Tesla Actually Gave Us the Answers We Needed (TSLA) ⚡️

以下是Dylan Loomis视频内容的中文总结，涵盖了他提出的每一个新闻点和分析： --- **一、引言与总体情绪：** * Dylan 觉得他应该向特斯拉社区道歉，他开玩笑说，他之前的帖子可能“削弱了”（nerfed）埃隆·马斯克的公开乐观情绪，导致了更“平淡的预期”。 * 埃隆“身体不适”，这可能影响了他的语气。 * 特斯拉在自动驾驶出租车（robotaxi）的推出方面，首次在很长一段时间内有了“适当的预期”和“重新校准”。 * 财报电话会议并不令人兴奋，但提供了“新的数据和见解”。 * 特斯拉未来的增长目前取决于自动驾驶出租车。 **二、Optimus (人形机器人)：** * Optimus 今年可能进入小规模生产，但早期产品将进入“Optimus 学院”。 * Dylan 批评分析师没有就 Optimus 提出具体问题： * 机器人在部署前将在学院待多久？ * 学院需要多少台 Optimus（近/远期）？ * Optimus 首次在特斯拉工厂/学院以外的实际部署是什么？ * Optimus 机器人何时能销售给普通人类客户的最佳猜测是什么？ * Optimus“太不确定了”；埃隆“泼了冷水”，强调了漫长的生产爬坡期。 * 没有关于销售成本（COGS）、早期部署或第三方客户定价的信息。 * 市场目前对 Optimus “几乎没有给予任何估值”。 **三、特斯拉股票与 Dylan 的展望：** * 迪伦此前停止讨论短期/近期股价预期，因为他不喜欢人们根据他的“猜测”（考虑到他前财务顾问的背景）做出财务决策。 * 现在，他觉得他应该发声了，因为他认为，由于估值原因，短期内“下行潜力”远大于上行潜力。 * **2027 财年远期市盈率（基于特斯拉股价约 313 美元）：** * 特斯拉：140 倍 * 苹果：35 倍 * 亚马逊：23 倍 * 谷歌：22 倍 * 微软：20 倍 * 英伟达：17 倍 * Meta：17 倍 * 他认为市盈率（PE）对于特斯拉/七巨头来说并非理想指标，但 2027 年的市盈率是最佳指标，因为它在不陷入未来不确定性的前提下，预估了自动驾驶出租车的增长。 * 特斯拉的高估值意味着市场“显然仍在很大程度上消化了这种自动驾驶出租车的增长”。 * 埃隆/团队向下调整预期导致“下行潜力”。 * 许多投资者没有意识到，大部分预期的规模扩张和性能提升“已经计入股价”。 * 特斯拉股票“在每股 305 美元时比 405 美元时更具吸引力”。 * 如果人们质疑他们的信念，他们可能没有理解这个故事或估值。 * 未来 3-6 个月的股价表现“可能很痛苦，也可能很无聊”。 * **长期乐观：** 潜在的 SpaceX 收购，特斯拉每年投入的资金是以前的三倍（作为一家资本效率高的公司）预示着高投资回报率，尽管回报“还需要几年”。 * 批评那些总是根据催化剂预测“特斯拉今年将飙升至五、六、七百美元”的分析师，因为市场已经知道这些信息了。 * 迪伦“绝不会提供财务建议”，但他认为“风险回报概况”在目前“更具吸引力”。 * 他“完全可以预见到特斯拉股价触及 280、260 美元”之后再上涨（这不是预测，也不是在玩期权）。 * 他指出，“直冲月球”（to the moon）的账号获得赞扬，而批评性账号“遭到批评”，但“到目前为止，他们的判断错少对多”。 * 迪伦“正在这里买入”，“没有卖出过一股”，“也没有出售计划”。 * 他对特斯拉未来 3-5 年的看法“一如既往地乐观”。 * 承认过去一年希望有更好的沟通，但埃隆最近放低姿态的态度允许人们翻过这一页。 **四、自动驾驶出租车推出数据与分析：** 1. **累计付费自动驾驶出租车里程图表：** * 这张图表*不*仅仅是无人监管里程（特斯拉于 2024 年 1 月开始无人监管，但图表显示了之前的数据）。 * 2025 年的数据点（约 60 万英里）对于奥斯汀一个城市来说可能过高。 * Dylan 假设这张图表包括了湾区，特斯拉可能将其归类为自动驾驶出租车（即使是网约车），因为它们收取付费乘车费用。 * 对图表的像素分析（外推法）： * 1 月：27 万付费里程 * 2 月：34 万付费里程 * 3 月：峰值 44.5 万付费里程 * 4 月：36.5 万（减速） * 5 月：19 万 * 6 月：15.5 万 * 最近的减速可能是特斯拉“正在湾区逐渐减少有人监管的乘坐，为无人驾驶做准备”。 2. **无人监管里程（新的关键指标）：** * Ashok（特斯拉自动驾驶/人工智能负责人）宣布，特斯拉已在“两个州的六个城市行驶了超过 38 万英里的无人监管自动驾驶出租车里程”。 * 不确定：不知道这些是否仅是付费里程，还是包括了“空驶里程”。 * 埃隆暗示湾区被纳入“自动驾驶出租车范畴”。 * 埃隆表示：“每周行驶里程增长超过 10%”。 * Ashok 澄清说，这指的是“无人监管里程”，并且从今年年初开始，预计将持续“到今年底”。 * “‘赛博出租车’的生产速度将与无人监管里程数的预期增长相匹配。”这现在是*关键*指标。 * **Dylan 的计算（从 1 月开始每周 10% 增长）：** * 从每周约 3500 英里开始，到 7 月中旬达到每周约 3.8 万英里（累计约 38 万英里）。 * 假设目前有约 50 辆无人监管车辆（例如，奥斯汀 30 辆，达拉斯、休斯顿、坦帕、迈阿密、奥兰多各 20 辆）。 * 每辆无人监管汽车平均“每天行驶略超过 100 英里”。 * 如果每周 10% 的增长持续到第三季度（10 月底）：每周约 13 万英里（接近 3.5 倍增长，约 250% 增加）。 * 这将要求无人监管车队在第三季度末增长到约 173 辆（增加约 123 辆）。 * *增长率递减情景：* 车队总数达到约 115 辆（增加约 65 辆）。 * Ashok 强调“最大化每辆车的效率”，目标是将每日里程从 100 英里提升到 200-300 英里。 * Dylan 更新的预期：“到第三季度末，实际部署的车辆可能多达 200 辆”。 * 在未来几个月内，扩展到数千辆自动驾驶出租车是不可能的。 3. **Waymo 比较：** * Waymo 平均*每天*行驶约 57 万英里无人驾驶里程。 * 特斯拉（以持续增长计算）*每周*将达到 13 万英里无人监管里程。 * Dylan 称这种比较“风马牛不相及”且“大体上是愚蠢的”，但他仍提供了这个背景。 4. **“零值得关注事件”：** * Ashok 关于“38 万英里内没有发生值得关注的事件”的声明。 * Dylan 指出：“值得关注”未被定义，并且它*仅指*真正的无人监管里程。 * NHTSA SGO 数据包括有人监管和无人监管里程。 * 38 万英里是“小样本量”，集中在“有限区域”，且在“有利条件”下。这不能推广到全国数百万英里的行驶。 * 在有限条件下零事故“不意味着在无限操作领域仍会保持这样”。特斯拉的表述中没有内在矛盾。 