Agent。瑞幸咖啡、肯德基、蜜雪冰城、东方航空成为首批测试伙伴——例如东航 Agent 可智能推荐行程方案、瑞幸 Agent 可主动建议"中午排队时间长、提前半小时点单"。Agent 具备记忆与主动规划能力，可在行程提醒、权益到期、复购推荐等场景主动服务。配套基础设施方面，5 月 20 日阿里云峰会发布的"千问云"（qianwenai.com）已集成 Qwen、GLM、Kimi、DeepSeek 等 150 余款模型 API，并将模型选型、认证配置、用量查询等核心链路封装为标准化 Skills 与 CLI 工具，全部在 GitHub 开源。第二子：阿里桌面智能体工作台 QoderWork CN 推出 AI 生产力计划，即日起免费发放百亿积分，用户下载即可领取 1 个月 PRO 会员（2000 积分），高校师生认证后额外获 4000 积分。QoderWork CN 支持 Qwen3.7-Max、Qwen3.7-Plus、GLM-5.1、DeepSeek-V4-Pro、Kimi-K2.6 等顶级模型，并上线设计、幻灯片、写作三大垂直工作台。
On June 3, Alibaba made two moves on the Agent ecosystem.First, Qianwen App announced full opening to third-party Agents and Skills, allowing all enterprises to operate their own branded Agents on Qianwen. Luckin Coffee, KFC, Mixue Bingcheng, and China Eastern Airlines became the first batch of test partners—for example, the China Eastern Airlines Agent can intelligently recommend travel plans, while the Luckin Agent can proactively advise “lunchtime queues are long, order half an hour in advance.” Agents have memory and proactive planning capabilities, enabling proactive service in scenarios such as travel reminders, expiring benefits, and repurchase recommendations. Regarding supporting infrastructure, “Qianwen Cloud” (qianwenai.com) released at the May 20 Alibaba Cloud Summit has integrated more than 150 model APIs including Qwen, GLM, Kimi, and DeepSeek, and has encapsulated core links such as model selection, authentication configuration, and usage queries into standardized Skills and CLI tools, all open-sourced on GitHub.Second, Alibaba’s desktop intelligent agent workbench QoderWork CN launched the AI Productivity Plan, beginning today with free distribution of 10 billion points—users who download can claim 1 month of PRO membership (2,000 points), and university teachers and students can receive an additional 4,000 points after authentication. QoderWork CN supports top models including Qwen3.7-Max, Qwen3.7-Plus, GLM-5.1, DeepSeek-V4-Pro, and Kimi-K2.6, with three vertical workbenches launched for design, slides, and writing.
来源 / Sources
- IT 之家 - 千问向第三方 Agent 全面开放
- 凤凰科技 - 千问第三方 Agent 首批接入
- 新浪科技 - 阿里 QoderWork CN AI 生产力计划
六、AI 三巨头 IPO 超级周期：合计融资近 2000 亿美元，联合国报告警示 AI 能耗压力
事件概要
全球 AI 行业正在经历有史以来最密集的 IPO 浪潮。SpaceX 已于 5 月 20 日正式向 SEC 提交 S-1 文件，目标 6 月 12 日以股票代码 “SPCX” 登陆纳斯达克，估值 1.75 万亿美元，募资 400-800 亿美元；Anthropic 于 6 月 1 日秘密递交 IPO 申请材料，最新估值 9650 亿美元（5 月 H 轮融资后），高于 OpenAI 的 8520 亿美元；OpenAI 据报道将在未来数周内秘密提交 IPO 申请。三家合计预计募资约 2000 亿美元，2026 年美股 IPO 募资总额将超过 2021 年 1600 亿美元的历史峰值。然而硬币的另一面，联合国大学水、环境与健康研究所 6 月 3 日发布《人工智能能耗的环境成本：碳、水和土地足迹》报告，警告到 2030 年全球数据中心用电需求将达 945 太瓦时（接近 2025 年的 2 倍），相关用水量将相当于 13 亿人一年的基本生活用水需求，土地占用面积将超过 14500 平方公里。报告呼吁建立 AI 算力的"碳-水-土地"三维核算体系。
The global AI industry is experiencing the most intensive IPO wave in history. SpaceX officially filed its S-1 document with the SEC on May 20, targeting a June 12 NASDAQ listing under the ticker “SPCX,” with a $1.75 trillion valuation and fundraising of $40-80 billion. Anthropic confidentially filed its IPO application materials on June 1, with the latest valuation at $965 billion (after the May Series H financing), higher than OpenAI’s $852 billion. OpenAI is reportedly preparing to confidentially file its IPO application in the coming weeks. The three combined are expected to raise approximately $200 billion, with the 2026 US IPO fundraising total set to exceed the historical peak of $160 billion in 2021. On the flip side, the United Nations University Institute for Water, Environment and Health released a report titled “The Environmental Cost of AI Energy Consumption: Carbon, Water, and Land Footprint” on June 3, warning that global data center electricity demand will reach 945 TWh by 2030 (nearly double 2025 levels), with related water consumption equivalent to the basic living water needs of 1.3 billion people for a year, and land occupation exceeding 14,500 square kilometers. The report calls for establishing a “carbon-water-land” three-dimensional accounting system for AI computing power.
