美国每年有3万到4万人死于交通事故。这个数字几乎不会出现在任何新闻标题里。
马斯克在特斯拉2026年第二季度财报电话会上主动提了这个对比,用它解释robotaxi(自动驾驶出租车)为什么没有一夜之间铺满全美。
如果我们哪怕让一个人受伤,那都会成为全世界的头条新闻,监管机构会立刻收紧对我们的限制。我们也不想让任何人受伤。(马斯克)
他还补了一句更狠的:理想情况下,连一只宠物都不要压到。
这场电话会的重心明显不在财务报表上。马斯克花了大量篇幅讲Optimus的量产难度、Terafab晶圆厂、Digital Optimus、Megapod、Starlink上车,以及美国电网里被浪费掉的一大半发电能力。财务数字反倒成了背景音乐。
把这些散落的信息拼起来,特斯拉正在同时推五件事:自动驾驶出租车、人形机器人、自研AI芯片、分布式AI算力、储能。五件事互相咬合,少一环整个循环就转不动。
Model Y 拿下全球销冠,顾客进店先问 FSD
第二季度交付量创了纪录。按马斯克的说法,Model Y 现在是全世界最畅销的汽车,不分车型、不分价位,而且在各个市场都在刷新纪录。
更有意思的是需求结构的变化。在已经批准 FSD(Full Self-Driving,全自动驾驶)的地区,FSD 的选装率高得异常。
很多人实际上是来买全自动驾驶的,车是附赠的。他们走进美国的门店,想要的是FSD,随便配一台什么车都行。(马斯克)
这句话听起来像营销话术,但它指向一个具体的商业逻辑:软件能力开始反过来定义硬件销量。过去买车的人问的是续航、空间、加速。现在一部分人问的是这套系统能不能自己开。
马斯克说,每批准一个新国家,都能看到类似的需求抬升。FSD 的国别审批节奏,因此成了特斯拉销量的一根隐形杠杆。
八条产线同时开动,二战后美国最快的工业扩张
电话会上被一笔带过的一串句子,拆开看相当惊人。
Cybercab 已经开始生产。Tesla Semi 已经开始生产。Optimus 即将开始生产。Megapack 3 即将开始生产。锂精炼厂投产了,正极材料厂投产了,电芯产能在扩。太阳能业务要从硅提纯一路做到电池片、组件,再到最后的电站部署。
这些产线分散在完全不同的行业里:整车、机器人、重型卡车、储能、锂化工、正极材料、电芯、光伏。任何一条单独拿出来,都要占用一家公司好几年的管理带宽。
马斯克自己的定位是:
我们正在做的事情,是人类历史上最雄心勃勃的先进基础设施制造产能建设。也许只有当年亨利·福特造T型车,或者二战期间美国企业大规模转产能相提并论。这可能是二战以来美国最快的工业扩张。这一点被低估了。(马斯克)
为什么要同时铺这么多摊子?他给了两条理由。
一条是交通电动化。另一条是AI。这两件事叠加起来,对电力的需求会大到今天的电网无法承受的程度。所以太阳能要从最上游的硅料做起,储能要扩到Megapack 3,电芯要自己造。横向扩张的动机在这里并不成立。真实的原因是下游需求曲线太陡,外部供应链根本跟不上。
资本效率可以牺牲一点,时间不能
有分析师问资本开支的节奏。马斯克的回答很坦率。
我让团队做的事情是,只要不太浪费,能花多快就花多快。我们不追求极致的资本效率,因为那会拖慢进度。(马斯克)
他把这件事讲成了一个净现值问题。省下一点钱、但把事情拖后半年,对公司的长期价值是负的。反过来,稍微不那么省、但提前投产,账算下来更划算。
这个判断放在重资产制造业里不算常见。多数公司会把资本效率当成KPI来管,因为财报上好看。马斯克选择的是另一条路:先把时间窗口占住,效率的事以后再说。
Robotaxi 的真实瓶颈:一个 9 一个 9 往上爬
robotaxi 已经在佛罗里达的几个城市、德州和湾区开放。开城的消息会第一时间发在 X 上的特斯拉官方账号,马斯克建议关心进展的人直接去那里看。
增长速度的数字是这样的:每周行驶里程增长超过10%。按周复合,这个曲线非常陡。
但马斯克说,真正卡住扩张的既不是车,也不是需求,更不是钱。
我们叫它「9的进军」。你需要多高的可靠性才能规模化?理想状态是99.999%。这是目前唯一真正限制robotaxi增长的东西。(马斯克)
从99%爬到99.9%,再从99.9%爬到99.99%,每一级付出的工程量都比上一级大一个量级。而这个行业的特点是,前面几十万次都平安无事,出一次问题,全世界都会看见。
需求端他一点都不担心。
robotaxi的经济性会非常有吸引力,需求会超过我们的服务能力。这种情况下,垂直整合是自然选择。(马斯克)
特斯拉打算自己运营整个robotaxi网络,像做其他业务一样,从车到调度到支付全握在手里。
Cybercab 缺的技术早就有了,缺的是里程
这里有个容易被忽略的工程细节。
Model 3 和 Model Y 在全球有几百万辆在路上跑,特斯拉手里握着海量的真实行驶数据,可以用来训练和验证自动驾驶系统。Cybercab 是一副全新的底盘,没有这个积累。
所以现在的做法是:给一部分 Cybercab 装回方向盘、刹车、油门,让它们带着安全员去跑里程,专门校准这台车的底盘特性。
马斯克说,等团队对这套校准有把握了,城市里 Cybercab 的数量会大幅增加。
这句话解释了一个外界常问的问题:为什么Cybercab已经投产,路上却看不到几台。限制它的是数据积累,产线产能反倒不是瓶颈。