5. **新城市/州推出：** * Ashok：预计“在新城市推出所需的时间将继续趋近于零”，目标是“将整个州作为一个整体而不是逐个城市推出”。 * Dylan“泼冷水”：这“今年不会发生”，“最早也要等到明年”。 * “趋近于零”可能是一个 5-10 年的过程。 * V15（最前沿的 FSD）正在自动驾驶出租车车队上运行。 * 在“紧密受控的环境”和充足数据下，在自动驾驶出租车上运行新的、未经测试的 V15 *更安全*，相比之下，V14 则在全球部署。 * 这意味着 V14 尚“不足以用于无人监管”。 * 分析师没有问及 V15 中“七个不同的主要改进方向”。 **五、收购：** * 第二季度 10Q 报告披露的细节：特斯拉以 19.5 亿美元（普通股/股权）收购了一家人工智能硬件公司。 * 其中 17.3 亿美元取决于服务条件和/或业绩里程碑。 * 约 2.22 亿美元分配给专利及相关技术。 * 这是一种“人才技术收购”，即收购技术、知识产权和员工，并依靠他们进行整合。 * Dylan 推测该收购是 Density AI（基于第一季度的传闻）： * 联合创始人/首席执行官加内什·文卡塔拉马南（Ganesh Venkataramanan）曾是特斯拉的员工（自动驾驶硬件高级总监，参与过 Dojo 项目）。 * 传闻：约 20 名其他特斯拉工程师（来自 Dojo 重组）加入了加内什的 Density AI。 * Density AI 专注于数据中心、汽车和传感器融合领域，用于高密度、高效推理（大型模型、长上下文工作负载）的专用 AI 加速器、硬件、全栈系统和内存架构。Dylan 认为这将是“一个很好的匹配”。 **六、FSD 订阅统计：** * 图表显示特斯拉“第二季度在北美 FSD 订阅的附带率（attach rate）显然达到了 55%”。 * Vybov 表示有 148 万个活跃的 FSD 订阅： * 45% 按月付费（约 66.6 万人）。 * 55% 一次性购买（约 81.4 万人）。 * FSD 增长速度是“特斯拉汽车车队增长速度的两倍”。 * 强调 FSD 是现有车辆基础的“利润升级”。 * **FSD 经常性收入计算：** 每年约 7.91 亿美元（148 万 * 0.45 * 100 美元/月 * 12 个月）。 * 这是“更令人鼓舞的统计数据之一”，因为附带率高于预期（10-20%）。 * 指出车辆交付并没有带来显著影响，即使是 FSD 收入“也尚未对整体财务状况产生显著影响”。 **七、财务（具体提及）：** * Dylan 表示如果需要，下周将进行更深入的分析。 * **能源业务毛利率：** 第二季度从约 39% 下降到约 20%。 * Vybov 预计未来能源业务毛利率将稳定在 20-25% 之间（此前认为会达到 30-40%）。这需要重新评估预期。 * **监管信用：** 首次出现“显著下降”，本季度降至 1.46 亿美元（此前最低记录是 3.8 亿美元）。 * CAFE 罚款的到期“可能终于在起作用了”。 **八、星链（Starlink）与 Semi 的 FSD：** * 埃隆：星链（Starlink）将在可用的情况下“适时”添加到特斯拉的其余车队中。 * 埃隆：特斯拉 Semi 的 FSD 将在“今年晚些时候或明年初”启用，但目前的重点是量产车型的 FSD。 **九、结论：** * 为视频因个人事务晚一天发布致歉。 * 计划在未来的视频中继续分析第二季度财报。 * 祝愿观众星舰飞行顺利，周末愉快。 * 感谢 Patreon 支持者。

Here's a comprehensive summary of the video, covering every single news item and analysis point made by Dylan Loomis: **I. Introduction & General Sentiment:** * Dylan owes the Tesla community an apology, jokingly suggesting his previous post may have "nerfed" Elon Musk's public optimism, leading to more "muted expectations." * Elon was "under the weather," potentially contributing to his tone. * For the first time in a long time, Tesla has "proper expectations" and "recalibration" regarding the robotaxi rollout. * The earnings call wasn't exciting, but provided "new data and insight." * Tesla's future growth currently hinges on robotaxis. **II. Optimus (Humanoid Robot):** * Optimus might enter low-scale production this year, but early units will go to "Optimus Academy." * Dylan criticizes analysts for not asking specific questions about Optimus: * Duration robots will be at the Academy before deployment. * Number of Optimide for the Academy (near-term/long-term). * First real-world deployment for Optimide outside Tesla factories/Academy. * Best guess for when an Optimus robot will be sold to a regular human customer. * Optimus is "far too uncertain"; Elon "poured cold water on it," highlighting the long production ramp. * No information on Cost of Goods Sold (COGS), early deployments, or pricing for third-party customers. * The market currently attributes "mostly nothing" to Optimus. **III. Tesla Stock & Dylan's Outlook:** * Dylan previously stopped discussing