The Dam Is Finally Breaking at Tesla: "Sorcery Is Coming" ⚡️
发布时间 2026-03-18 00:34:19 来源
以下是视频中所有新闻内容的摘要：
**频道与内务管理：**
* **赞助者致谢：** 感谢了新的/更新的赞助者：Mark P、Roger R、YouTube Patreon 和 Todd L。
* **缩略图更新：** 承认了评论中偏爱旧缩略图的反馈；新缩略图将运行几周以适应算法，然后会在X平台和YouTube上进行投票决定。
**特斯拉制造与供应链扩张：**
* **三星在德克萨斯州的第二家工厂：** 三星宣布将在德克萨斯州泰勒市建设第二座270万平方英尺的晶圆厂（FAB 2），计划于明年开始量产，规模与FAB 1相同。这是因为美国主要科技公司在台积电产能受限的情况下寻求替代方案。
* **三星泰勒园区的潜力：** 三星在泰勒市购买了1,268英亩土地，足以容纳多达10座晶圆厂，主要专注于高性能计算和汽车领域（2纳米工艺）的芯片。他们已为该园区争取到121家客户。
* **特斯拉与三星的角色：** FAB 1将于明年开始量产，这是三星与特斯拉签订的价值165亿美元的AI5和AI6芯片协议的一部分。特斯拉作为FAB 1的主要客户（该厂最初难以找到客户）所展现的信心，被认为是FAB 2能够快速获取客户的原因。
* **FAB 2与特斯拉：** 尽管FAB 2不像FAB 1那样主要为特斯拉专用，但考虑到它们现有的合作关系，一部分产能很可能会流向特斯拉。
* **2纳米需求激增与行业瓶颈：** 行业面临硅供应瓶颈，特别是先进逻辑和内存制造。由于消费电子和AI加速器需求汇合，台积电的3纳米（N3）芯片正经历“需求冲击”。
* **代工厂多元化：** 产能限制正推动公司转向代工厂多元化，特斯拉通过与台积电和三星合作正在这样做。
* **AI与消费电子产品：** AI基础设施客户现在优先于消费电子产品获得先进芯片，因为其平均售价更高且要求复杂。
* **台积电产能问题：** 由于洁净室空间有限，台积电在未来两年内无法增加足够的N3产能。
* **特斯拉的TeraFab愿景：** 埃隆·马斯克认为行业对洁净室的优化不足。特斯拉的TeraFab预计将采用2纳米工艺，目标是实现独特的“完全密封隔离的晶圆”或“先进微环境”方法。
* **封装技术（Co-os/2.5D）：** 解释了Co-os（Chip on Wafer on Substrate，晶圆上芯片在基板上）作为台积电的封装步骤。描述了2.5D封装，其中逻辑芯片和高带宽内存（HBM）芯片并排放置在硅中介层上，然后再放置在基板上。
* **特斯拉TeraFab的优势：** 特斯拉旨在TeraFab开发自己的逻辑、内存和最终封装工艺，以满足其确切需求。
* **封装的未来：** 2.5D集成正被推向极限，导致需要决定哪些问题值得在硅层面解决。特斯拉的垂直整合使其处于有利地位。
* **TeraFab时间表：** TeraFab可能在接近2030年之前都“接近于零”，但这是特斯拉实现独立的战略性长期举措。
* **3D封装：** 提到了3D封装，作为未来芯片直接堆叠到基板上的潜力。
* **半导体行业复杂性：** 强调了硅行业迷人的复杂性，包括地缘政治和氦气供应等资源问题。
* **特斯拉/LG能源磷酸铁锂电池协议：** 内政部宣布特斯拉与LG能源达成一项供应协议，将在密歇根州兰辛市建设一座耗资43亿美元的磷酸铁锂（LFP）方形电池电芯工厂，并于明年投产。
* **MegaPack 3与储能：** 该工厂生产的美国制造磷酸铁锂电芯将为特斯拉在休斯顿生产的MegaPack 3和其他储能系统供电。
* **协议背景：** 该消息在2023年7月就已传出，但客户（特斯拉）当时并未公布。最初，这是LG与通用汽车之间Altium Cells协议的一部分；通用汽车出售了其股份，LG将该工厂转为为特斯拉生产磷酸铁锂电池。
* **LG磷酸铁锂协议的好处：** 避免了中国关税（目前磷酸铁锂电芯的关税约为70%），符合IRA国内激励措施，并加强了美国电池供应链。
* **信用结构：** LG全资拥有并运营密歇根州的工厂，并获得45X生产税收抵免（每千瓦时35美元），这将间接降低特斯拉的电芯成本。特斯拉随后将因在休斯顿组装模组和电池包而获得每千瓦时10美元的补贴。
* **LG磷酸铁锂供应时间表：** 尽管MegaPack 3的生产将于今年晚些时候在休斯顿以小规模开始，但密歇根州磷酸铁锂电芯的供应预计要到2027年8月，这意味着将有大约1年的时间由内华达州或海外供应来弥补。
* **特斯拉自己的磷酸铁锂工厂：** 特斯拉自己的磷酸铁锂工厂（约7吉瓦时产能）将于今年开始小规模生产，但需要时间提升产能。
* **能源业务利润：** 预计到2028年，这项磷酸铁锂供应将为特斯拉的能源业务带来“非常可观的利润增长”。
* **V4超级充电桩部署：** 纽约超级工厂已经建造了最后一个V3超级充电柜（七年内超过15,000个），现在正加大V4充电柜的生产，以在全球部署。
* **澳大利亚最大的电动汽车充电站：** 特斯拉正计划在澳大利亚昆士兰州麦凯市建设一个拥有超过25个充电位的超级充电站，很可能是能够提供高达250千瓦功率的V4充电桩。
**特斯拉自动驾驶与机器人出租车发展：**
* **拉斯维加斯机器人出租车目击：** 在内华达州亨德森市的一个停车场，发现了一批配备摄像头清洁器（与客户车队不同）的Model Y，拉斯维加斯是计划于2024年上半年推出的城市之一。这些车辆挂着德克萨斯州车牌。
* **达拉斯机器人出租车目击：** 在达拉斯发现了一辆红色Model Y，配备后置摄像头清洗器和德克萨斯州车牌，正在测试，模拟接送服务。达拉斯也是计划于上半年推出的城市之一。
* **奥斯汀机器人出租车规模化：** 奥斯汀车队中现在有三辆无人监管的机器人出租车，这可能表明在经历了一段有限的地理围栏测试期后，正呈现缓慢而积极的规模化趋势。
* **Cybercab可访问性：** 埃隆表示，Cybercab的座椅高度设计符合标准轮椅高度，车门运动学设计允许轮椅平行停放于座椅旁。
* **Cybercab控制：** 在快闪活动中，Cybercab通过手机进行控制。
* **自动驾驶汽车的法规参与：** Cybercab首席工程师Eric在首届全国自动驾驶安全论坛上与交通部会面，强调“创新的美国制造自动驾驶汽车如Cybercab将极大提升道路安全”。
* **NHTSA FMVSS现代化：** 美国国家公路交通安全管理局（NHTSA）提议修订联邦机动车安全标准（FMVSS）102，以豁免配备自动驾驶系统但无手动控制的车辆，使其无需遵守变速箱换挡位置显示要求（例如：驻车、空挡、前进）。意见征询期将于4月15日结束。
* **特斯拉对FMVSS的立场：** 特斯拉声称Cybercab已符合所有当前的FMVSS要求。
* **欧洲的全自动驾驶（FSD）：** 一位欧洲特斯拉客户看到用户界面通知，称当从意大利进入奥地利时，“监督式FSD将不可用”。埃隆预计，监督式FSD的RDW（荷兰道路交通管理局，常用作欧洲型式认证的代指）批准将于3月20日获得。
**更广泛的AI、科技与埃隆的愿景：**
* **英伟达轨道数据中心：** 在GTC大会上，黄仁勋宣布英伟达将推出用于轨道数据中心的自有计算平台，包括用于太空任务的Avira Rubin Space 1模块。
* **埃隆先前的说法得到证实：** 黄仁勋的声明被认为是证实了埃隆早前受到批评的关于太空数据中心的言论。
* **空间冷却挑战：** 黄仁勋承认了工程上的障碍，特别是在太空中的冷却问题（缺乏对流），尽管人们普遍误解太空很冷。
* **AI竞赛：** 埃隆表示AI竞赛将归结为扩大算力和芯片产量。
* **埃隆的AI预测：** 针对谷歌DeepMind将赢得AI竞赛的说法，埃隆表示“几年之内”，然后“SpaceX将远远超越所有其他公司总和”，这引发了“SpaceX”可能隐性包含特斯拉的猜测。
* **SpaceX收购特斯拉：** 日益增长的协同效应和SpaceX的双层股权结构使得SpaceX收购特斯拉“越来越有意义”。
**竞争对手新闻与行业转变：**
* **英伟达/优步L4级机器人出租车合作：** 英伟达和优步宣布深化合作，目标是到2028年实现L4级无人驾驶机器人出租车。优步的服务将于2027年上半年在洛杉矶和旧金山启动，并于2028年扩展到28个城市，但“无人驾驶运营”预计要到2028年才能实现。
* **对竞争对手自动驾驶的怀疑：** 主持人表示怀疑，指出自动驾驶“最后的2%”非常困难，英伟达提供的是工具而非可运行的神经网络，以及传统汽车制造商在电动汽车盈利能力上挣扎，这些公告更多是为了安抚投资者。
**其他新闻：**
* **Roadster发布：** 埃隆表示新款Roadster的发布“可能在4月下旬”，称其将是“震撼人心、达到新高度”。
* **特拉维斯·卡兰尼克谈论Roadster：** 优步创始人表示，“知情人士”将Roadster即将到来的发布描述为“魔法和魔术”。
* **证券交易委员会 (SEC) 半年度财报提案：** SEC正在准备一项提案，允许公司选择性地半年度而非季度报告财务业绩。主持人评论说，这可能对公司有利，但对投资者/粉丝不利。
* **特斯拉股价表现：** 特斯拉股价收盘价为179.27美元（上涨0.94%），而纳斯达克100指数（NDX）上涨0.51%。交易量比平均水平低22%，表明交易持续低迷（5900万股，而去年下半年为9500万股）。
Here's a summary of every single news item from the video:
**Channel & Housekeeping:**
* **Patron Shoutouts:** Thanked new/updated patrons: Mark P, Roger R, YouTube Patreon, and Todd L.
