美东时间1月28日，特斯拉2025年第四季度财报电话会议上，CEO埃隆·马斯克将核心议题锁定 芯片自主化，明确芯片（尤其内存芯片）是未来增长的核心 瓶颈，官宣兆级晶圆厂（TerraFab）建设、自研AI5/AI6芯片等重磅规划 ，同步披露2026年大额资本支出与配套产能布局，彰显其摆脱供应链束缚、支撑 AI与机器人业务的战略决心。 亮眼财报为芯片战略提供坚实支撑：特斯拉Q4营收249亿美元、调整后EPS 0.50美元，均超市场预期；汽车业务毛利率升至17.9%，现金及投资超440亿美 元，充足的资金储备成为芯片自主化与产能扩张的重要保障，也印证了其加大核心领域投入的可行性。 马斯克在会议中直言，随着Cybercab自动驾驶出租车（4月量产）、Optimus人形机器人（目标年产能100万台）推进，特斯拉对AI芯片、内存芯片的需求 呈爆发式增长，仅Optimus年产能达100万台时，年芯片需求就将达60亿美元。而当前全球高端芯片（含内存芯片）产能高度依赖台积电、三星、美光等外 部厂商，不仅存在产能分配优先级不足的问题，更面临地缘政治引发的供应链中断风险，现有供应链已无法匹配其长期发展目标，内存芯片的供应短缺更 是 ...

导读：马斯克再一次游走于风险边缘——一边是特斯拉日渐萎靡的汽车业务，一边则是他压上特斯拉未 来的AI和机器人业务。 （文 / 观察者网 周盛明 编辑 / 高莘） 游走在风险边缘，压上身家和声誉，忍受现实的压力——马斯克似乎早已习惯在"逆风局"中找到实现目 标的快感。 作为特斯拉、SpaceX、xAI等多家公司的首席执行官，马斯克的创业过程展现出极端的风险偏好和近乎 于偏执的"赌徒"性格。 埃隆·马斯克 美联社 随着特斯拉2025年财报的公布，马斯克的这一性格再次显露无疑——虽然特斯拉的全年营收出现历史首 次下滑、净利润同比近乎腰斩，但马斯克毅然决然的宣布将Model S/X停产，并集中大量资金投资尚未 商业闭环的AI和机器人领域。 虽然交出了不甚如意的财报，但特斯拉的股价波动并不明显。如今，特斯拉的市值依旧高达1.35万亿美 元（约合人民币9.38万亿元），远远超过同类汽车公司——背后真正作为支撑的，正是马斯克给予投资 者的"对未来的希望"。 但是希望并不能给予商业化的回报。在"赌徒马斯克"的带领下，特斯拉再一次游走于风险边缘——一边 是特斯拉日渐萎靡的汽车业务，一边则是他压上特斯拉未来的AI和机器人业务。 ...

马斯克表示，特斯拉最早推出的两款高端车型将在下一季度结束生产，获得"光荣退役"。此举意在腾出加州弗里蒙特工厂的产能，转而用于建立一条年产能达 100万台的人形机器人Optimus生产线。 马斯克直言，公司正在迈向一个"基于自动驾驶的未来"，并强调现在是预订这两款绝版车型的最后时机。 除了产品线的重大调整，特斯拉进一步深化了与马斯克旗下另一家公司xAI的绑定。除了确认20亿美元的投资外，双方还达成了一项"框架协议"，旨在探索更深 层次的合作。特斯拉表示，这一合作将有助于提升其自动驾驶技术、车载娱乐体验以及制造机器人的研发能力。 与此同时，公司披露2026年将是"巨大的投资年"，资本支出（Capex）预计将超过200亿美元，主要用于算力基础设施、电池供应链及AI芯片的建设。 特斯拉正在加速剥离传统汽车制造商的身份，全力押注自动驾驶与人形机器人业务。 特斯拉周三盘后公布的第四季度财报电话会议透露出激进的战略转型信号，马斯克宣布将 永久停产旗舰车型Model S和Model X ，并确认了向人工智能（AI） 公司xAI投资20亿美元的计划。 数据显示，去年这两款车型仅占特斯拉159万辆总交付量的3%。马斯克表示，尽管 ...

