特斯拉 2017 年股东大会上,马斯克花了一大段时间,公开说造 Model X 的方式是“傲慢至极”。
他给了两个比喻。第一个:Model X 像一颗法贝热彩蛋,是一件了不起的作品,但里面塞了太多炫技的东西,这些东西本该分批放到第二代、第三代去做。
第二个更直接。他说从 Model S 平台派生出一辆 SUV,这个决定从一开始就错了。
“我们试图从 Model S 的平台衍生出 Model X,这是个错误。更好的做法是按一辆 SUV 该有的样子,去设计一辆 SUV。”(马斯克)
台下坐的是股东。一个 CEO 在自己公司的股东大会上,把过去三年最贵的一款产品当成反面教材讲,这种场面不多见。
但这场演讲的很多内容,都围绕同一件事展开:特斯拉接下来要做的事,就是不停做减法。
Model 3 的配置器里只有两个选项:颜色和轮毂尺寸
马斯克说第一批 Model 3 的配置会“非常简单”,简单到只有两项:你要什么颜色,你要多大的轮毂。
其他版本什么时候有,会给一个时间表。但一开始不开放。
原因他讲得很清楚:复杂度是产能爬坡的头号敌人。配置每多一项,生产线上就要多一条分支,供应链上就要多一种零件,出错概率成倍往上翻。
他拿 Model S 举了个正面例子。Model S 刚投产的时候只有一个配置。为什么?因为当时特斯拉只会做那一个配置,再多就做不出来了。
然后他补了一句自嘲:后来我们变聪明了,于是有了 Model X。
“Model X 是一次彻头彻尾的傲慢。它是一颗法贝热彩蛋。这确实是一辆了不起的车,但它塞进了太多酷东西,这些本该放到第二版、第三版去做,那才是合理的节奏。我们当时太自信,造出了一个可能再也不会被复制的产品,也许本来就不该被复制。”(马斯克)
双电机版本是个具体例子。Model 3 的两台电机用了两套不同的架构,一台为高速巡航优化,一台为走走停停的城市路况优化,城市里续航最好,高速上续航也最好,加速还猛。
但马斯克说,一开始只上单电机。理由很直白:两台电机意味着出问题的概率翻倍,而产能爬坡阶段最不该做的事就是翻倍风险。双电机的配置要到今年年底,更可能是明年年初。
一万个零件,速度由最慢的那一个决定
马斯克在台上说了一句很朴素的话:造车难得离谱。
他的解释是,一辆车大约有一万个不同的零件,整条产线的速度,由最慢的那个零件决定。这一万个零件里任何一个出问题,车就下不了线。
这还不是最麻烦的部分。他接着说,特斯拉自己的工厂只占一环,往外看,供应链铺满整个地球。
“你想想,我们的供应链覆盖了整个地球。那么问题来了:地球上某个地方发生某种意外事件的概率有多大?相当高。所以我们不能从整个地球继承不可抗力风险。”(马斯克)
他给了一个实用主义的解法:缓冲库存。多备一些零件,某个供应商断供了还能顶一阵。但他紧接着补了一句,如果那个零件压根就没被生产出来,你备多少库存都没用。
这段话听起来像抱怨,实际上点出了 Model Y 的核心设计逻辑之一:简化供应链,减少对全球各地零散供应商的依赖。
一座工厂的电池产能,超过全球其他所有电池厂之和
讲到内华达的超级工厂(Gigafactory),马斯克的语速明显快了一档。
他放了一张当时的施工现场照片,然后说,照片上看到的还只是最终占地的一小部分。
“我们相信,几年之内,这一座工厂的电池产能,会等于全世界所有其他锂电池工厂加起来的总和。把美国、中国、欧洲,把其他所有地方的工厂全算上,这一栋楼的产出会超过它们之和。”(马斯克)
他承认这个规模不身临其境很难体会,用了一个词:令人瞠目。
规模带来的是议价能力。马斯克说,极高的产量把规模效应拉满,目标是拿到全世界最低的每千瓦时成本,同时还要做出最先进的电池。最好的技术配最低的成本,这个组合才是竞争力所在。
他还强调了一件事:全世界没有第二家公司在按这个量级规划电池产能。别人连尝试都没有。
超级工厂本身也被他当成一个产品来做。用他的说法,超级工厂就是一台巨大的机器,特斯拉会不断打磨它、把它产品化,然后复制到世界各地。最终会有 10 座、12 座,也许 20 座。
工厂安全:比行业平均水平低 32%
工伤率这组数字,马斯克讲得很简短,但给出的量级很具体。
当年到目前为止,特斯拉工厂的工伤率比汽车行业平均水平低 32%,而且逐月在改善。他预计这个数字会走到行业平均的一半以下,也会好过美国任何一家工厂。
安全数据值得单拎出来看,因为这类指标最难作假。它反映的是管理颗粒度,跟营销预算没关系。
马斯克自己的定性是:这件事极其重要。
Autopilot 从 Mobileye 切换到自研视觉系统,阵痛期快结束了
2017 年是特斯拉辅助驾驶最尴尬的一年。硬件二代(HW2)已经装车,但软件还没跟上,体验一度不如装了 Mobileye 方案的一代车。
马斯克在台上没回避这一点。他说从 Mobileye 的视觉方案切换到特斯拉内部的视觉系统,是一段非常艰难的路,但“我们差不多要到岸了”。
他把硬件配置重新说了一遍:8 个摄像头、12 个超声波传感器、一个雷达、一套很好的 GPS 和惯性测量单元(IMU)。他用的说法是,二代硬件的车已经装好了实现完全自动驾驶所需要的全部硬件,剩下的事只是不断升级软件。
时间点他给得很具体:下一个软件版本也许下周就推送,届时体验会第一次超过一代硬件的车,之后进步会非常快。
他还重申了那个著名的目标路线:
“目标没有变:从加州的一个停车场出发,一路开到纽约的一个停车场,全程不碰任何一个控制件。”(马斯克)
当时很多人把这话当成营销话术。放回 2017 年的语境里看,它描述的其实是工程层面的验收条件:在一个具体路线上,在具体条件下,达成零接管的端到端行驶。“我们的自动驾驶很好”是形容词,无从验证;一条路线跑通与否,可以对赌。
超级充电站要开始提供配套服务了
充电网络的规划,马斯克讲了三层。
第一层是覆盖密度。他的原话是:让人们在任何时候、想去任何地方,都能方便地到达。超级充电桩的布局逻辑沿公路线铺开,覆盖的是城际通行,跟城市里的充电桩是两套东西。
第二层是充电速度。持续升级超级充电系统的能力,让充电更快。
第三层是体验。他说会开始在超级充电站增加配套设施,先从那些大型站点开始。
“目标是让你开特斯拉跑一趟长途的时候,会喜欢上这段经历。”(马斯克)
这里的产品逻辑值得多想一层。燃油车加油是纯粹的等待,五分钟,枯燥,你只想快点结束。充电至少要十几二十分钟,这段时间不可能被消除,那就只能被利用。咖啡馆、便利店、休息区在这些站点里承担的功能,是把“被迫等待”转化成“主动停留”。
充电时间长,从缺点变成了可以设计的空间。
半挂卡车:客户已经在参与设计,他们只问两个问题
电动重卡是 2017 年最不被看好的方向之一。电池重量、续航、载重、充电时间,每一条都是硬约束。
马斯克的表态很简短:很多人认为做不出长续航的电动重卡,但我们确信这件事能做成。
发布时间定在当年 9 月底,会上会展示一台能跑的原型车。
他接着说,这台车已经给好几家大批量采购重卡的公司看过了。反应几乎一模一样:
“他们的反应都一样:这东西我们很喜欢。然后就只问两个问题,我们能买多少台,什么时候能拿到。”(马斯克)
这些大客户还参与了设计过程。马斯克的说法是,重卡最大的几个买家在帮忙确保车的规格符合他们的需求,所以这台车身上没有什么秘密,他们早就知道自己需要什么,也已经告诉特斯拉了。
剩下的事情就是把产量拉起来。
这个流程反过来看很有信息量。传统车企造商用车,是先造出来再拿去卖。特斯拉的做法是先让客户定义规格,再造。对一个新品类来说,这是降低市场风险的唯一办法,因为重卡买家的需求极其具体,猜错一个参数,整个产品就没有意义。
Model Y 不该是 Model 3 的拉高版
讲到 Model Y 的时候,马斯克主动提了一个批评。