* **Thumbnail Update:** Acknowledged comments preferring older thumbnails; will run new ones for a few weeks for algorithm, then poll on X and YouTube to decide.
**Tesla Manufacturing & Supply Chain Expansion:**
* **Samsung Second Factory in Texas:** Samsung announced a second 2.7 million sq ft FAB (FAB 2) in Taylor, Texas, set for mass production next year, identical in size to FAB 1. This is due to major US tech companies seeking alternatives to TSMC amid capacity constraints.
* **Samsung's Taylor Campus Potential:** Samsung bought 1,268 acres in Taylor, large enough for up to 10 FABs, focusing on chips for high-performance compute and automotive (2nm process). They've secured 121 customers for the campus.
* **Tesla's Role with Samsung:** FAB 1 is starting mass production next year as part of Samsung's $16.5 billion deal with Tesla for AI5 and AI6 chips. Tesla's confidence as a major customer for FAB 1 (which struggled to find customers initially) is credited for FAB 2's rapid customer acquisition.
* **FAB 2 & Tesla:** While FAB 2 isn't primarily dedicated to Tesla like FAB 1, a portion is likely to end up in Tesla's hands given their existing partnership.
* **2nm Demand Surge & Industry Bottlenecks:** The industry faces a bottleneck in Silicon supply, specifically advanced logic and memory fabrication. TSMC's 3nm (N3) chips are experiencing a "demand shock" due to convergence of consumer electronics and AI accelerator demand.
* **Foundry Diversification:** Capacity constraints are pushing companies towards foundry diversification, which Tesla is doing by working with both TSMC and Samsung.
* **AI vs. Consumer Electronics:** AI infrastructure customers are now getting priority over consumer electronics for advanced chips due to higher average selling prices and complex requirements.
* **TSMC Capacity Issues:** TSMC cannot add enough N3 capacity for the next two years due to limited clean room space.
* **Tesla's TeraFab Vision:** Elon Musk believes the industry is sub-optimizing clean rooms. Tesla's TeraFab, expected to be 2nm, aims for a unique "fully sealed and isolated wafers" or "advanced micro-environments" approach.
* **Packaging Technology (Co-os/2.5D):** Explained Co-os (Chip on Wafer on Substrate) as TSMC's packaging step. Described 2.5D packaging where logic and HBM (High Bandwidth Memory) dies are placed side-by-side on a silicon interposer, then on a substrate.
* **Tesla's TeraFab Advantage:** Tesla aims to develop its own processes for logic, memory, and final packaging at TeraFab, specializing for its exact needs.
* **Future of Packaging:** 2.5D integration is being pushed to its limits, leading to decisions on which problems are worth solving in silicon. Tesla's vertical integration positions it well for this.
* **TeraFab Timeline:** TeraFab might be "close to zero" until closer to 2030, but it's a strategic long-term move for Tesla's independence.
* **3D Packaging:** Mentioned 3D packaging as a potential future where dies are stacked directly onto the substrate.
* **Semiconductor Industry Complexity:** Highlighted the fascinating complexity of the silicon industry, including geopolitics and resource issues like helium supply.
* **Tesla/LG Energy LFP Battery Deal:** Department of Interior announced a supply agreement between Tesla and LG Energy to build a $4.3 billion LFP (Lithium Iron Phosphate) prismatic battery cell facility in Lansing, Michigan, launching production next year.
* **MegaPack 3 & Energy Storage:** American-made LFP cells from this plant will power Tesla's MegaPack 3 and other energy storage systems produced in Houston.
* **Deal Background:** This news was known in July 2023, but the customer (Tesla) wasn't named. Originally, it was part of an Altium Cells deal between LG and GM; GM sold its stake, and LG pivoted the facility to LFP for Tesla.
* **Benefits of LG LFP Deal:** Avoids Chinese tariffs (currently ~70% on LFP cells), qualifies for IRA domestic incentives, and strengthens US battery supply chain.
* **Credit Structure:** LG fully owns/operates the Michigan plant and receives the 45X production tax credits ($35/kWh), which will indirectly lower cell costs for Tesla. Tesla will then receive $10/kWh for assembling modules and packs in Houston.
* **Timeline for LG LFP Supply:** While MegaPack 3 production begins in Houston at smaller scales later this year, the Michigan LFP cell supply is not expected until August 2027, creating a ~1-year gap covered by Nevada or overseas supply.
* **Tesla's Own LFP Factory:** Tesla's own LFP factory (~7 GWh capacity) is starting small-scale production this year but will take time to ramp up.
* **Energy Business Margins:** This LFP supply, by 2028, is expected to be "very margin accretive" for Tesla's energy business.
* **V4 Supercharger Rollout:** Gigafactory New York has built its last V3 Supercharger cabinet (over 15,000 units in seven years) and is now ramping up V4 cabinet production for worldwide deployment.
* **Australia's Biggest EV Charging Site:** Tesla is planning a Supercharger site with over 25 bays in Mackay, Queensland, Australia, likely V4 stalls capable of up to 250 kW.
**Tesla Autonomy & Robotaxi Development:**
* **Las Vegas Robotaxi Sightings:** A parking lot full of Model Ys with camera cleaners (unlike customer fleet) was spotted in Henderson, NV. Las Vegas is a planned launch city for H1 2024. The vehicles had Texas plates.
* **Dallas Robotaxi Sightings:** A red Model Y with a rear camera washer and Texas plates was spotted testing in Dallas, simulating pick-up and drop-off. Dallas is also a planned H1 launch city.
* **Austin Robotaxi Scaling:** Three unsupervised robotaxis are now in the Austin fleet, potentially indicating a slow, positive scaling trend after a period of limited geofence testing.
* **Cybercab Accessibility:** Elon stated Cybercab's seat height is designed for standard wheelchair height, and door kinematics allow wheelchairs to park parallel to the seat.
* **Cybercab Controls:** At pop-ups, the Cybercab was controlled using a phone.
* **Regulatory Engagement for AVs:** Cybercab lead engineer Eric met with the Department of Transportation at the first national AV safety forum, emphasizing "innovative American-made AVs like the Cybercab will vastly improve roadway safety."