Core Viewpoint - Tesla's financial report indicates a significant decline in total deliveries and revenue for 2025, marking the first revenue drop since its IPO, with net profit nearly halved. The company plans to gradually cease production of Model S and Model X starting next quarter [1][3][16]. Financial Performance - In Q4 2025, Tesla's revenue was $24.9 billion, a 3% year-over-year decline. Revenue from automotive-related businesses was $17.693 billion, down 11% year-over-year, with automotive sales revenue at $16.75 billion, also down 10% [10][12]. - The total annual sales for Tesla in 2025 reached 1.636 million units, a decrease of 8.6% year-over-year, marking two consecutive years of decline. However, deliveries in the Asia-Pacific region set a historical record [5][8]. - Tesla's GAAP net profit for 2025 was $3.794 billion, a 46% decrease from $7.1 billion in 2024, while non-GAAP net profit fell 26% to $5.858 billion [16][18]. Strategic Shift - Tesla is shifting its focus towards AI and robotics, with plans to transform production lines for Model S and Model X into facilities for the Optimus robot, aiming for an annual production target of 1 million units [23][25]. - The company is also set to begin mass production of the long-awaited electric semi-truck, Semi, in the first half of the year, and the new Roadster is expected to debut in April [27][29]. AI and Robotics Development - Tesla's FSD (Full Self-Driving) software has seen a 38% increase in paid users, reaching 1.1 million by the end of 2025. The company is actively expanding its Robotaxi fleet, with over 500 units deployed in select U.S. cities [32][36]. - The third-generation Optimus robot is expected to be released within months, featuring advanced capabilities for learning and task execution [40][42]. Financial Health - As of the end of 2025, Tesla's cash and investment balance reached a record $44.059 billion, a 21% increase from the previous year, supported by a positive free cash flow of $6.26 billion [20].

受此消息影响，周四美股夜盘，特斯拉股价上涨2.4%，报441.83美元。 营收承压但超预期 财报显示，2025年第四季度，特斯拉总营收达249.01亿美元，较去年同期的257.07亿美元下滑3%，略超 预期的247.9亿美元。 盈利端承压明显，四季度营业利润为14.09亿美元，同比下滑11%；净利润8.4亿美元，同比暴跌61%。 非GAAP口径下调整后的每股收益（EPS）为0.50美元，同比下降17%，显著高于预期的0.45美元。 年营收首降、Q4净利暴跌61%，特斯拉交出一份不理想成绩单。 由于核心汽车业务承压，特斯拉全年营收同比下滑3%，创下历史首次年度营收降幅。 不过，公司四季度营收与调整后每股收益均超出市场预期。 | Printered | | | | | | | | --- | --- | --- | --- | --- | --- | --- | | ($ in millions, except percentages and per share data) | Q4-2024 | Q1-2025 | Q2-2025 | Q3-2025 | Q4-2025 | YoY | | Total auto ...

据业内人士26日透露，三星电子晶圆代工部门有望在明年扭亏为盈。该部门自2022年以来已累计 亏损数万亿韩元。KB证券研究主管金东元表示："受特斯拉人工智能芯片供应量增加的推动，三星 晶 圆 代 工 部 门 的 产 能 利 用 率 有 望 提 高 ， 从 而 使 其 业 绩 从 去 年 的 7 万 亿 韩 元 亏 损 转 为 明 年 的 盈 利。"一位熟悉三星晶圆代工部门情况的人士也表示："我们正在加快订单进度，目标是明年实现盈 利。除了3纳米及以下先进工艺的良率趋于稳定外，高利润的4-8纳米工艺节点的产能利用率也已达 到峰值。" 三星晶圆代工业务的盈利预期与泰勒晶圆厂的产能提升时间密切相关。三星电子正在德克萨斯州泰 勒市建设一座大型晶圆代工生产基地，计划到2030年投资超过370亿美元（约合54万亿韩元），并 计划于今年下半年开始量产。目前，三星晶圆代工业务已进入最后阶段，包括派遣工程师和启动生 产线，以期在今年上半年进行试生产。三星晶圆代工的目标是赢得美国大型科技公司的订单，并实 现芯片的量产。 如果三星晶圆代工通过泰勒晶圆厂量产特斯拉人工智能芯片，预计盈利能力将有所提升。三星晶圆 代工计划明年量产AI5和 ...

2025年12月马斯克表示特斯拉在奥斯汀正在进行车内无安全员的Robotaxi测试，但尚未面向公众运营。 2026年1月22日特斯拉车内无安全员的Robotaxi车队开始在奥斯汀面向公众服务，目前仅从少数车辆开 始，后续数量将逐步增加，马斯克在达沃斯论坛中也表示2026年底Robotaxi服务在美国将得到非常广泛 的应用。特斯拉1月28日举办的四季度业绩会同样值得期待，或将披露最新Robotaxi运营情况及未来规 划。 Cybercab进展持续推进，马斯克披露后续硬件规划 截至2026年1月Cybercab已在美国五个州进行测试，计划4月生产，马斯克表示早期生产速度可能较慢但 最终会非常快，且预计大规模运营时每英里价格可能低于0.20美元。随着Cybercab量产在即，马斯克也 在不断披露硬件的更新情况：2026年1月17日马斯克表示AI5芯片设计几乎完成，AI6已进入早期开发阶 段，AI7/8/9设计周期目标为九个月。此外，马斯克表示AI4本身将实现远超人类的安全水平，AI4就能 实现无监督，无需额外硬件；AI5将使车辆几乎完美；AI6将用于Optimus和数据中心；AI7/Dojo3将采用 空间AI计算。 ...