当时有不少声音说,Model Y 应该直接从 Model 3 的平台派生,省时省力。他的回应是:这个思路听起来对,但我们已经在 Model X 上犯过一次这个错误了。
“按一辆 SUV 该有的样子去设计 SUV,按一辆轿车该有的样子去设计轿车。否则你只是在把不合适的东西硬塞进去。”(马斯克)
除了设计逻辑,他还提到了制造层面的改进。Model Y 会采用一批非常重要的制造工艺革新,让整车可以用一种和以往完全不同的方式生产,成本会大幅下降。
他给了一个具体到有点惊人的数字:从 Model 3 到 Model Y,资本支出(capex)可以降低一半。
资本支出减半意味着什么。同样一笔钱可以建两倍的产能,或者同样的产能只需要一半的钱。特斯拉当时正在疯狂扩张、现金流紧绷,这个数字的分量比任何性能参数都重。
时间点上,Model Y 的目标是 2019 年前后上路。他还给了一个判断:Model Y 的需求会超过 Model 3。
这个判断在当时听起来大胆。但 SUV 在全球乘用车市场的占比确实一直在压过轿车,尤其在北美和中国。这一判断的依据是整个品类的销量走势,跟单款车的竞争力关系不大。
工厂选址:同时认真考虑三个地方,但先不透露地点
有股东问接下来建多少工厂、建在哪。
马斯克的回答很谨慎。目前有三个选址在认真考虑,但在确定地点和时间之前,会选择按兵不动。
他用了“keep the powder dry”这个说法,字面意思是保持火药干燥,引申为留着弹药不急着打。
长期目标没变:全世界至少 10 座超级工厂,最多可能到 20 座。
现在下单 Model 3,要等到明年年底之后
有股东问,如果现在还没下 1000 美元订金,要等多久。
马斯克的回答很实在:会等很久。他估计现在下订,大概要等到明年年底之后才能提车。
但他补了一句:每周都还有更多订金进来,所以如果你想要,就赶紧交订金,因为队伍不会变短。
关于产能爬坡,他的态度是宁可慢也不能乱:必须把爬坡做对,保证质量,保证安全。这句话配上前面“一万个零件”的论述,逻辑是自洽的。爬坡阶段出的每一个质量问题,代价都比晚交付几个月高得多。
电池能升级,但什么时候升有讲究
有人问老车能不能换新电池。
马斯克的回答分两层。
技术上可以。特斯拉已经做了一些,未来会覆盖几乎所有车型,并且每年都会让这件事更容易。
但有一个物理约束:最轻的电池包和最重的电池包之间重量差距很大,所以不能从最轻的直接换到最重的,会超出车辆设计的承载范围。
更实用的建议是时机。他的说法是,别急着升级,等现有电池衰减得差不多了再换。如果一台车还很新,换电池包在经济上不划算,这时候更好的选择是把旧车卖掉,换一台新的或者二手的。
二手特斯拉会放到官网显眼位置,但不叫“认证二手车”
最后一个问题关于二手车的流通。
马斯克说,官网的二手特斯拉板块会得到多得多的曝光,尤其是那些车龄四五年、里程数比较高的 Model S,会是一个不错的入口。
然后他顺便吐槽了一个行业惯用词:
“我们不会叫它 pre-owned。那是个假名字,一堆胡扯。它就是二手的。”(马斯克)
这几句话和整场演讲的调子是一致的。去掉包装,直接说事。
这场演讲真正在讲什么
把上面这些内容串起来看,一条线很清晰。
Model X 的错误在于加法。太多功能、太多配置、太多零件、太多供应商,复杂度失控,产能爬坡变成灾难。
Model 3 的答案是减法。两个配置选项,单电机起步,简单的产线。
Model Y 的答案是重构。不派生、不将就,重新设计,把资本支出砍一半,把供应链缩短。
超级工厂的答案是规模。用极致规模换极致成本,再用成本优势换取持续扩张的空间。
半挂卡车的答案是共创。让最大的买家定义规格,把市场风险前置消解。
这五件事放在一起,是同一套方法论的五个侧面:把复杂度当成成本来管理,而非当成卖点来炫耀。
这对做产品的人有直接参考价值。功能堆得越多,配置给得越花,供应链铺得越广,看起来越强大;实际上每一项增加都在悄悄抬高失败的概率,而这个概率在规模化生产时会被放大到致命。
2017 年的特斯拉正卡在 Model 3 产能地狱的门口。马斯克在股东大会上把过去的错误一条条摊开讲,为的就是接下来那套“极度克制”的做法能被接受。要说服所有人认同一个只有两个选项的配置器,最好的办法就是先承认上一个产品错在哪。
- 做新产品时,先把第一版的配置选项砍到最少。问自己:如果只能保留两个可选项,留哪两个。剩下的全部推到第二版。
- 盘一遍你的供应链或依赖清单,找出那些“地球上某个地方出事了就会断”的环节。能用库存缓冲的就缓冲,不能缓冲的就在下一版设计里去掉。
- 把规模当成一种战略资产来算账。如果你的成本结构里规模效应不明显,说明你还没有找到真正的护城河。
- 推新品类之前,先找三到五个最大的潜在买家,把规格交给他们定。猜需求是这个世界上最贵的错误。
- 遇到“衍生一个版本更快”的诱惑时,回想 Model X。派生产品省下的是开发时间,付出的是长期的制造和体验代价。
- 复盘时公开讲清楚自己错在哪,比讲清楚自己对在哪更有用。前者能换来团队对“做减法”的认同。
0:00it's not going to keep growing it is as fast as we can to make sure people can go wherever they want whenever they want with convenience keep upgrading the capability of the supercharger system so it can charge things faster we're going to start adding amenities to the supercharger stations so the really big stations will start adding amenities and just make it so that you when you do a long distance trip in a Tesla you love the experience but that's the overarching goal and then an autopilot
0:40it's definitely a tough tough slog transitioning from the mobile eye vision ship to the Tesla's internal vision system but I think we're almost there in terms of exceeding the ability of the hardware one car so the hardware two cars have all of the hardware necessary for full autonomy so eight cameras twelve of the most advanced ultrasonic sonar sensors a radar a very good GPS
1:08and IMU and everything necessary to go full autonomy it's really about upgrading the software over time so I think with the next release of software which is maybe as soon as next week we will finally exceed the experience of the hardware one cars and then it's going to advance very rapidly from there and a goal remains being able to drive