* **NHTSA FMVSS Modernization:** NHTSA is proposing to amend FMVSS 102 to exempt ADS-equipped vehicles without manual controls from transmission shift position display requirements (e.g., Park, Neutral, Drive). The comment period ends April 15th.
* **Tesla's Stance on FMVSS:** Tesla claims the Cybercab already meets all current FMVSS requirements.
* **FSD in Europe:** A European Tesla customer saw a UI notification that "supervised FSD will become unavailable" when crossing from Italy to Austria. Elon expects RDW approval for supervised FSD by March 20th.
**Broader AI, Tech & Elon's Vision:**
* **Nvidia Orbital Data Centers:** At GTC, Jensen Huang announced Nvidia's own compute platforms for orbital data centers, including the Avira Rubin Space 1 module for space missions.
* **Elon's Prior Claims Vindicated:** Jensen's announcement was noted as vindicating Elon's earlier, criticized statements about data centers in space.
* **Space Cooling Challenges:** Jensen acknowledged engineering hurdles, particularly cooling in space (lack of convection), despite common misconceptions about cold space.
* **AI Race:** Elon stated the AI race will come down to scaling power and chip output.
* **Elon's AI Prediction:** Responding to a claim that Google DeepMind would win the AI race, Elon said "for a few years," then "SpaceX will far exceed everyone combined," leading to speculation that "SpaceX" might implicitly include Tesla.
* **SpaceX Acquisition of Tesla:** The growing synergies and SpaceX's dual-class share structure are making a SpaceX acquisition of Tesla "make more and more sense."
**Competitor News & Industry Shifts:**
* **Nvidia/Uber Partnership for L4 Robotaxis:** Nvidia and Uber announced a deeper partnership aiming for Level 4 driverless robotaxis by 2028. Uber's service would start in LA and San Francisco in H1 2027 and scale to 28 cities by 2028, with *driverless operations* not expected until 2028.
* **Skepticism on Competitors' Autonomy:** The host expressed skepticism, noting the difficulty of the "final 2%" of autonomous driving, that Nvidia provides tools but not a working neural network, and that legacy auto is struggling with EV profitability and making these announcements to appease investors.
**Other News:**
* **Roadster Unveil:** Elon stated the new Roadster unveil is "probably in late April," calling it a "banger next level."
* **Travis Kalanick on Roadster:** The Uber founder said those "in the know" describe the Roadster's upcoming reveal as "sorcery and magic."
* **SEC Semi-Annual Earnings Proposal:** The SEC is preparing a proposal to allow companies to report financial earnings semi-annually instead of quarterly, optionally. The host commented this would be good for companies but bad for investors/fans.
* **Tesla Stock Performance:** Tesla stock closed at $179.27 (+0.94%), while NDX was up 0.51%. Volume was 22% below average, indicating continued depressed trading (59M vs. 95M in H2 last year).
摘要
Save 50% on your system and your first month is free when you sign up for professional monitoring.
GPT-4正在为你翻译摘要中......
中英文字稿 
This video is sponsored by SimplySafe and let me tell you, there's no safe like SimplySafe. Welcome to Electrified, it's your host, Dylan Umis, quick shout out to my new or updated patrons, Mark P, Roger R, YouTube Patreon and Todd L. Thank you all for choosing to support the channel. Real quick, on the new thumbnail situation, I've seen the comments that some of you strongly prefer the older version, so I hear you, I'm gonna run the new ones just for a few weeks, I just need something different for now in the recommendation algorithm, but after a few weeks, I will run a poll here and on X and we'll see what the majority prefer and then I'll decide what to do from there.
这个视频由SimplySafe赞助，让我来告诉你，SimplySafe给你无与伦比的安全感。欢迎来到《Electrified》频道，我是你的主持人，Dylan Umis。先感谢一下我新的和更新的资助者，Mark P、Roger R、YouTube Patreon 和 Todd L。谢谢你们选择支持这个频道。关于新的缩略图，我注意到有些观众更喜欢旧版本。所以我知道大家的意见。我会只在接下来的几周内使用新的缩略图，因为我现在需要在推荐算法上做些调整。但几周之后，我会在这里和X平台上进行投票，看大家的多数意见，然后再决定接下来的做法。
But as always, thank you for the feedback. We may be in a bit of a lull when it comes to the Robotaxi expansion right now, but basically everywhere else you look at Tesla right now, there are signs of significant expansion underway. Not only is Tesla actually moving forward with its own tariff app, but we now have Samsung who's more of a partner with Tesla announcing a second factory in Texas.
但一如既往，感谢你的反馈。目前，Robotaxi 的扩展可能有些停滞，但除此之外，特斯拉的各个方面都显示出显著的扩展迹象。不仅如此，特斯拉正在推进其自己的收费应用程序，三星作为特斯拉的合作伙伴之一，现在还宣布将在德克萨斯州建造第二家工厂。
This one also in Taylor set for mass production next year. It decided on this second factory as major US tech companies line up to place their orders amid constraint capacity at TSMC. The project has entered the early stages of regulatory preparation, and Council document show the extension allows the city to continue supporting regulatory review for the planned second factory. Those documents show the motion passed unanimously. The planned FAB 2 facility is said to be about 2.7 million square feet the same size as FAB 1, which we know is currently under construction.
这家公司的这项计划也将在明年进行大规模量产。考虑到TSMC产能有限，而美国主要科技公司纷纷下单，他们决定建设第二家工厂。这一项目已经进入了监管准备的初期阶段，而市议会的文件显示，这一扩展允许城市继续支持对规划中的第二工厂进行监管审查。文件显示，这一提议获得了一致通过。据称，计划中的FAB 2工厂大约占地270万平方英尺，与目前正在建设的FAB 1工厂的规模相同。
And construction of FAB 2 will begin once the building permits are issued. But the news that was largely being overlooked beyond FAB 2, more factories could follow. Samsung bought 1268 acres of land in Taylor to build at semi cluster, which is large enough to accommodate up to 10 FABs. The campus will focus on producing chips for high performance compute and automotive applications using the company's 2 nanometer process. Now we have reporting that Samsung has already secured 121 customers for the campus.
一旦建筑许可颁发，FAB 2的建设将立即开始。然而，有一个消息常常被忽略，那就是在FAB 2之后，可能会有更多的工厂跟随。三星在泰勒市购买了1268英亩的土地用于建造一个半导体集群，这个地方足够容纳多达10个工厂。这个园区将专注于使用公司2纳米工艺生产高性能计算和汽车应用的芯片。目前有报道称，三星已经为这个园区成功签约了121位客户。
Potential large orders could come from major tech firms like Google, and Bight Dance. The first plant is preparing to start mass production next year as part of Samsung's deal with Tesla. As a refresher, FAB 1 is preparing to start mass production next year as part of Samsung's deal with Tesla. They signed at 16.5 billion dollar contract to which Elon said that was just the beginning and it would likely far exceed that number in time.
可能的大订单可能来自像谷歌和字节跳动这样的主要科技公司。作为三星与特斯拉协议的一部分，第一座工厂正在准备明年开始大规模生产。提醒一下，FAB 1作为三星与特斯拉协议的一部分，也在准备明年开始大规模生产。他们签署了一份合同，金额达165亿美元，对此埃隆表示这只是个开始，未来可能会远远超过这个数字。
But it'll start with making AI5 and AI6 chips. But for now it remains unclear whether the extra orders from Tesla have expanded the scope of the deal which previously covered only AI6 chips. Don't forget at FAB 1 for a few months Samsung was struggling to find any customers until Tesla came in and basically saved the day. As far as we know, Tesla scooped up nearly all of the expected available capacity at FAB 1.
这将从制造AI5和AI6芯片开始。不过，目前尚不清楚特斯拉的额外订单是否扩大了交易范围，因为之前的协议只涉及AI6芯片。别忘了，在几个月内，FAB 1工厂一直难以找到客户，直到特斯拉的介入才扭转了局面。据我们所知，特斯拉几乎抢购了FAB 1预计的所有产能。
And now just a few months later with that vote of confidence from Tesla, they apparently have 121 different customers lined up for FAB 2. And as you'll see shortly there's very good reason for why there is a surge for two nanometer demand right now. So to be clear, FAB 2 is not primarily dedicated to Tesla like FAB 1.