Optimus 为什么慢:整台机器人没有一条现成供应链
马斯克对Optimus的判断没有变过:这会是史上最大的产品。但他这次把难度讲得比以往都细。
要做一个通用人形机器人,你随便让它做一件事,或者给它看一段视频,它就能完成,不需要任何编程。这个目标从来没有人实现过。(马斯克)
难点分三层。
第一层是机电设计。要做到足够的灵巧度,同时足够可靠、足够耐用。马斯克举了个很具体的画面:一台70公斤的机器人,如果在现场趴下了,你得把它像抬一具遗体那样搬出去。这种产品没法商用。
第二层是智能。他特意提醒,网上那些看起来很惊艳的机器人演示,绝大多数是预编程或者远程遥控的。真正能处理通用任务的人形机器人,目前一台都没有。Optimus 想做第一个。
第三层是量产,也是他认为最难的一层。
这会是特斯拉造过的最难量产的产线,因为机器人身上每一个零件都是新的。量产爬坡的难度,和零件的新颖程度成正比。(马斯克)
他拿汽车做对比。电动车的动力系统是新的,但轮子、后视镜、挡风玻璃、车身覆盖件这些东西,外面有一整条成熟的供应链,打个电话就能订货。Optimus 没有这样的供应链。整条链要么自己建,要么干脆自己做。
Fremont 工厂原来生产 Model S 和 Model X 的那条产线,已经被改造成了 Optimus 产线。马斯克说,去现场看相当震撼。
但他同时给了一个不那么好听的预期:Optimus 会走一条正常的S曲线,只是这条曲线的前半段会又平又长。原因就是零件太新。
关于手,他讲得有点动情。人手是个了不起的东西,越研究越觉得神奇,远不止一个对生拇指那么简单。Optimus 的设计目标是达到人类的灵巧度,最终超越人类。
Optimus 4 的目标是每年一千万台
电话会上给出了两代产品的产能口径。
Optimus 3 的目标是年产100万台。Optimus 4 放在奥斯汀生产,垂直整合程度高得多,目标年产1000万台,整整一个数量级。马斯克说,Optimus 4 会有大量PCB自己做。
他给这些数字加了很重的限定词:量产爬坡极其困难,所有目标都带前提。
把1000万台这个数字放在人形机器人这个品类里看,它的分量落在成本结构上。一台机器人如果要卖到普通家庭买得起的价格,唯一的路径就是把产量拉到汽车的量级。市场预测的事可以先放一边,先看产线够不够快。
没有芯片就没有机器人,所以特斯拉要自己建晶圆厂
Terafab 是这场电话会上信息量最大、披露最少的项目。
马斯克说,选址很快公布,但会有单独的发布会。他的解释是,这件事太大,塞进财报电话会是浪费。
没有Terafab,我们扩产Optimus的能力就会被卡住,因为根本没有足够的AI芯片。内存、逻辑、封装,三样都得解决。(马斯克)
奥斯汀的开发晶圆厂已经下了设备订单。这栋楼的设计值得单独说:光刻掩模生产、逻辑芯片、内存、封装、芯片测试,全部放在同一个屋顶下。
我不认为地球上存在这样一栋建筑。(马斯克)
把全流程塞进一栋楼,换来的是迭代速度。一颗新芯片设计出来,几天内就能流片、测试、看结果,而不是排几个月的队。这让特斯拉可以在AI芯片上下一些高风险高回报的赌注。
他还专门感谢了几家供应商。台积电在亚利桑那建厂,三星在德州建厂,两家都投入了数百亿美元为特斯拉的AI算力扩产。松下在电芯上投了几十亿美元。美光给了特斯拉一份相当可观的内存配额,而且条件合理。马斯克的原话是,这几年内存价格「疯得离谱」,美光能在这种时候腾出产能,很不容易。
顺带一提,他开场时嗓子不太舒服,自己说「今天状态有点不在」。这场电话会是在他身体不太舒服的情况下开的。
芯片路线图也给了时间点。AI4 的一个中等幅度改进版本,大概明年年中量产。AI5 同样瞄准明年年中,会先上Optimus。马斯克对AI5的评价极高,说它会是世界上最好的边缘计算芯片,如果有人的芯片比它更好,他愿意握手致意。
还有一个对存量车主有用的信息:凡是带摄像头的特斯拉,未来某个时间点升级到新一代AI板,财务上都是划算的。因为不做升级,改装成本太高。
Digital Optimus:给电脑屏幕装一套自动驾驶
这个概念第一次被讲得这么清楚。
开车这件事,本质是像素进、控制出。摄像头拍到画面,神经网络输出转向和油门。Digital Optimus 做的是同一件事,只不过对象换成了电脑屏幕:屏幕上的像素流进来,鼠标和键盘的控制指令流出去。
所以特斯拉为自动驾驶开发的那套AI技术,可以几乎原样迁移过去,做出一个「会自己开电脑」的系统。
技术细节上,它靠的是特斯拉自研的低成本AI电脑,实时处理高帧率视频,中间不靠隔几秒截一张图。这个区别决定了它能不能流畅操作复杂的软件界面。
Digital Optimus 对物理形态的Optimus是刚需。一台人形机器人走到触摸屏前面,不能站在那里发呆。它必须具备通用的触屏操作能力。
这个项目跟SpaceX合作。分工上,SpaceX那边的大模型扮演「经理」,负责拆解任务、下发指令,Digital Optimus 负责执行。
车里也已经上了Grok,Grok 同时在帮 Digital Optimus 干活。
Megapod:把 AI 算力塞进超级充电站
这个设计很有意思。
把一颗特斯拉 AI4 芯片和一颗 x86 芯片配成一对,装进一个箱子里,封装方式类似Megapack。这个箱子就叫Megapod。