short/near-term stock expectations due to discomfort with people making financial decisions based on his "guesstimates" (given his former financial advisor background). * Now, he feels he should have spoken up, as he believes there's "far more downside potential" than upside potential in the near term due to valuations. * **Forward PE Ratios for FY2027 (based on Tesla at ~$313/share):** * Tesla: 140X * Apple: 35 * Amazon: 23 * Google: 22 * Microsoft: 20 * Nvidia: 17 * Meta: 17 * He argues PE isn't ideal for Tesla/Mag 7, but 2027 PE is the best metric as it pulls forward robotaxi growth without going too far into future uncertainty. * Tesla's high valuation means the market is "clearly still pricing in a lot of this robot taxi growth." * Downward expectation shifts from Elon/team lead to "downside potential." * Many investors don't realize much of the expected scaling and performance is "already priced in." * Tesla stock is "far more attractive at $305 a share than it was at $405." * If people are questioning their conviction, they might not have understood the story or valuations. * The next 3-6 months "might be painful, it might be boring" for stock performance. * **Long-term optimism:** Potential SpaceX acquisition, Tesla spending 3x more per year than ever (as a capital-efficient company) signals high ROI, though payoff is "a few more years." * Critiques analysts who always predict "Tesla's ready to run to five, six, $700 a share this year" based on catalysts, as the market already knows about these. * Dylan will "never give financial advice" but sees the "risk reward profile" as "much more attractive where we are now." * He "could absolutely see Tesla stock touching 280, 260" before moving up (not a prediction, not playing options). * He notes that "to the moon" accounts get praise, while critical accounts "get criticized" but "so far, they've been more right than wrong." * Dylan is "buying here," "has not sold a single share," "has no plans to." * He is "as bullish on Tesla as I've ever been" for the next 3-5 years. * Acknowledges wishing for better communications over the past year, but Elon's recent toned-down approach allows moving past it. **IV. Robotaxi Rollout Data & Analysis:** 1. **Cumulative Paid Robotaxi Miles Chart:** * This chart is *not* only unsupervised miles (Tesla started unsupervised in Jan 2024, but the chart shows data before that). * The 2025 data point (~600,000 miles) is likely too high for Austin alone. * Dylan assumes this chart includes the Bay Area, which Tesla might classify as robotaxi (even if rideshare) because they charge for paid rides. * Pixel analysis of the chart (extrapolations): * January: 270,000 paid miles * February: 340,000 paid miles * March: Peak at 445,000 paid miles * April: 365,000 (deceleration) * May: 190,000 * June: 155,000 * The recent deceleration might be Tesla "slowly toning down the supervised rides in the Bay Area to prepare for driverless." 