几个月后，得益于特斯拉的信任支持，他们已经为FAB 2准备好了121个不同的客户。很快你就会明白，为什么现在对2纳米技术的需求如此旺盛。需要明确的是，FAB 2并不像FAB 1那样主要为特斯拉服务。
However, given Tesla's existing partnership with Samsung and what Tesla did for Samsung to get them that first major order and major customer in Taylor, there's a very strong chance that at least a portion of the FAB 2 will end up in Tesla's hands. So why the rush for two nanometer demand? Well, this is an article from semi-analysis and Dylan Patel who's widely regarded as one of the best semi-conductor industry analysts that we have.
然而，考虑到特斯拉与三星已有的合作关系，以及特斯拉为了让三星获得泰勒的第一个大订单和主要客户所做的努力，很有可能FAB 2至少一部分会落入特斯拉手中。那么，为什么要急于满足2纳米的需求呢？这篇文章来自半导体分析领域，被誉为我们当今最优秀的半导体行业分析师之一的Dylan Patel。
The bottlenecks in this industry continue to shift every few months, but right now it is Silicon supply. There's simply not enough advanced logic and memory fabrication. And remember, Tesla's TerraFab will be unique because they're combining all three. The logic the memory and the packaging all under one roof. TSMC calls its three nanometer chips N3 and up until today N3 demand has been driven primarily by consumer electronics, smartphones, laptops, computers, etc.
在这个行业中，瓶颈每隔几个月就会变化，但目前的瓶颈在于硅材料的供应。先进的逻辑和存储芯片制造能力远远不够。此外，特斯拉的TerraFab会很独特，因为他们将逻辑芯片、存储芯片和封装统统集中在一个地方进行生产。台积电（TSMC）将其三纳米芯片称为N3，迄今为止，N3的需求主要是由消费电子产品如智能手机、笔记本电脑和电脑等推动的。
But this year we have all of the main AI accelerator families also transitioning to N3. So this sudden convergence of N3 adoption coupled with continued growth of AI compute demand has resulted in a huge demand shock for N3 wafer capacity. And these capacity constraints may therefore push customers to explore greater foundry diversification, which is exactly what Tesla's doing going to both TSMC and Samsung for AI56 and beyond.
今年，我们所有主要的AI加速器家族也正在转向使用N3技术。因此，N3技术的采用突然激增，再加上AI计算需求的持续增长，导致了对N3晶圆产能的巨大需求冲击。这种产能限制可能会促使客户探索更多样化的代工厂选择，这也正是特斯拉所做的，即同时选择台积电和三星来生产AI56及以后的产品。
And look, if we want to continually have an excellent handle on Tesla as a company and as investors and fans, we have to understand the Silicon world now. And of course, in the months and years ahead, I'll do everything in my power to continually educate everybody including myself on the depths of this industry. But this year, AI infrastructure customers are getting clear priority over consumer electronics. AI accelerator designs usually have bigger die sizes and more complex packaging requirements, which translate to higher average selling prices.
看，如果我们想不断地掌握特斯拉作为一家公司的动态，并作为投资者和粉丝保持关注，我们现在就必须了解硅谷的世界。当然，在未来的几个月和几年里，我会尽我所能不断地向大家（包括我自己）深入介绍这个行业的知识。但今年，人工智能基础设施的客户明显优先于消费电子产品。AI加速器的设计通常具有更大的芯片尺寸和更复杂的封装要求，这意味着更高的平均售价。
And why can't TSMC just add more N3 wafer starts, like memory suppliers, TSMC is constrained by available clean room space. Additional usable fab area must first be built before equipment can be installed and new capacity brought online. And for the next two years, TSMC will not be able to add enough capacity to fully meet demand. So clearly, a bottleneck exists and we cannot forget about Elon's bet that at Tesla's new Cara Fab expected to be two nanometers, Elon will be able to eat cheeseburgers and smoke cigars in the clean room.
为什么台积电不能像存储器供应商那样增加更多的 N3 晶圆产能呢？这是因为台积电受限于现有的无尘室空间。在安装设备和增加新产能之前，必须先建造出额外的洁净厂房。在接下来的两年里，台积电将无法增加足够的产能来完全满足需求。所以，很明显存在瓶颈。此外，我们还不能忘记埃隆在特斯拉新 Cara Fab 工厂2纳米技术上的赌注，他说将可以在无尘室里吃汉堡和抽雪茄。
So obviously, Elon believes that the industry is doing something sub optimally with the clean rooms. And all we've really heard about Elon's idea is that it has to do with fully sealed and isolated wafers or these advanced micro environments. And then for today, one term that will come up often going forward is co-os, which just stands for chip on wafer on substrate. This is TSMC's naming for its packaging step of the process.
显然，埃隆相信，在无尘室这个领域，业界的做法并不够理想。目前，我们所了解到的埃隆的想法主要涉及到完全密封和独立的晶圆，或者这些先进的微环境。今天，有一个术语将会经常出现，那就是 "co-os"，即芯片-晶圆-基板。这是台积电用来描述其封装步骤的命名。
And zooming out industry wide, 2.5D is the overall type of packaging that the industry is largely using right now. And it gets that naming because it's somewhere between 2D and 3D. And so the packaging process is largely where you take the logic die, which is the system on chip, and then you actually pair it or connect it to the HBM or the high bandwidth memory. And the HBM is typically made up of a stack of somewhere between 8 to 12 dies.
从行业整体来看，2.5D是目前普遍使用的封装类型。之所以称为2.5D，是因为它介于2D和3D之间。这个封装过程大致包括将逻辑芯片（即系统芯片）与高带宽内存（HBM）配对或连接起来。HBM通常由8到12层芯片堆叠而成。
And rather than being a fully 3D process where you would stack the SOC and the HBM on top of each other, they are side by side. And then effectively connected by these microscopic copper wires, where both are connected to a silicon interposer and then ultimately to a substrate, which acts as the mechanical structure. So the memory and the logic and the interposer are all very brittle, so the substrate is effectively what holds it together, gives it some structure.
而不是采用完全的3D堆叠工艺，将SOC（系统级芯片）和HBM（高带宽存储器）彼此叠放；它们是并排放置的。然后通过微小的铜线连接在一起，这些铜线将它们与硅中介层连接，最终再连接到充当机械结构的基板。因此，存储器、逻辑芯片和中介层都非常脆弱，基板实际上起到了将它们结合在一起并提供一定结构的作用。
And again, it has copper and solder balls for actually bringing power up through the motherboard and into the interposer and transferring that power evenly to the HBM and SOC. All that to say, one of the things I'm most excited for for Tesla's Terrafab is that they'll be able to develop their own processes for the logic, for the memory, and for that final packaging step. Already, these foundries have their own versions of how they do things and what they specialize in, but Tesla will be able to specialize for exactly what it needs.
再说一次，计算机主板通过使用铜和焊球来接通电源，将电力传输到中介层，并均匀地分配到高带宽内存（HBM）和系统级芯片（SOC）上。我最期待特斯拉的Terrafab项目的一点是，他们将能够自主开发适合自身需求的逻辑、存储器和最终封装工艺。目前，各大代工厂都有各自的工艺和专长领域，但特斯拉将能够专注于他们的具体需求。
And again, the plan is to have that specialization at all three steps. And this year, packaging engineers and foundries have converged on the same conclusion. 2.5D integration is now the primary tech responsible for power delivery, signal integrity, at extreme bandwidths, mechanical stability across large areas, and in some cases, even active functionality. AKA, it's where everything comes together.
再说一次，计划是在所有三个步骤上都有专业化。今年，包装工程师和代工厂在同一个结论上达成了一致：2.5D集成现在是主要技术，负责在极高带宽下的电源传输、信号完整性、机械稳定性，并且在某些情况下，还提供了主动功能。简单来说，这就是所有元素交汇的地方。
And this 2.5D packaging is being pushed to its limits, so the question is what will the next phase look like, as these companies are now being forced to decide what problems are worth solving in silicon and which are not. And at inflection points like that for a company that's so heavily focused on vertical integration, it'll be a great thing for Tesla to get to that point.
这种2.5D封装技术已经达到了极限，因此问题在于下一阶段将是什么样子，因为这些公司现在被迫决定哪些问题值得在硅片上解决，哪些不值得。在这样的转折点上，对于一个高度专注于垂直整合的公司来说，特斯拉达到这一点将是一件好事。
Now, we have to understand Tesla's tariff app might be close to a zero until closer to 2030. But we know the time is going to pass anyway, so for Tesla to set itself up through the 2030s, in position A, where it's less reliant on the rest of the industry, makes such a big risk like this actually makes sense. I was able to find a simpler graphic comparing the 2.5D process, where we are now, to a potential 3D future.