Megapod 可以放在全国任何地方,甚至放到国外。它不需要依附于某个超大规模数据中心,只要有电就能跑。
特斯拉手里有什么?全球大约7吉瓦的超级充电站功率,而且还在增长。这些站点本来就接好了电网,分布在高速公路、城市边缘、商业区。把Megapod摆在超充站旁边,等于用分布式的电力来扩AI算力。
这可能是个相当大的事情。(马斯克)
这条思路避开了AI行业最硬的约束。建一个吉瓦级数据中心,要解决土地、电网接入、审批、散热,周期以年计。而把算力拆成一个个箱子,撒到已有的电力节点上,扩张速度完全是另一个量级。
Starlink 上车,因为硅谷也有信号黑洞
Starlink 会集成进Cybercab,也会集成进所有特斯拉车辆,前提是当地有Starlink覆盖。
理由非常朴素:robotaxi 不能失联。
硅谷有很多地方蜂窝信号很差,甚至完全没有,这在硅谷挺让人意外的。我自己开车上班,前10到15分钟根本打不了电话,连接太差了。(马斯克)
一台无人驾驶车如果开进了信号盲区,调度系统联系不上,乘客也没法求助,这种事出一次就够上头条。Starlink 补的是覆盖上的窟窿,网速快慢倒在其次。
第二个场景是乘客体验。车里有人的时候,人们想办公、想看视频。Starlink 能让乘客在车里看4K直播体育赛事,每GB数据的成本低到蜂窝网络做不到。
还有一个延伸用法:车上的Starlink终端本身可以当地面中继站,给附近的手机和需要WiFi的人提供连接。固定式的Starlink终端同样能干这活。
Semi 的自动驾驶先让路半年
Tesla Semi 已经开始生产,但它的自动驾驶功能被排在了后面。
马斯克给出的理由有两层。
一层是需求。美国卡车司机严重短缺,愿意开长途卡车的人越来越少,而卡车运输是整个美国物流的骨架。一台能自动驾驶的Semi,直接对着这个缺口。
另一层是资源分配。Semi 的产量还很小,到今年年底占特斯拉总车队比例依然很低。把自动驾驶的研发资源投在几百万辆Model 3、Model Y和即将上量的Cybercab上,回报高得多。
所以未来六个月,自动驾驶团队的重心放在3、Y和Cybercab的「9的进军」上。Semi 的自动驾驶大概今年底或明年初实现,赶在Semi大规模量产之前完成。
美国发电能力是平均用电的 2.5 倍,缺的是电池
能源业务这部分,马斯克讲了一组很容易被忽略的数字。
美国的装机发电容量大约1.2到1.3太瓦,但平均用电只有0.5太瓦。也就是说,发电能力是平均用电的2.5倍。
电网的问题出在电和用的对不上。白天光伏出力最高的时候用电不一定最高,晚上用电高峰光伏又停了。结果就是大量产能闲置,同时高峰期还要靠调峰机组救场。
光靠电池,美国的能源输出理论上就能翻倍。(马斯克)
Megapack 干的活里有削峰,但分量更重的是电网平衡,把风电和光伏的波动抹平。马斯克说,太阳能加电池的组合,会是未来全球绝大部分电力的来源。Starlink 卫星就是这么供电的,可以把特斯拉看成一颗巨大的卫星。
然后他把话题接到了AI上。
电力正在成为AI最硬的约束。只是找不到电还好办,更难的是把功率抹平。
在一次训练任务里,功耗可以在100毫秒内下降70%。你需要响应极快的先进电力电子设备,才能平滑这种剧烈的波动,尤其是在训练的时候。(马斯克)
这解释了SpaceX为什么买了那么多Megapack。外界以为是给数据中心备用供电,实际主要用途是给训练任务做功率平滑。
还有一层商业上的巧劲。如果特斯拉告诉公用事业公司,一年里最糟糕的那几个小时不用你们供电,电池能顶住,那拿电就变得容易多了。电网公司最怕的就是尖峰负荷,谁能帮它削掉尖峰,谁就更容易拿到并网许可。
这也是为什么马斯克说Megapack未来的需求会非常高。
几条可以自己跟的线索
这场电话会里没有给出明确时间表的东西很多,但可以观察的抓手其实不少。下面几条是可以自己长期跟的。
- 盯FSD的国别审批节奏。每批准一个新市场,都会出现一轮需求抬升,这是销量的先行指标,比交付数据早一个季度左右。
- 盯robotaxi每周行驶里程的增速。10%的周复合是很陡的曲线,一旦掉到5%以下,说明撞上了「9的进军」里的某一道坎。
- 盯Cybercab在城市里的实际投放数量。这个数字取决于底盘数据的积累进度,比产线投产时间更能反映真实节奏。
- 盯Terafab的独立发布会和选址。AI5瞄准明年年中量产,晶圆厂的时间表直接决定Optimus的产能天花板。
- 如果你是制造业从业者,Optimus 那条空白的供应链是机会清单。机器人身上的每一个零件,目前都还没有成熟的外部供应商。
- 盯Megapod的落地位置。如果它开始成规模地出现在超级充电站旁边,说明分布式AI算力这条路走通了,这会改变整个AI基础设施的成本结构。
- 盯Megapack在数据中心订单里的占比变化。功率平滑正在变成训练集群的刚需,这块需求的增速可能快过储能行业的整体增速。
- 别用「机器人演示视频」判断Optimus的进度。电话会上说得很清楚,网上那些演示基本都是预编程或遥控的。真正的信号是量产产线的爬坡速度,不是视频效果。