据特斯拉首席执行官埃隆·马斯克最新透露，该公司将重启此前被搁置的超级计算机项目Dojo 3。这距离 2025年8月该项目被全面叫停已过去五个月，标志着特斯拉芯片战略的重大转向。 而且这一次，Dojo3的核心使命不再局限于地球上的自动驾驶模型训练，而是正式进军"太空人工智能 （AI）计算"。 Dojo3"复活" 五个月前，特斯拉突然叫停Dojo 3项目，解散核心团队，项目负责人彼得・班农离职，约20名核心成员 跳槽成立新公司DensityAI。 当时马斯克公开解释，分散资源同时开发两种截然不同的AI芯片设计缺乏合理性，决定将全部精力集 中于AI5、AI6及后续芯片，这类芯片既能满足高效推理需求，也能胜任核心训练任务。 他还称，特斯拉计划增加对英伟达和其他合作伙伴（如AMD）在计算方面的依赖，以及对三星在芯片 制造方面的依赖，而不是继续开发自己的定制芯片。因此，马斯克的最新言论表明，这一战略可能再次 发生了转变。 马斯克在最新帖子中指出，基于AI5芯片设计进展顺利的现状，公司决定重启Dojo。他还发布"招聘启 事"，邀请对"研发世界上产量最高芯片"感兴趣的工程师加入，并要求应征者用三个要点简述其解决过 的关键技术 ...

Core Viewpoint - Tesla is reviving its Dojo 3 supercomputer project, which was previously halted, indicating a significant shift in its chip strategy towards space AI computing [3][4][10]. Group 1: Dojo 3 Project Overview - Dojo 3 was initially designed as a supercomputer for machine learning training, with a focus on architecture restructuring and cost optimization [7]. - The project was paused five months ago due to a strategic decision to concentrate resources on AI5 and AI6 chips, moving away from developing custom chips [5][8]. - The revival of Dojo 3 is based on the successful progress of the AI5 chip design, with Tesla actively recruiting engineers for this initiative [5][8]. Group 2: Space AI Ambitions - The new mission for Dojo 3 includes advancing into "space artificial intelligence computing," aligning with Musk's vision of deploying AI computing centers in space [4][10]. - Musk believes that operating large-scale AI systems in orbit will be more cost-effective than on Earth within the next four to five years, leveraging solar energy and cooling technologies [11]. - This vision connects Tesla's various business interests, potentially positioning the company as a major beneficiary if successful [11]. Group 3: Challenges Ahead - Achieving the goal of space-based AI data centers faces significant challenges, including orbital debris, regulatory approvals, and international space policies [12]. - Cooling high-power computing devices in a vacuum presents practical difficulties, despite theoretical advantages [12][13]. - Constructing large AI data centers in geostationary orbit (GEO) is feasible but requires extensive infrastructure, which poses logistical and financial challenges [13][14].

Core Insights - Musk is aggressively pushing Tesla's AI chip development, aiming to disrupt the semiconductor market dominated by companies like NVIDIA and AMD with a series of new chips: AI5, AI6, and AI7 [2][3] - The AI5 chip, designed for both vehicles and robots, is expected to outperform its predecessor by 50 times and is a key component in Tesla's strategy to unify its hardware and software across different applications [4][5][6] - The AI6 chip aims to revolutionize the AI infrastructure by combining training and inference capabilities on a single chip, potentially transforming Tesla vehicles into distributed computing nodes [7][8] - The AI7 chip is targeted for space applications, indicating Musk's ambition to extend AI capabilities beyond Earth, particularly for SpaceX's Starship and Starlink [8] - Tesla plans to establish its own semiconductor fabrication facility, TeraFab, to gain control over its chip supply chain and reduce reliance on external suppliers like NVIDIA [13][16] Group 1: Chip Development - Musk announced the completion of the AI5 chip design, which will connect Tesla's smart vehicles and robots [3] - The AI5 chip is expected to have performance comparable to NVIDIA's Hopper architecture at a lower cost and power consumption [3][4] - The AI6 chip is designed to break the traditional separation between training and inference chips, allowing for more versatile applications [7][9] Group 2: Strategic Implications - The rapid iteration cycle of nine months for new chips is driven by the need to keep pace with the fast-evolving AI algorithms that outstrip current hardware capabilities [10][11] - Musk's vision includes a significant increase in annual AI chip demand, projecting between 100 million to 200 billion chips needed to support Tesla's growth [11][12] - The establishment of TeraFab aims to address supply chain vulnerabilities exposed during the global chip shortage, allowing Tesla to optimize production and costs [16][17] Group 3: Ecosystem and Future Vision - Tesla's AI ecosystem is designed to create a feedback loop where data collected from vehicles and robots continuously improves AI models [18][19] - The integration of AI7 with Starlink aims to provide global connectivity and computational power, enhancing the capabilities of Tesla's products in remote areas [19][20] - Musk's overarching goal is to secure control over AI computing power, positioning Tesla as a leader in the race towards artificial general intelligence (AGI) [20]