现在，我们必须理解特斯拉的关税可能会接近于零，至少要到接近2030年才会有所改变。但我们也知道，时间总会过去，因此特斯拉为了在2030年代中处于领先地位，让自己减少对其他行业的依赖，这样冒一个大风险其实是说得通的。我找到了一张更简单的图表，可以用来比较我们现在所处的2.5D工艺与未来可能实现的3D工艺。
Again, the 3D is where the dies would be stacked on top of each other directly onto the substrate. Whereas today, the stack of high-bandwidth memory dies and the logic dies are still next to each other. So again, if you hear me say co-os in the future, that's TSMC's process for the packaging step. Chip, on wafer, on substrate.
再次说明，3D技术是指将芯片层直接堆叠在基板上。而目前，高带宽内存芯片层和逻辑芯片层仍然是并排放置的。因此，如果你以后听到我提到"co-os"，那是指台积电在封装步骤中的工艺，即芯片、在晶圆、在基板上。
And one more oversimplified diagram to show how these connections in the interposer and through the substrate actually work. And this connection between the high-bandwidth memory and the logic happens at the packaging phase. I will say it might feel a bit overwhelming at first trying to get into the silicon space, but I have found it to be so incredibly fascinating.
再提供一个简化的图表，展示中介层和基板中的这些连接是如何实际运作的。其中，高带宽存储器与逻辑之间的连接是在封装阶段实现的。我想说，刚开始涉足芯片领域时，可能会觉得有些难以理解，但我发现这确实非常有趣。
Not just because so much of our world is now directly reliant on this technology, but when you understand that these chips that are mostly made of sand and all of these EUV machines, the most complicated machines that humans are making right now, that are massive to do this microscopic etching on this sand to then power most of the technology that we know and love is really just mind-blowing stuff. And when you consider geopolitics and consequences of the war, like a potential shortage in helium that there's currently no replacement for in silicon production. You just have this fascinating web of factors and all of the biggest AI CEOs and tech CEOs now speedrunning to build out their own data centers both on earth and in space.
不仅是因为我们现在的世界在很大程度上直接依赖于这项技术，还因为当你了解到这些芯片主要是由沙子制成的，以及所有这些极紫外光刻机——这是目前人类制造的最复杂的机器——在这个沙子上进行微观蚀刻，从而为我们所熟知和热爱的绝大多数技术提供动力时，你会感到非常震撼。再想到地缘政治和战争的后果，比如目前在硅生产中无法替代的氦气可能短缺，就会发现这是一张引人入胜的复杂网络。而现在所有的大型人工智能和科技公司的CEO们都在加速建设他们自己的数据中心，既在地球上，也在太空中。
Here's something I think about often. How many times a day do you check your test lab? Honestly, the charge status, century mode alerts, dash cam viewer, cabin temperature before you go outside. Some of us can be a little obsessed with monitoring our cars from our phones. I'll admit, I am. But it took me an embarrassingly long time to realize I should have that same level of visibility into my actual home. And not just a camera, but a real system, sensors, alerts, live feeds, the works. And that realization is a big part of why I partnered with simply safe the sponsor of this video.
这里有一些我经常思考的事情。你每天查看实验室设备的次数有多少？说实话，无论是充电状态、百年模式警报、行车记录仪查看器，还是外出前查看车内温度。有些人可能会有点痴迷于通过手机监控自己的汽车。我承认，我就是这样。但是，我花了很长时间才意识到，我应该对自己的家也有同样程度的信息了解。不仅仅是装个摄像头，而是真正的系统，包括传感器、警报、实时视频等等。而这个意识正是我为什么与这段视频的赞助商Simply Safe合作的一个重要原因。
I told you guys about our old house flooding, it was quite devastating, insurance back and forth for weeks, dozens of hours of cleanup, and a ton of sentimental property ruined forever. What I haven't really talked about is our family's anxiety that stuck with us after that. So now I have simply safe water sensors placed in the spots that matter. Anywhere water might come in the house under sinks, basement walls. So if anything goes wrong, I get an alert on my phone right away, just like I would get a century mode alert with my Tesla. And that's what I love about the system, you can tailor it to your life.
我之前和你们提到过我们老房子被水淹的事，那次真的很糟糕。保险公司来回折腾了好几周，清理工作花了几十个小时，还有很多有纪念意义的物品被永久破坏了。不过，我还没怎么谈过这个事件给我们家带来的焦虑。现在，我在重要的地方装了简易水传感器，比如水可能渗入的地方，像水槽下方、地下室墙壁等。所以一旦发生漏水，我马上就在手机上收到提醒，就像收到特斯拉的哨兵模式警报一样。我喜欢这个系统的原因是，它可以根据你的生活进行定制。
I have motion sensors I activate when actually and I travel, door sensors on every entry point, and cameras around the house that I can pull up live from anywhere. And the app is genuinely where it all comes together. One tap and you're looking at live camera feeds, you can arm and disarm the system, check the sensor status, it's clean, it's fast, and it really is the kind of interface that a Tesla owner would appreciate. It just works. And what sealed it for me when I was researching is that simply safe has no long term contracts, which was huge. Monitoring plans start at around a dollar a day with no hidden fees.
我有运动传感器，一旦我们旅行的时候就会激活，每个入口都有门传感器，家周围装有摄像头，我可以随时随地查看实时画面。所有这些功能通过一个应用程序整合在一起。只需轻轻一点，你就能看到实时摄像头画面，还可以启用或禁用系统，检查传感器状态，界面简洁且操作迅速，非常适合特斯拉车主使用，一切运转良好。让我最终决定选择它的原因是，Simply Safe 没有长期合同，这点很重要。监控计划大约每天一美元起，没有隐藏费用。
I did the entire installation myself, but they do have professional installation too if that's more your speed. And there's a reason they've been named the best home security system by US News and World Report for five years in a row. So if you already live the monitoring everything from your phone lifestyle like I do, I think you owe it to yourself to have that same visibility into your home. I partnered with SimplySafe to get you guys an exclusive deal so you can save 50% on your SimplySafe security system when you sign up for a professional monitoring plan and you'll get your first month free.
翻译如下：
我自己完成了整个安装，但如果你更喜欢专业安装，他们也提供这项服务。并且，他们已经连续五年被《美国新闻与世界报道》评为最佳家庭安防系统，是有原因的。因此，如果你和我一样喜欢用手机监控一切，我建议你也应该拥有对自己家里的同样的掌控。我和SimplySafe合作，为大家争取到了一个独家优惠：当你签约专业监控计划时，可以在购买SimplySafe安防系统时省下50%的费用，并且第一个月免费。
Visit SimplySafe.com slash electrified to customize yours. And remember there's no safe like SimplySafe. The department of the interior just put out a press release saying Tesla and LG energy are expanding their partnership with a supply agreement to build a $4.3 billion LFP, prismatic battery cell facility and Lansing Michigan production to launch next year. Meaning American made sales will power Tesla's MegaPak 3 and energy storage systems produced in Houston.
访问 SimplySafe.com 斜杠 electrified 来定制您的产品。请记住，没有什么比 SimplySafe 更安全。内政部刚刚发布了一份新闻稿，宣布特斯拉和 LG 能源正在扩大合作，达成供应协议，将在密歇根州兰辛建立一个价值43亿美元的LFP棱柱形电池生产设施，预计将在明年启动生产。这意味着美国制造的电池将为特斯拉在休斯顿生产的 MegaPak 3 和能源储存系统提供动力。
We've actually known about this news since July of last year when LG made it public that they had a supply contract for over $4 billion with some company but at the time they did not name the customer. However, the expectation speculation was it was Tesla. And depending on your perspective this one might be even sweeter because originally this was part of Altium cells meaning it was a deal between LG and GM. But spring of last year GM sold its entire stake back to LG and then LG pivoted the facility from NMC chemistry to LFP and to prismatic cells.