Thank you. So, yes, it's been a great quarter. We achieved record Q2 deliveries. Model Y, I believe it is now, I think it's the best-selling car of any kind in the world and is setting records across the board. So its popularity is increasing tremendously. And we're seeing in locations that have FSD approved, we're seeing a very high take rate of FSD. And in fact, I think, for a lot of people, they're actually buying Tesla full self-driving with a car attached as opposed to a car with FSD. They're coming into our stores in the U.S. and they want the full self-driving and with whatever car comes with essentially. So, clearly, the -- this is a significant demand driver.
And as we get approval for FSD in different part -- different countries, I think we'll see a similar uptick in demand. So, yes, it's looking very good on that front. The energy business is also growing incredibly fast, and I think it will be crucial for the scale-up of artificial intelligence, data centers. We're investing a lot in growing the core business and really preparing for the future. So this is a massive CapEx year, but I'm confident that all the things that we're investing in are -- will yield incredible returns, really the best CapEx returns that we've ever seen. And the Tesla team has performed incredibly well.
The Cybercab has started production, and we also have started production and are ramping in -- we will soon start production with Optimus. We have started production with the Tesla semi-truck. We will soon start production with Megapack 3. We started production with the lithium refinery, the cathode refinery. We obviously have -- we're scaling up battery cell production battery cell production. And we're also preparing to do a massive solar panel like solar cells and panel production.
Actually, this is going all the way from silicon refinement to producing the solar cell and then deployment of solar because there's going to be tremendous need for electricity in the future due to electrification of transport and AI, obviously. So we're working on what we believe is the most ambitious build-out of advanced infrastructure manufacturing capacity ever in history. It's bigger. So, with robotaxi, our goals are very ambitious for robotaxi, but we do need to be cautious about causing any accidents or causing any harm to anyone.