1:34autonomously from a parking lot in California to park light in New York without touching a control at any point along the long way [Applause] so the the gigafactory pro is going going quite well that that's the think the latest picture of the status obviously be ultimate footprint will be quite a bit bigger than what you see even there and and it's we believe it
2:11will be within a few years have a capacity capacity equal to all other lithium-ion battery factories in the world combined in one building scratch everything out of everything in the US China Europe and Everclear everywhere else this one factory will output more than all of them combined so it's really the sheer scale of this is difficult to appreciate unless you're there in person it is just staggering the enormous that
2:41allows us to achieve high economies of scale so that with height with very high production rates we're maximizing economies of scale everything has to get the lowest cost per kilowatt hour in the world a lot but at the same time have the most advanced batteries so a combination of the the best technology at the lowest cost I think is it just a
3:05very quickly obviously a great position to be in and I think it's a fundamental part of Tesla strategy that there's no one else who's attempting anything as far as we are even attempting anything on the scale and so I think that puts us in a very very strong competitive position to sustain the growth of the company for for several years to come and and then over time there will be several gigafactory zai think eventually you know 10 to 10 or 12 maybe 20 at an
3:45equivalent a lot so it's like a giant machine so we keep refining this and productizing it and then and then building Giga factories around the world I'm poor thing that we're making progress on is factory safety so the so far this year Tesla is 32 percent below the order automotive industry entry rate
4:16and that that trend is better and better with each passing month so I think we're on track to be less than half the entry rate of the automotive industry and by far better than any other US factory which i think is something that's extremely important and then we've got some future products so the semi-truck we're going to unveil at the end of September and I think that's that's very exciting a lot of people don't think you can do a heavy duty long range truck
4:54that's electric but we are confident that this is this can be done so we'll be showing off a working prototype it ended up too long for now the end of September and we've shown it to a number of the organizations that buy heavy duty trucking and they all love it they just want to know how how many can they buy and how soon like we're cool and we're involving them in again [Applause]
5:29and we're going to impose be involved in the design process so the biggest customers of the heavy-duty Tesla semi are helping ensure that is that it is specified to their needs so it's not a mystery they already know that it's going to meet their needs because they help to decide what they've told us what those needs are so it's it's going to really just be a question of scaling volume to make as many as as we can and