实际上，自从去年七月LG公开宣布与某公司签订了超过40亿美元的供应合同以来，我们就已经知道了这则消息。但当时他们没有透露客户的名字。然而，人们普遍猜测这家公司可能是特斯拉。根据不同的观点，这条消息可能更加令人兴奋，因为最初这笔交易是Altium Cells的一部分，这意味着LG和通用汽车（GM）之间的合作。然而，去年春天，通用汽车将其持有的股份全部卖还给了LG，随后LG将工厂的生产方向从NMC电池化学转向LFP技术和棱柱电池。
So Tesla scooped up all of that supply that GM did not have the demand for. And this is a triple threat. Number one avoids those Chinese tariffs on LFP cells. It will then qualify all of this LFP supply for some domestic incentives through the IRA. And third it will strengthen not just Tesla's supply chain but the US national supply chain for battery supply as well. A very important distinction to make here this site is fully owned and operated by LG meaning all of the 45X credits the production tax credits will be going to LG.
所以，特斯拉抢购了通用汽车没有需求的那部分供应。这是一个三重优势。首先，这样可以避免中国对磷酸铁锂（LFP）电池的关税。其次，这将使所有的LFP供应符合《美国降低通胀法案》的本土激励政策。第三，这不仅会加强特斯拉的供应链，还会增强美国国家的电池供应链。这有一个重要的区别需要注意：这个工厂由LG完全拥有和运营，这意味着所有的45X生产税收抵免将归LG所有。
However the standard industry practice is for these battery producers getting those credits to lower the cost of the cells or the price of the cells for the ultimate consumer which in this case is Tesla. Meaning in directly Tesla is likely to benefit greatly from these production tax credits they just won't be receiving them directly. However don't forget Tesla will still be able to most likely get some direct credits being the $10 per kilowatt hour for taking those cells and assembling them into modules and packs in Houston. Again specifically for MegaPak 3. So the 35 dollars per kilowatt hour will go direct to LG and then they will lower the price of the cells for Tesla. Then Tesla will buy those cells as the consumer. Assemble those cells into modules and packs in Houston and then get that $10 per kilowatt hour credit directly.
然而，行业的标准做法是这些电池生产商获得税收抵免后，会降低电池芯的成本或价格，最终受益的客户就是特斯拉。这意味着特斯拉虽然不会直接收到这些生产税收抵免，但会间接大幅受益。不过，别忘了，特斯拉大概率仍然可以直接获得一些抵免，比如每千瓦时10美元的抵免，因为他们在休斯顿把这些电池芯组装成模块和电池组。再次强调，特别是用于MegaPak 3的电池。每千瓦时35美元的抵免将直接给到LG，然后LG会降低电池芯对特斯拉的售价。之后，特斯拉作为消费者购买这些电池芯，将其在休斯顿组装成模块和电池组，并直接获得每千瓦时10美元的抵免。
Just keep in mind that wow MegaPak 3 production is expected to begin in Houston at smaller scales later this year. This new LFP cell supply from Michigan is not expected to be available for Tesla until sometime around August 2027. So there will be roughly a one year period where Tesla has to get LFP supply from either Nevada or overseas. But again long term this is a big win for Tesla because largely Tesla's getting its LFP cells now from CATO.
请记住，预计MegaPak 3将在今年晚些时候在休斯顿开始小规模生产。然而，来自密歇根的新型LFP电池预计要到2027年8月左右才会供应给特斯拉。因此，大约有一年的时间，特斯拉需要从内华达州或海外获取LFP电池。然而，从长远来看，这对特斯拉来说是一个巨大胜利，因为特斯拉目前主要从CATO获得LFP电池。
And right now the current tariff rate on cells from China is in the neighborhood of 70%. And Tesla's own LFP factory around 7 gigawatt hours of eventual capacity is starting production on a very small scale this year. But as we know that will certainly take some time to ramp up and be ready. So not only will Tesla eventually avoid those high tariffs but they'll also benefit from the direct manufacturing incentives available in the states.
目前，从中国进口电池的关税大约是70%。而特斯拉自家的磷酸铁锂（LFP）电池工厂，预计产能大约为7吉瓦时，今年开始小规模生产。但正如我们所知，要完全提高产能需要一些时间。因此，特斯拉将来不仅可以避免这些高额关税，还可以在美国享受直接的制造业激励政策。
So the energy business never gets the love and attention it deserves. But starting sometime in 2028 this should be very margin accretive for Tesla. We had the Zacker sharing a short video of a parking lot full of model wise in Henderson in the suburbs of Las Vegas. He posed the question, could it be Tesla getting ready for a Robotaxi launch in Las Vegas? And primarily the speculation is driven that these model wise have all of the camera cleaners which so far the customer fleet has not gotten.
因此，能源业务从未得到应有的关注和重视。但从2028年开始，这一领域对特斯拉的利润贡献应该会显著提升。一位名叫Zacker的人分享了一段简短视频，视频中展示了拉斯维加斯郊区亨德森一个满是Model Y车辆的停车场。他提出了一个问题，这是否意味着特斯拉正在为在拉斯维加斯推出无人驾驶出租车做准备？主要的猜测是，这些Model Y车辆都配备了摄像头清洁器，而这是目前顾客车辆中尚未配置的功能。
And here are some photos to confirm the camera washer. It should be noted that parking lot is indeed a self-serve test drive lot for Tesla. But as we know Las Vegas is one of the cities that Tesla plans to launch in in the first half of this year. So nothing official but I would say those camera washers and the red color that Tesla's been using for unsupervised vehicles in Austin is an encouraging sign. Another city you'll find on Tesla's list for the first half is Dallas.
以下是一些照片，用来确认摄像头清洗装置。需要指出的是，停车场确实是一个特斯拉的自助试驾场地。但正如我们所知，拉斯维加斯是特斯拉计划在今年上半年进入的城市之一。所以虽然没有官方的消息，但我认为这些摄像头清洗装置和特斯拉在奥斯汀用于无人监督车辆的红色标识是一个令人鼓舞的信号。特斯拉上半年计划进入的城市名单中还有达拉斯。
And Sawyer got a photo from a follower again of a red model Y with the rear camera washer doing some testing in Dallas, Texas. And both these vehicles and the ones in the lot in Las Vegas all had Texas plates. Plus one of the model wise in Dallas was spotted simulating a pickup and a drop off. So we're roughly halfway through the first half of this year and we've had very little robot taxi motion so in the second half of the first half things should really pick up.
Sawyer 又收到一位粉丝发来的照片，照片上是一辆红色的 Model Y，后置摄像头正在达拉斯进行测试。这辆车和那些停在拉斯维加斯的车辆都挂着德州的车牌。而且，在达拉斯拍到的一辆 Model Y 还在模拟接送任务。我们现在大约处于今年上半年的中期，而到目前为止，自动驾驶出租车的进展还很少。所以在上半年的剩余时间里，这方面的进展应该会加速。
And Jesse shared a video saying another unsupervised robot taxi was added to the fleet now tessels up to three in Austin. As I've said for a while I do not think we should be placing so much emphasis on the numbers from the robot taxi tracker. But there is a chance we are in the midst of unsupervised slowly scaling in Austin right now. We started with around eight unsupervised vehicles then Tesla went down to one in a limited geofence for a while but now we may be reversing that trend going positive once again.
Jesse 分享了一个视频，说现在在奥斯汀车队中新增加了一辆无人监督的机器人出租车，现在总共有三辆。正如我之前所说的，我认为我们不应该过分关注机器人出租车追踪器上的数量。但现在有可能我们正处于在奥斯汀进行无人监督车辆缓慢扩展的阶段。我们最开始有大约八辆无人监督的车辆，然后特斯拉一度减少到只有一辆，并且限定在一个有限的区域内运行，但现在这种趋势可能正在逆转，再次向积极方向发展。
Too early to say if that limited geofence has been expanded for unsupervised rides but certainly something we will keep an eye on. And on the cyber cab Eric also said we built the seat height to be in line with standard wheelchair seat height. Accessibility is very important in autonomous vehicles freedom of transportation for everyone is critical.
目前尚无法确定这种有限的地理围栏是否已被扩展用于无人驾驶服务，但这肯定是我们将密切关注的一个方面。关于自动驾驶出租车，Eric 还提到我们把座椅高度设计成与标准轮椅座椅高度一致。可访问性在自动驾驶车辆中非常重要，因为交通自由对每个人都至关重要。
Eric said this combined with the door kinematics a wheelchair can park parallel to the seat with plenty of room unlike traditional doors. And just a fun fact when Tesla was positioning the cyber cab at the pop up they were using a phone to control it. We also just had Sean Duffy say are you ready to ditch your steering wheel? That referring to the first ever national AV safety forum that the Department of Transportation held.