Although there are, I think, 30,000 to 40,000 automotive deaths per year in the United States alone, most of those do not generate any press or -- you never really read about almost any of those. But if we injure even one person, it will be worldwide headline news. And regulators will immediately clamp down on our activities. So -- and we don't want to injure anyone. So we're going as fast as humanly possible in scaling robotaxi with while trying to ensure that we do not harm anyone at all and ideally do not even run over a pet. So that's really the constraint is we're -- we want to grow as fast as possible with robotaxi without harm to anyone.
And we're providing live updates for robotaxi launch as we open up in a city. We immediately post that on X. So if you just follow the Tesla account on X, but we'll keep you informed as we scale to new cities. We've opened up a number of cities in Florida and in Texas and obviously in the Bay Area. So we'll continue to scale, I think, very, very rapidly on -- with it looks like more than 10% a week in terms of miles driven. So it's a very high compound growth rate. So, let's see, Optimus, as you've heard me say before, I think Optimus will be the biggest product ever, but it is a very complex problem to solve.
It's one of the hardest things to solve to make an autonomous humanoid robot that can do tasks that if you simply ask it to do something or show a video, it can do the task without any programming. Like no one's ever achieved this. And you've got -- there are many challenges in the electromechanical design of the robot to achieve sufficient dexterity and rely it to be very reliable and have a long wear and tear, like meaning it needs to be out in the field and not break down. Otherwise, you've got sort of a 70-kilogram robot that just flopped over and you got to carry it out like a body. It's our wheels. And we don't want Optimus to go haywire.
So it's a lot of work to scale -- to get the design right and to scale production. And I really want to emphasize here that the production scaling challenge is very substantial. This is going to be the hardest product to scale manufacturing that we've ever made at Tesla because everything on the robot is new. And the difficulty of scaling the production ramp is proportionate to the newness of the parts in the robot. And at least with electric cars, the nonelectric powertrain portions of the cars had an existing supply chain. So you could go to existing suppliers for wheels and side mirrors and windshield glass and that kind of thing, body panels. But with Optimus, there is no supply chain.
So we've had to build up the supply chain in its entirety or in-house the production. And we actually have in-house a tremendous amount. The Optimus production line that we're rolling out in Fremont in place of what used to be the model S & X production looks incredible. I mean it's quite stunning to see. So I just want to make sure to calibrate people correctly. Optimus will follow the sort of normal S-curve of a manufacturing ramp, but the initial portion of the S-curve will be quite flat and long because of the newness of the parts and the robot.
So you've probably seen lots of impressive demonstrations of robots on the Internet, but those demonstrations you're seeing are preprogrammed or remote controlled. So there is no humanoid robot that is actually able to do generalized tasks. Optimus will be the first one that is capable of doing that, where it's not just a demo, it's genuinely useful in day-to-day life. So -- and Optimus is designed to have full human dexterity. So a hand that has the same level of dexterity, if not higher than a human hand. And the human hand is an incredible, incredible thing. The more you study the human hand, the more amazing you realize how amazing hands are. It's more than just the opposable thumbs.
I mean the nuances of how human hands work are amazing as the closer you look, the more amazed you are. And Optimus will have that capability. You will have human and then super human dexterity. So it's really going to be quite something. Let's see, Terafab. The Terafab, we expect to announce the location soon and provide sort of more details about our plans in that regard, which we'll leave that to the product sort of the launch announcement rather than try to squeeze it into an earnings call because I think this is a very big announcement and to have its own sort of day in the spotlight and not be sort of squeezed into an earnings call.
But I do think Terafab is going to be an amazing initiative and a necessary one and one without which we will be constrained in our ability to scale Optimus production because we simply won't have enough AI chips. So it's crucial to solve that, and we'll have to solve memory, logic and packaging for -- in order to scale Optimus. So if anyone likes it, we'll leave that to an announcement. We've placed equipment orders for our development fab in Austin. And that development fab, I think, is pretty cool because it's intended to have lithography mask production and then logic, memory and packaging and chip testing all under one roof.
So you can have a very fast iterative cycle and try out new chip designs very quickly and see if they work. I don't think such a building exists anywhere on earth. So this is really going to be super helpful as we try some exciting adventurous high-risk, high payoff bets on AI chips. But it focused, like I said, on primarily Optimus, like how do we make Optimus as useful as possible. And we have like we have Digital Optimus, which is basically driving a computer screen, I guess you can call it computer use or something like that, but driving a computer screen in the same way you drive a car. So, for the car, it's sort of pixels in or photons in and controls out.
The same thing is true for Optimus and Digital Optimus. It's photons in, controls out. So we feel confident we can adapt the same Tesla AI technology that we developed for self-driving cars to have a self-driving computer screen, self-driving computer essentially. So the self-driving computer where the Tesla -- the very low cost, Tesla AI computer can handle all of the sort of short-term tasks. They're doing real-time video control of the screen of the computer. So it's not like screenshots type of things, it's real-time video and high frame rate. And Digital Optimus will be important, obviously, for physical optimus because physical optimus needs to be able to operate computers.