6:00then model why I'm really excited about model why it's you know there's been some criticism like we should sort of do it derived from the model 3 platform but I think actually we made a mistake in trying to derive the model X from the Model S platform but it would have been better to just design design an SUV the way an SUV should be designed zionists and the way is span should be designed otherwise you're just trying to shoehorn something in that doesn't make sense also there are a number of I think really major manufacturing improvements that can be done that allow us to pull
6:45the car in a way that the cars that have been built before it does expenditures I think would be substantially less I'm confident that we could drop the capex by a factor of 2 between model 3 and model y which i think is a really big deal and accelerate its readiness despite the new technologies so we're
7:13aiming for that you know to hit the roads in 2019 approximately and probably the demand for the model Y will exceed them the demand for the Model 3 so there's a few other things I haven't mentioned share good here is I just like really recommend showing up for the semi truck unveiling
7:43maybe those will more than we're saying here maybe could be who knows model twe configurators so expect that a configurator to go live towards the end of next month when we deliver the first production model three so yeah we definitely on track to deliver the first production model 3 next month that's going to be real exciting and then we'll have the configurator go live right at that point now I should say that the we've kept the initial configurations from model 3 very simple this is critical to achieving a rapid production ramp big mistake we made with the DX which primarily my responsibility way
8:29too much complexity right at the beginning that was route that was very foolish so you know we think of it like going back to the launch of the Model S one less only had one configuration at start of production and now it had one configuration because that's all we could do number because we were really clever [Laughter] and and then model X of hubris
8:57extraordinaire we who actually we added I mean it is like a Faberge egg of cause I think it's a big it's really an amazing product but it's it has way too many cool things in it that should have really been rolled in with version 2 version 3 that would have been the sensible way to do it we got to get overconfident and created something great that probably will never be made again and perhaps should not be but it is an amazing car and as we keep refining the software and model X it's just going to get better and better so initially the multi configurators it's kind of going to be like what color do you want and what size wheels do you want that's basically going to be the
9:45configurator and then and then we'll show what other configure what other versions are coming later as soon as we get configuration one right but they would a timeline associated with with each configuration for example we were going to start off with dual motor but but that's like wait a second we just doubled the probability of something going wrong if we get two motors because they're two different motor architectures one motor is optimized for highway travel and ones optimized for stop-and-go traffic which is great for maximizing your your mileage in City and