来源 / Sources
- 央视财经 - Anthropic 秘密申请 IPO
- 凤凰财经 - AI 三巨头 IPO 浪潮
- 中国汽车报 - 联合国 AI 能耗报告
- 新华网 - 联合国大学 AI 环境报告
行业数据速览
| 指标 | 数据 | 备注 |
|---|---|---|
| 微软 MAI-Thinking-1 SWE Bench Pro 得分 | 51.2% | 与 Claude Sonnet 4.6 持平 |
| 微软 MAI-Thinking-1 上下文窗口 | 128K token | 35B 活跃参数，无蒸馏训练 |
| Surface RTX Spark Dev Box AI 算力 | 1 petaflop | 100W TDP，20 核 CPU，128GB 统一内存 |
| 博通 2026 财年 AI 芯片销售额指引 | 560 亿美元 | 已向 OpenAI 交付定制芯片 |
| 博通 AI XPV 平台 2028 年算力目标 | 20+ GW | 与阿波罗、黑石联合创建 |
| 比亚迪"尧舜禹"研发团队规模 | 4000 人 | 第七代原型机已出炉 |
| 比亚迪人形机器人 2026 年内部部署目标 | 2 万台 | 4S 经销商网络向消费者销售 |
| 汽车与机器人核心部件通用率 | 60%+ | 刀片电池/伺服电机/域控制器 |
| 千寻智能 3 个月累计融资 | 近 50 亿元 | 4 轮密集融资 |
| 千寻智能本轮（A+）融资金额 | 15 亿元 | 一线美元基金+产业+国资 |
| 联合国 2030 年数据中心用电预测 | 945 TWh | 较 2025 年接近翻番 |
| Anthropic 估值 | 9650 亿美元 | 已秘密递交 IPO |
| OpenAI 估值 | 8520 亿美元 | 数周内秘密递交 IPO |
| SpaceX 估值 | 1.75 万亿美元 | 6/12 登陆纳斯达克 |
| AI 三巨头合计预计募资 | ~2000 亿美元 | 创历史新高 |
媒体人解读
今天的资讯呈现出 AI 行业**"技术-资本-能耗"三线并发**的鲜明特征，每一条线都指向拐点级别的产业事件。
技术线：微软 Build 2026 用 7 款自研模型宣告"Agent 优先"时代正式到来。MAI-Thinking-1 在 SWE Bench Pro 上与 Claude Sonnet 4.6 持平的意义，远超一个技术参数——它意味着微软第一次具备了"独立推理模型"的能力，而不再完全依赖 OpenAI 的转售与蒸馏。"无蒸馏训练"是这场发布会最具行业风向标意义的设计选择，OpenAI、Anthropic 的护城河正在被快速侵蚀。Surface RTX Spark Dev Box 用 100W 功耗实现 1 petaflop 算力，则给"端云一体"Agent 部署提供了硬件基础。
资本线：Anthropic 9650 亿美元估值+千寻 50 亿元融资+博通 560 亿美元 AI 收入指引，资本对 AI 产业链的押注已经从"模型层"向"基础设施层"和"具身智能层"双向延伸。千寻 Spirit v1.6 在英伟达自家发起的 RoboArena 上击败 Cosmos3 是个关键信号——"客场夺冠"的故事性远超技术参数本身，它意味着中国具身智能公司第一次在"世界级权威榜单"上拿到了规则制定者的认可。
能耗线：联合国报告给狂欢的资本市场踩了一脚刹车。945 太瓦时的用电量、13 亿人年用水当量、14500 平方公里土地占用——这些数字不是抽象的环境议题，而是数据中心选址、电力供应审批、水资源分配的现实约束。下半年"AI 算力可持续性"将从 ESG 报告里的修辞变成各国能源政策的硬约束。远景科技在 5 月提出的"AI 电力系统"概念，会从产业提案变成国家战略。
战略含义：比亚迪"尧舜禹"项目 60% 以上的零部件与汽车通用，意味着汽车产业正在变成"具身智能的最大母生态"。这不仅是车企多元化，更预示着未来 5 年人形机器人量产的最大推力，将来自"汽车级供应链"——这与传统机器人公司（波士顿动力、特斯拉 Optimus）的路径完全相反，值得所有具身智能创业者重新评估竞争格局。
Today’s news presents a striking three-line concurrent feature of “technology - capital - energy consumption” in the AI industry, with each line pointing to a turning-point-level industry event.