Eric 说，这种设计结合了车门运动学，轮椅可以平行停靠在座椅旁边，空间充足，与传统车门不同。而且，一个有趣的事实是，特斯拉在展示他们的Cyber Cab时，用手机来控制车子。我们刚刚还听到Sean Duffy问，你准备好抛弃方向盘了吗？这句话是关于交通部举行的首次全国自动驾驶安全论坛。
Then the lead cyber cab engineer Eric shared that post saying thank you and your team for meeting with us this week at the AV safety forum. Innovative American made AVs like the cyber cab will vastly improve roadway safety and we look forward to working with you on launching them at scale this year. And we just got this from the federal register the FMVSS modernization for number 102 to accommodate ADS equipped vehicles.
首席网络出租车工程师埃里克分享了那篇帖子，并表示感谢你和你的团队在本周的自动驾驶安全论坛上与我们会面。由美国创新制造的自动驾驶汽车（如网络出租车）将极大地提高道路安全性，我们期待在今年大规模推出这些车辆时与你们合作。此外，我们刚刚从《联邦公报》上了解到，FMVSS第102号规定将进行现代化改进，以适应配备自动驾驶系统的车辆。
Nitzah is proposing to amend FMVSS 102 and in short the proposed modification would accept vehicles equipped with automated driving systems that do not have manually operated driving controls from the requirement for a transmission shift position meaning a vehicle like the cyber cab that will not have wheels or pedals would no longer need to display park neutral drive. So it's not groundbreaking and there will be a comment period that last until April 15th this year but we will take what we can get and remember all of these FMVSS modifications might be moot because Tesla already said even as they stand now the cyber cab meets all of them. So unless Nitzah challenges Tesla's interpretation of all of that we are no longer at the mercy of the FMVSS at least when it comes to cyber cab.
Nitzah 正在提议修改美国联邦机动车安全标准（FMVSS）102。简而言之，该提议的修改将允许配备自动驾驶系统且没有手动驾驶控制的车辆不受传动档位显示的要求限制。这样的车辆，如未来可能推出的“Cyber Cab”（可能类似于特斯拉的产品），将不再需要显示停车、空挡、驱动等档位信息。这并不是一个突破性的改变，但公众可对该提案进行意见反馈，截止日期为今年4月15日。不过，特斯拉已经表示，即便在现行标准下，Cyber Cab 车型也符合所有要求。因此，除非Nitzah 对特斯拉的标准解读提出质疑，否则至少在Cyber Cab 这款车型上，我们不再受FMVSS 限制。
Yesterday in video at GTC announced its own compute platforms for orbital data centers. Jensen said space computing the final frontier has arrived as we deploy satellite constellations and explore deeper into space intelligence must live wherever data is generated. The first time Elon said anything about data centers in space there was immense backlash across all different industries saying that Elon was crazy and that it was not ready. And while we may have some time between now and when we have this at any sort of scale very clearly for Jensen to announce this publicly having Nvidia's backing we can be sure that the rest of the industry is going to try to follow suit and be a part of this next frontier and it really is so typical to have Jensen come after the fact and tout orbital data centers and everybody's clapping and cheering and high-fiving.
昨天，在GTC的一个视频中宣布了属于他们自己的轨道数据中心计算平台。Jensen说，随着我们部署卫星星座并深入探索太空，空间计算这一最终前沿已经到来，因为智能必须在数据生成的地方存在。最初埃隆提到太空数据中心时，各个行业都极力批评，说他疯狂，认为这一构想还未成熟。尽管从现在到这个概念大规模实现之间可能还有一段时间，但Jensen公开宣布这一消息并得到Nvidia的支持，显然整个行业将会尝试跟随这一趋势，成为这一新前沿的一部分。Jensen总是在事情后期出现，并推崇轨道数据中心，而在场的人都为此鼓掌欢呼，击掌庆祝，这种情况再典型不过了。
But months ago when Elon first brought up this idea and really started making the push for it it was largely met with backlash and mocking. Anyway in video said that its Avira Rubin Space 1 module will be used on space missions led by multiple companies. The chips are specifically engineered for size weight and power constrained environments. Importantly Jensen said Nvidia is working with partners on a new computer for orbital data centers but there are still engineering hurdles to overcome. In space there's no convection there's just radiation and so we have to figure out how to cool these systems out in space but we've got lots of great engineers working on it.
几个月前，当埃隆首次提出这个想法并开始大力推动时，它大多遭到了反对和嘲讽。然而，在视频中提到，Avira Rubin Space 1 模块将用于由多家公司领导的太空任务。这些芯片专为尺寸、重量和电力受限的环境而设计。更重要的是，黄仁勋说，英伟达正在与合作伙伴一起开发用于轨道数据中心的新型计算机，但仍需克服工程方面的困难。在太空中没有对流，只有辐射，所以我们必须想办法在太空中给这些系统降温，不过我们有很多优秀的工程师在努力研究这个问题。
I think that's an important point because you'll see some people saying space is so cold so cooling really won't be a challenge. But according to Jensen clearly for now it still is a challenge. So yes there will be hurdles and obstacles between now and then however directionally this all but confirms where everyone is going to be running. And as I've always said any competition is actually a great thing for Tesla and when these engineering bottlenecks get figured out because they will get figured out removing all of those major bottlenecks on earth. The environmental impacts all of the permitting all of the land availability concerns, the liquid cooling requirements and the water consumption, the power availability, electricity supply, the interconnection cues, all of that largely gets wiped out when those engineering challenges for orbital data centers are solved.
我认为这是一个重要的观点，因为有些人会说，太空非常寒冷，所以降温不会是个问题。但是根据詹森的说法，目前这仍然是个挑战。所以，是的，从现在到那时会有很多障碍和挑战，不过总体上这几乎可以确定大家都会朝这个方向努力。正如我一直所说，任何竞争对特斯拉来说实际上都是一件好事，当这些工程瓶颈被解决时——而它们终将被解决——地球上的所有主要瓶颈都会被消除。包括环境影响、所有的许可问题、土地可获得性问题、液体冷却需求、水资源消耗、电力可用性、电力供应、连接延迟，所有这些问题在解决轨道数据中心的工程难题后基本都会被消除。
And Elon just said the AI race will come down to scaling power and chip output. So this is exactly why Elon is strategically positioning his companies as we see him doing right now. And on the AI race a chuto said Google DeepMind is destined to win the AI race to which Elon said for a few years then SpaceX will far exceed everyone combined. My question here was Elon thinking that this SpaceX entity would include Tesla at that time. Because if it didn't we know Tesla is an AI company and it would be very hard for SpaceX to exceed Tesla if they were in the group of everyone else combined. Now that's obviously reaching but could Elon be embedding something in that statement.
埃隆刚刚表示，AI竞赛最终取决于扩展能力和芯片产量。这正是他目前正在战略性地布局自己公司的原因。在AI竞赛中，有人说谷歌DeepMind注定会赢得比赛，对此埃隆表示，这种情况会持续几年，然后SpaceX将远超所有公司之和。我的疑问是，埃隆在说这话时是否认为SpaceX会包括特斯拉。如果不是这样的话，我们知道特斯拉是一家AI公司，那么SpaceX要超越特斯拉所在的“所有公司之和”将会非常困难。这显然是个大胆的猜测，但埃隆是否在这句话中暗藏什么深意呢？
You guys know I'm usually not one to entertain things like this but I'll be honest I was the one that kept going back to that thought in this case. And given SpaceX's dual-class share structure that could make things easier for Elon's ownership a SpaceX acquisition of Tesla in the next few years continues to make more and more sense. Not that it wouldn't come without any trade-offs but the synergies and the overlap and the partnership continues to grow by the week. Max from Tesla said the team is all in on V4 cabinets. The brand new V4 cabinet line at Giganyork is automotive grade and spooling up for charging ubiquity around the world.