2. **Unsupervised Miles (The New Key Metric):** * Ashok (Tesla's Head of Autopilot/AI) announced Tesla has driven "more than 380,000 miles of unsupervised robotaxi across six cities in two states." * Uncertainty: Don't know if these are paid miles only or include "deadhead miles." * Elon implies Bay Area is considered under the "robotaxi umbrella." * Elon stated: "more than 10% a week in terms of miles driven" growth rate. * Ashok clarified this refers to "unsupervised miles" and has been happening since early this year, expected to continue "through the rest of this year." * "Cyber cab production rate will be matching with the expected growth of the unsupervised mile number." This is now *the* key metric. * **Dylan's Calculations (10% weekly growth starting January):** * Starting at ~3,500 miles per week, reaching ~38,000 miles per week by mid-July (cumulative ~380,000 miles). * Assumes a current fleet of ~50 unsupervised vehicles (e.g., 30 in Austin, 20 across Dallas, Houston, Tampa, Miami, Orlando). * Each unsupervised car averages "just over 100 miles per car per day." * If 10% weekly growth is sustained through Q3 (end of October): ~130,000 miles per week (nearly 3.5x growth, ~250% increase). * This would require the unsupervised fleet to grow to ~173 vehicles by end of Q3 (adding ~123 cars). * *Decaying growth rate scenario:* Fleet hits ~115 cars total (adding ~65 cars). * Ashok emphasized "maximizing the efficiency of each vehicle," aiming to push daily mileage from 100 to 200-300 miles. * Dylan's updated expectation: "maybe up to 200 cars actually deployed" by end of Q3. * "Scaling to thousands of robotaxis anytime in the next few months is off the table." 3. **Waymo Comparison:** * Waymo averages ~570,000 driverless miles *every day*. * Tesla (at sustained growth) would hit 130,000 unsupervised miles *every week*. * Dylan calls this comparison "apples to oranges" and "largely foolish" but offers it for context. 4. **"Zero Notable Incidents":** * Ashok's statement about "zero notable incidents over 380,000 miles." * Dylan points out: "notable" is undefined, and it refers *only* to truly unsupervised miles. * NHTSA SGO data includes both supervised and unsupervised miles. * 380,000 miles is a "small sample size," concentrated in "limited areas" with "favorable conditions." This cannot be extrapolated to millions of miles across the country. * Zero incidents in limited conditions "does not mean it's going to stay that way" in unlimited operational domains. There is no inherent contradiction in Tesla's messaging. 