Like you can't come up to a touchscreen and not know what to do. It's got to have a generalized touchscreen and computer use capability. So that's looking promising. And this is in partnership with SpaceX. SpaceX is sort of the big model that is kind of like the manager of Digital Optimus and kind of tells Digital Optimus what to do, provides it with a series of tasks and then digital Optimus goes and does those tasks. We're also building out a megapod design that has Tesla AI4 computers with x86. So pairing x86 computer with the Tesla AI4 computer in a box, it's got Digital Optimus in a box. And in a megapod, kind of like the Megapack packaging.
And so we put just a large number of AI4+x86 combos in a giant box essentially. And these boxes can be placed anywhere in the country or outside the country. And this allows us to scale AI compute using disaggregated electricity production. So because there's lots of places all around the world, including in our superchargers. I think we've got something like 7 gigawatts of power at our superchargers and growing. And we can place megapods at many of these superchargers and have distributed power for AI. So, I think that, that could be quite a big deal. So, in conclusion, we're super excited about our autonomy and robotics road map. There's so much awesome stuff coming.
It's, like I said, hard to squeeze into an earnings call. But we'll have a lot of product announcements. And this is going to be a great year for Tesla, I think one of our best years ever. And then I think next year will be even better. So, in conclusion, I want to thank the Tesla team for their excellent execution and all of our supporters for joining us in this journey. Thank you.
Sorry, yes, our suppliers have been great, and they've made -- they have made and are making tremendous investments in support of Optimus and robotaxi and whatnot. Samsung and TSMC, in particular, are building fabs at TSMC in Arizona and Samsung in Texas and putting in tens of billions of dollars for -- to build AI compute for Optimus and robotaxi. Panasonic has also invested many billions in increasing battery cell production. I'm a little under the weather here. So I sound a little kind of struggling with some a little bit today. So, yes, but our partners have been great. And I'd like to thank them for their support and their investment and their hard work.
Yes. I actually I'd also like to thank Micron for giving us the memory allocation. They're going to make some very tough decisions on memory allocation, and we really appreciate Micron making room for Tesla in the years to come and giving us actually a very significant allocation on reasonable terms given the pretty insane pricing of memory these days.
Yes. We expect to be vertically integrated with robotaxi as we are in the rest of our business. So I don't think we're going to have any demand challenges with robotaxi. So this is -- the economics will be so compelling that I think we will really have a lot more desire to use the service then like I think demand will outstrip our ability to service the demand. And so obviously, you just -- there's -- you can just go vertical in that situation. So it's really just about going through what we call the March of 9s of reliability where you need like how many reliability do you need to scale? Ideally, you want 99.999% reliable. And so that's the -- it's just the March 9 of reliability.
I think the only thing really constraining our growth in robotaxi.
Well, I mean, as you can tell from the many collaborations on so many fronts with SpaceX and there's a lot -- there's more and more overlap, especially with Terafab, that's really going to be a gigantic project. So -- but obviously, we can't talk about combining companies and that kind of thing on an earnings call. It's got to be done with the appropriate process. And with that, I'll turn it over to Brandon, our General Counsel.
Yes. And there's many other things, too. Obviously, you've got Grok in the car. So -- and Grok helping drive Digital Optimus. You also got Starlink being integrated into the Cybercab and Starlink will be integrated into all of our vehicles, at least for markets that Starlink is active. Because for a robotaxi situation, you need to have coverage everywhere. And there are many places even in Silicon Valley where the cellular coverage is terrible or sometimes nonexistent, which is surprising for Silicon Valley. But I know when I want to drive to work with the first 10, 15 minutes, I can't actually do any calls because the cellular connectivity is so bad.
So we can't have robotaxis getting stuck in these like Bermuda triangles of lack of cellular connectivity. So Starlink with its ability to do connectivity anywhere is actually quite important. So we don't have robotaxis missing in action. And then obviously, if people are sitting in the car that they're going to want to do high productivity stuff or entertainment. And with Starlink, you can watch 4K live sports in the car and with very low cost per gigabyte of data that's really not feasible via the cellular system. And there's many other situations.
Yes. One of the things for Cybercab should be noted is because it is a new vehicle chassis, we need to accumulate driving data that is specific to the Cybercab before we can put a lot of them on the road. So unlike say Model 3 and Model Y and our other vehicles where we've got a lot of vehicles on the road, millions of vehicles on the road. We don't have that for Cybercabs. So we actually have to accumulate miles with Cybercab that are retrofitted with steering wheels and brakes acceleration braking battles that kind of thing to calibrate to the Cybercab chassis. So you'll see -- as we are confident about that, the number of Cybercabs in cities will increase dramatically.