你们都知道，我平常不怎么会关注这样的事情，但坦白说，这次是我自己一直在反复思考。考虑到SpaceX的双重股权结构，这可能会让埃隆更容易持有股份，因此在未来几年内，SpaceX收购特斯拉显得越来越有道理。当然，这也有一些权衡在内，但两家公司间的协同效应、重叠和合作每周都在增强。来自特斯拉的马克斯说，团队正全力投入V4充电柜项目。纽约超级工厂的全新V4生产线达到了汽车工业标准，正为实现全球范围内的充电普及加紧准备。
He shared a pretty cool black and white photo but the big news was that Giganyork has built their last V3 Supercharger cabinet marking the end to 15,000 plus V3 cabinets over seven years. Now the V4 cabinet line is ramping up which means all new Supercharger locations going forward will be true V4 sites. And Tesla is now planning Australia's biggest EV charging site with more than 25 bays. Site plans say it'll be located in the town of McKay in Queensland. The reporting is saying they're likely to be V4 Supercharger stalls capable of up to 250 kilowatts. Just remember the Cybertruck is not currently available in Australia so those higher supercharging speeds aren't really applicable.
他分享了一张很酷的黑白照片，但更大的新闻是Giganyork已经完成了最后一个V3超级充电桩柜的建造，这标志着在七年间超过15,000个V3充电柜的结束。现在V4充电柜线正在加速生产，这意味着未来所有新的超级充电站都将是真正的V4站点。而且，特斯拉现在计划在澳大利亚建造最大的电动汽车充电站，拥有超过25个充电位。根据规划，这个站点将位于昆士兰的麦凯镇。报道指出，这些充电桩很可能是能支持高达250千瓦的V4超级充电桩。请注意，Cybertruck目前在澳大利亚还没有上市，因此这些更高的充电速度暂时还不适用。
This image was shared by a European Tesla customer and the UI said upcoming country border supervised FSD will become unavailable. And that was while using enhanced autopilot from Italy to Austria. To be clear this one could be nothing these notifications have been around before but I'm sharing as a reminder that Elon said he's expecting supervised FSD approval from the RDW by March 20th which is just three days.
一位欧洲特斯拉客户分享了这张图片，显示用户界面提示即将在下一个国家边境时，FSD（全自动驾驶功能）将无法使用。而且这是在从意大利到奥地利使用增强自动驾驶功能时出现的提示。需要说明的是，这可能并不是什么大问题，因为这种通知以前也出现过。但我分享这个信息是为了提醒大家，马斯克之前表示预计RDW（荷兰车辆管理局）将在3月20日批准监督下的FSD，而这仅有三天的时间。
Nvidia and Uber announced a deeper partnership aiming for driverless level 4 rubber taxis by 2028. In the press release they do say the service would start in LA in San Francisco in the first half of 2027 and scale to 28 cities globally by 2028. But if you actually read the release you'll find the driverless operations are not expected until 2028. For now until we get more information on the program I've said what I want to say. Number one getting 98% of the way to autonomous driving right now is relatively speaking pretty easy.
英伟达和优步宣布加深合作，目标是在2028年前推出实现L4级别（Level 4）自动驾驶的出租车。在新闻发布中，他们表示这一服务将在2027年上半年首先在洛杉矶和旧金山启动，并在2028年拓展到全球28个城市。不过，如果你仔细阅读新闻稿，会发现实际上无人的操作预计要到2028年才实现。目前，在我们获得更多关于该计划的信息之前，我已经说了我想说的。首先，现在要实现98%的自动驾驶技术相对来说还算比较容易。
So companies making announcements and having demos that look pretty good is not that difficult but what people fail to understand is how excruciatingly difficult the final 2% is and further how long that actually takes. And I've already done a few episodes on how Nvidia is just providing tools legacy auto still needs to take these tools and know what to do. So these companies can't buy these tools and then magically overnight have a working neural network that is capable of FSD.
所以，公司发布看起来不错的公告和演示其实并不难，但人们往往忽视了最后2%的工作是多么艰难，以及这需要多长时间。我之前已经做过几个节目，讲述英伟达（Nvidia）只是提供了工具，传统汽车公司仍然需要使用这些工具并知道如何操作。因此，这些公司不能仅仅买了这些工具，就能在一夜之间神奇地拥有一个可以实现全自动驾驶（FSD）的神经网络。
So personally I would be very surprised if any of these legacy auto companies are able to remove the driver even by 2028. Because for now that's the plan but I am expecting delays. And as we've seen across the legacy auto industry and all of these companies pulling back from EVs taking tens of billions of dollars of write downs they are struggling profitability is struggling they need to do or say something to their investors to avoid mass Exodus.
所以，依我个人来看，如果这些传统汽车公司能在2028年前实现无人驾驶，我会非常惊讶。因为目前这确实是他们的计划，但我预计会有延迟。正如我们所见，传统汽车行业的许多公司正在缩减电动汽车的投入，损失达数百亿美元，他们在盈利上面临困境，需要向投资者展示或宣称些什么，以避免大量投资者撤资。
So they really have no choice but to say hey guys look here's our solution we're going to figure this out too with the help of Nvidia. And the unfortunate reality is somebody that's still invested in a GM or a Ford no offense at all if that is you everybody has different objectives but I'm saying the average investor that's not closely plugged into the autonomous driving space isn't going to know the difference they'll see the demos and the partnerships and how good things look already but 2028 is going to roll around and those people will be wondering what in the world is going on.
所以他们其实别无选择，只能说：“嘿，大家看，这是我们的解决方案，我们也要在Nvidia的帮助下解决这个问题。” 不过，不幸的现实是，如果你仍然投资于通用汽车或福特，我没有冒犯的意思，每个人都有不同的投资目标。但我要说的是，那些对自动驾驶领域不太了解的普通投资者是不太会分辨出差别的。他们会看到演示和合作伙伴关系，以及事情已经看起来有多好，但到了2028年，那些人会不禁感到疑惑：这究竟是怎么回事。
When Tesla is scaling its rubber taxi network while most of legacy auto is still nowhere near unsupervised autonomy. Elon said the new Roadster unveil is probably in late April and he also said the new Roadster unveil hopefully next month. It will be a banger next level. So late April is the new expectation pushed back from April 1st but that new date is not set in stone.
当特斯拉在扩大其自动出租车网络时，大多数传统汽车制造商在无人驾驶技术方面仍然没有实质性进展。埃隆·马斯克表示，新款Roadster可能会在四月底发布，同时他也提到，希望能够在下个月推出。这将是一项突破性的升级。因此，现在预计的发布日期推迟到了四月底，而不是之前计划的4月1日，但这个新日期也还没有最终确定。
And Travis Kalanik the founder of Uber said when I've run into people who are in the know I inquire they tell me nothing but their eyebrows raise and their eyes widen in a way that can only mean something of sorcery and magic is coming. The SEC is now preparing a proposal that would allow companies to report financial earnings semi-annually instead of quarterly. This proposal could be released as soon as April and the proposal is expected to make quarterly reports optional rather than eliminating them entirely.
Uber的创始人特拉维斯·卡兰尼克说，当我遇到知情人士并向他们打听时，他们什么都不说，但他们扬起眉毛、瞪大眼睛，这种反应只能意味着某种神奇的事情即将发生。美国证券交易委员会（SEC）现在正在准备一项提案，该提案将允许公司每半年而不是每季度报告一次财务收益。该提案最早可能在四月份发布，并预计会使得季度报告成为可选项，而不是完全取消。
So as I mentioned on X this would be a great thing for the companies but not so great for consumers, investors and fans. And selfishly for us to only get a peek behind the curtain twice a year for Tesla that would be brutal. Tesla stock closed the day at $399.27 up 0.94% while the NDX was up 0.51%. The volume was 22% below the average still very depressed from where it was even late last year.
正如我在X上提到的，这对公司来说可能是件好事，但对消费者、投资者和粉丝来说就不那么好了。而且，自私地说，每年只能两次窥探特斯拉的内部情况，对我们来说将会是残酷的。特斯拉的股票当天收盘价为399.27美元，上涨了0.94%，而NDX上涨了0.51%。成交量比平均水平低22%，仍旧比去年年底时的水平要低很多。
Over the last 30 days the average volume is right around 59 million whereas in the second half of last year it was closer to around 95 million. Don't forget check out simply say flinged below if you want to support the channel and most importantly protect your family even if you just start with one camera and work your way in over time. I really have found it to be a great system. Hope y'all have a wonderful day and a huge thank you to all of my patreon supporters.
在过去的30天里，平均交易量大约是5900万，而去年下半年则接近9500万。不要忘记查看下面简单说明，如果你想支持这个频道，并且最重要的是保护你的家人，即使你只是从一台相机开始，慢慢地逐步完善。我真的发现这是一个很棒的系统。希望大家有美好的一天，非常感谢我所有的Patreon支持者。