5. **New City/State Rollout:** * Ashok: Expect "time to launch in a new city will continue to trend towards zero," aiming for "entire states as a whole instead of city by city." * Dylan's "cold water": This is "not going to happen this year," "earliest sometime next year." * Analysts should have pressed for timelines. "Trending towards zero" could be a 5-10 year process. * V15 (bleeding-edge FSD) is running on the robotaxi fleet. * Running new, untested V15 on robotaxis is *safer* in a "tightly knit controlled environment" with abundant data, compared to V14 deployed globally. Robotaxi is the "tip of the spear." * This implies V14 is "not good enough for unsupervised" yet. * Analysts didn't ask about the "seven different major improvement tracks" in V15. **V. Acquisition:** * Details revealed in Q2 10Q: Tesla acquired an AI hardware company for $1.95 billion (common stock/equity). * $1.73 billion is subject to service conditions and/or performance milestones. * ~$222 million allocated to patents and related technology. * This is an "aqua-hire," acquiring technology, IP, and employees, relying on them for integration. * Dylan speculates the acquisition is Density AI (based on Q1 rumors): * Co-founder/CEO Ganesh Venkataramanan previously worked at Tesla (senior director for autopilot hardware, involved with Dojo). * Rumor: ~20 other Tesla engineers (from Dojo reorganization) joined Ganesh at Density AI. * Density AI focuses on specialized AI accelerators, hardware, full-stack systems, memory architecture for high-density, efficient inference (large models, long context workloads) for data centers, automotive, and sensor fusion. Dylan believes this would be "a great fit." **VI. FSD Subscription Statistics:** * Chart shows Tesla's "apparent 55% attach rate for FSD in Q2 in North America." * Vybov stated 1.48 million active FSD subscriptions: * 45% paying monthly (~666,000 people). * 55% bought outright (~814,000 people). * FSD growth rate is "twice the rate of Tesla's car fleet." * Highlights FSD as a "margin upgrade" for the installed vehicle base. * **FSD recurring revenue calculation:** ~$791 million per year (1.48M * 0.45 * $100/month * 12 months). * This was "one of the more encouraging stats" as the attach rate was higher than expected (10-20%). * Notes that vehicle deliveries are not moving the needle, and even this FSD revenue "isn't moving the needle yet" for overall financials. **VII. Financials (Specific Call-outs):** * Dylan offers to do a deeper dive next week if desired. * **Energy gross margins:** Came down from ~39% to ~20% in Q2. * Vybov expects future energy gross margins to settle between 20-25% (previously thought to push 30-40%). This requires re-rating expectations. * **Regulatory credits:** First "meaningful decline," down to $146 million for the quarter (lowest watermark before this was $380 million). * The expiration of CAFE penalties "may finally be playing out." **VIII. Starlink & FSD on Semi:** * Elon: Starlink will be added to the rest of Tesla's fleet "in time" where offered. * Elon: FSD on Tesla Semi will be active "later this year or early next year," but current focus is FSD for high-volume vehicles. **IX. Conclusion:** * Apologizes for the video being a day late due to personal matters. * Plans for future videos to continue analyzing Q2. * Wishes viewers well for the Starship flight and the weekend. * Thanks Patreon supporters.