Yes, absolutely. As you alluded to, there is a really serious shortage of truckers. So there just aren't enough people who around -- who want to drive trucks, which are crucial for transport throughout America. So an autonomous Semi is actually going to be very important to address the shortage of truck drivers. And obviously, it will be great for improving safety and making it easier on truck drivers to have a self-driving Semi is going to, I think, dramatically improve safety and comfort for truck drivers that use the Tesla Semi.
Since the units -- the total number of units of the Tesla Semi is still low and will continue to be a very small percentage even by the end of this year of our total vehicle fleet, it makes sense for us to focus our efforts, our self-driving efforts on our high-volume vehicles. So Model 3 and Y and solving self-driving for -- and really getting to the point where it's generalized unsupervised self-driving for those vehicles and Cybercab. So we expect to do -- get self-driving working on the Tesla Semi probably around the end of this year or early next year. And -- but I just don't want it to be a distraction on the March of 9s of self-driving of safety for self-driving on 3Y and Cybercab.
That's -- so it's taking a bit of a backseat for the next six months or so for autonomous semi. But it will be -- it will definitely be a bit working next year and in time for the scale up to high production of Tesla Semi. For communications, yes, I think there's an interesting angle there where you could use the Starlink terminals that are in the Tesla cars as like sort of cell towers, I suppose, or connectivity towers basically for providing relays on the ground to cell phones and to anyone who would like WiFi basically. But yes, I mean it is a different subject, but obviously, we could do that with the stationary stalling terminals, too.
You're talking about making like microcontrollers and like the various computers that are in the car that operate like the steering system and the braking system.
Yes. Optimus has a lot of really specialized power electronics and circuit boards. It's all Tesla design, design stuff, but it is fabricated by suppliers. For Optimus 4, which will be built in Austin, that will be a much more vertically integrated supply system for Optimus 4. And that we aim to have an order of magnitude more production of Optimus 4 than Optimus 3. So sort of aspirationally 10 million units a year versus 1 million units a year of Optimus 3. But with all the caveats there, which is intensely difficult to scale production. So -- and yes, Optimus 4 will be much more vertically integrated. So probably doing a lot of the sort of PCV work in-house for that.
I think it's going to make sense for -- to upgrade all cars with that are have less than hardware for almost any cars that have cameras basically because otherwise, it probably too many modifications. But anything that's got sort of -- that's set up for cameras, it will be financially sensible at some point to upgrade them. And I think we want to upgrade them to the next generation of next-generation AI board, so to be either like we have an upgraded AI chip that's sort of medium -- sort of moderate improvement over AI4, probably reaches production around the middle of next year. And then there's also AI5, which hopefully is in volume production around the middle of next year.
AI will initially go into Optimus and actually making very good progress. So Tesla chip team is really doing great work and making incredibly fast progress on AI. And we're -- I'm very excited about the design of the Tesla AI chip. I think it's -- I think it's going to be the best edge computing chip in the world. And if somebody has got a better one, I'd like to need that person to shake your hand because the AI chip is just so good. So I think we've got -- I think it's going really well on the chip front. And again, I'd like to thank TSMC and Samsung and Micron for their support.
Yes, it's a good question. I've actually my -- what I've asked the team to do is like really, we should be spending on CapEx as fast as we can spend -- as fast as we can without it being too wasteful. So we're not trying to aim for like some extremely high efficiency capital spend because that would slow things down. So it's a balance between like how much sort of capital efficiency versus time. So if we can be -- it's okay to be a little less capital efficient if we get things done sooner because that's actually going to be the higher NPV outcome for the company. So, overall, I'm pretty happy with how things are going.
I mean we're doing an incredible amount of construction and production growth in so many different arenas simultaneously. I don't think there's ever been a company that's done this at the scale. But maybe proportionately back when Henry Ford did the Model T or something like that, but World War II when companies were making -- had to sort of scale up reference production. But I think probably this is the fastest industrial scale-up since World War II in America.
This is underrated by the way. Please go ahead.
Yes. Our CapEx efficiency, I think, is all scale good.
Well, it's still mostly it's not just peak shaving, but it's also grid balancing. So it's not just like the facts are on briefly, but for balancing power from wind generation and solar generation, batteries are awesome. I mean the solar battery combination will be how the vast majority of energy in the world is produced in the future. That's really, of course, how all of the Starlink satellites are powered with solar panels and batteries. So -- and you can just think of us like a giant satellite. There's far more power available from the sun than anything else by far. And we think power constraints are going to be -- they really are a major issue for AI, just turning on the AI computers.