Technology Line: Microsoft Build 2026 used 7 self-developed models to formally announce the arrival of the “Agent-first” era. MAI-Thinking-1’s tie with Claude Sonnet 4.6 on SWE Bench Pro carries significance far beyond a single technical metric—it means Microsoft has, for the first time, “independent reasoning model” capability, no longer fully dependent on OpenAI’s reselling and distillation. “No distillation training” is the most industry-beacon-meaningful design choice of this launch. OpenAI and Anthropic’s moats are being eroded quickly. The Surface RTX Spark Dev Box achieves 1 petaflop of computing power at 100W TDP, providing the hardware foundation for “end-cloud integrated” Agent deployment.
Capital Line: Anthropic’s $965 billion valuation + Qianxun’s 5 billion yuan financing + Broadcom’s $56 billion AI revenue guidance show that capital bets on the AI industry chain have extended from the “model layer” to the “infrastructure layer” and the “embodied intelligence layer” in both directions. Qianxun’s Spirit v1.6 defeating Cosmos 3 on NVIDIA’s own RoboArena is a key signal—the storytelling of “winning away from home” goes far beyond technical parameters themselves, meaning that a Chinese embodied AI company has, for the first time, received recognition from the rule maker on a “world-class authoritative leaderboard.”
Energy Consumption Line: The UN report has hit the brakes on the reveling capital market. 945 TWh of electricity, the equivalent of 1.3 billion people-years of water, 14,500 square kilometers of land occupation—these numbers are not abstract environmental issues, but real constraints on data center site selection, power supply approval, and water resource allocation. In the second half of the year, “AI computing power sustainability” will shift from rhetoric in ESG reports to hard constraints in national energy policies. The “AI Power System” concept proposed by Envision Group in May will evolve from industrial proposal to national strategy.
Strategic Implications: BYD’s “Yao Shun Yu” project with over 60% component commonality with automobiles means the automotive industry is becoming the “largest mother ecosystem of embodied intelligence.” This is not just automaker diversification, but signals that the biggest push for humanoid robot mass production in the next 5 years will come from “automotive-grade supply chains”—a path completely opposite to traditional robot companies (Boston Dynamics, Tesla Optimus), worth re-evaluating the competitive landscape for all embodied intelligence entrepreneurs.
参考来源
英文来源
- The Verge - Microsoft Build 2026 Coverage
- Broadcom Investor Relations
- TechCrunch - Latest AI News
- Reuters Technology
- Bloomberg - Anthropic Valuation Report
中文来源
- 界面新闻 - 微软 Build 2026 七款 MAI 模型
- 财联社 - 博通 2026 财年 AI 芯片 560 亿美元
- 新浪财经 - 博通 AI 芯片业务
- 第一财经 - 比亚迪确认自研人形机器人
- 东方财富 - 李柯透露比亚迪人形机器人
- 腾讯科技 - 比亚迪人形机器人战略
- 上证报 - 千寻智能 3 个月近 50 亿元融资
- 新浪科技 - 千寻打破具身融资纪录
- 亿邦动力 - 千寻智能 RoboArena 登顶
- IT 之家 - 千问向第三方 Agent 全面开放
- 凤凰科技 - 千问第三方 Agent 首批接入
- 新浪科技 - 阿里 QoderWork CN AI 生产力计划
- 央视财经 - Anthropic 秘密申请 IPO
- 凤凰财经 - AI 三巨头 IPO 浪潮
- 中国汽车报 - 联合国 AI 能耗报告
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4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