The AI compute demand is so high that people having -- even the -- the hyperscalers are having trouble turning on their AI compute and finding the power and then smoothing the power with -- especially for the training runs where the power cycles dramatically in a very short period of time, you can have -- during a training run, the power consumption can drop by 70% for 100 milliseconds. And you really need fast-acting advanced power electronics to be able to smooth out these massive -- massive changes in power, especially during the training runs. So that's why SpaceX has bought so many megapacks for the data centers. It's actually mostly for smoothing out the power for the training runs.
-- will also be useful for being able to tap more power from the grid by -- if you tell utilities that they can that they will not need to supply power during the worst hours or the worst days of the year because the batteries can handle it, then that makes it much easier for them to give you power. In fact, I mean, probably the best way to scale total energy production in the United States is simply using batteries because there's, I think, around 1.2 or 1.3 terawatts of power production in the U.S., but the average power usage is only 0.5 terawatt. So you've got 2.5x as much power generation as the average usage.
That means you could potentially double the energy output in the United States just with batteries. So really, we see demand for Megapack being very, very high in the future.
All right. Thanks everyone. This article is a transcript of this conference call produced for The Motley Fool. While we strive for our Foolish Best, there may be errors, omissions, or inaccuracies in this transcript. Parts of this article were created using Large Language Models (LLMs) based on The Motley Fool's insights and investing approach. It has been reviewed by our AI quality control systems. Since LLMs cannot (currently) own stocks, it has no positions in any of the stocks mentioned.
As with all our articles, The Motley Fool does not assume any responsibility for your use of this content, and we strongly encourage you to do your own research, including listening to the call yourself and reading the company's SEC filings. Please see our Terms and Conditions for additional details, including our Obligatory Capitalized Disclaimers of Liability. The Motley Fool has positions in and recommends Tesla. The Motley Fool has a disclosure policy. *Average returns of all recommendations since inception. Cost basis and return based on previous market day close. Invest better with The Motley Fool.
Get stock recommendations, portfolio guidance, and more from The Motley Fool's premium services. Making the world smarter, happier, and richer. © 1995 - 2026 The Motley Fool. All rights reserved. Market data powered by Xignite and Polygon.io. About The Motley Fool Our Services Around the Globe Free Tools Affiliates & Friends Record vehicle deliveries and robotaxi expansion drive capital spending surge.
所以,目前还没有哪个人形机器人真正能完成通用任务。Optimus 将是第一个有能力做到这一点的,它不是演示,而是在日常生活中真正有用。
So there is no humanoid robot that is actually able to do generalized tasks. Optimus will be the first one that is capable of doing that, where it's not just a demo, it's genuinely useful in day-to-day life.
我们预计 robotaxi 会像我们其他业务一样垂直整合。所以我认为 robotaxi 不会有什么需求挑战。
We expect to be vertically integrated with robotaxi as we are in the rest of our business. So I don't think we're going to have any demand challenges with robotaxi.
你们还会看到 Starlink 被整合进 Cybercab,而且 Starlink 将整合到我们所有车辆中,至少在 Starlink 活跃的市场如此。因为对 robotaxi 来说,你需要到处都有网络覆盖。
You also got Starlink being integrated into the Cybercab and Starlink will be integrated into all of our vehicles, at least for markets that Starlink is active. Because for a robotaxi situation, you need to have coverage everywhere.
关于 Cybercab,有一点应该注意:因为它是全新的车辆底盘,我们需要积累 Cybercab 专属的驾驶数据,然后才能把大量车辆投放到路上。
One of the things for Cybercab should be noted is because it is a new vehicle chassis, we need to accumulate driving data that is specific to the Cybercab before we can put a lot of them on the road.
所以,自动驾驶 Semi 对于解决卡车司机短缺问题实际上会非常重要。
So an autonomous Semi is actually going to be very important to address the shortage of truck drivers.
而且我们的目标是 Optimus 4 的产量比 Optimus 3 高一个数量级。所以,理想情况下是每年 1000 万台,而 Optimus 3 是每年 100 万台。
And that we aim to have an order of magnitude more production of Optimus 4 than Optimus 3. So sort of aspirationally 10 million units a year versus 1 million units a year of Optimus 3.
而且我们——我对 Tesla AI 芯片的设计非常兴奋。我认为它——我认为它将成为世界上最好的边缘计算芯片。
And we're -- I'm very excited about the design of the Tesla AI chip. I think it's -- I think it's going to be the best edge computing chip in the world.
但我认为这可能是美国自二战以来最快的工业扩张。
But I think probably this is the fastest industrial scale-up since World War II in America.
我的意思是,太阳能与电池的组合将成为未来世界上绝大多数能源的生产方式。
I mean the solar battery combination will be how the vast majority of energy in the world is produced in the future.
资源与出处
- 视频在 YouTube 收看
- 档案silicon-valley.tech 档案页本页视频与文稿数据来自 Elon Musk Archive
- 精读本篇精读 Markdown 版