10:24maximizing a highway your mileage on on a freeway and having incredible acceleration but it's too much complexity right off the bat so it'll just be single motor to begin with and then we'll have the dual motor it config for lucky to end of this year or more likely early next so I wouldn't worry too much about the configurator because you only have to decide like what color do you want and what size really it for the initial production in terms of how
10:53many factories are in the works we're really giving serious consideration to three three factory locations right now but we're going to try to hold our powder dry keep the powder dry until we're confident of of the locations and the timing but like I said ultimately probably does at least ten of these worldwide and maybe maybe as much as twenty
11:31how long ago it will be four customers running model 2 and haven't placed a 1k deposit this will be a long wait there's a lot of people that have ordered the car I'm guessing if you put a deposit down on model three now it's probably well it's gonna be over a year end of next year before you get it something like that on the other hand there are
12:01more and more deposits every week so if you want it then definitely put down the deposit line isn't getting shorter doing our best to do this ramp but we've got to do the ramp right we're going to make sure the quality is good the safety is good and it's crazy hard to make cars I
12:27tell you this is like 10,000 unique items and it will move as fast as the slowest item so and then even beyond the stuff that's internal Tesla if you say like okay who's the least lucky supplier we have out there or what supplier cares the least or whatever in case maybe and then you look at our supply chain it's like wow our supply chain is like covering earth so like what are the odds that there's going to be some portions your event somewhere on earth like pretty.i so either one of things I want
12:59to be with model wise also just simplify the supply chain so that we're not inheriting force majeure risk from earth because earth is big and something like a swinging bad happens on earth at any given point in time so we must stop inheriting fortunate zero risk from from all of earth you can sort even help you can solve this by buffering parts but if the parts aren't made to begin with you can't bother them so yeah other times
13:30place to offer battery upgrades yeah so we do after that already I wouldn't recommend doing a battery upgrade until the back the existing batteries you have has a fair bit of life on it and the awesome limitation is because there can be a pretty big weight difference between the lightest battery pack and the heaviest so we can't go from lightest to heaviest but we definitely can upgrade battery packs
13:59really in every car and we will we will offer that and make that easier with each passing year there's always got a new battery pack it's not going to make financial sense to upgrade that just the battery pack better to sell the colleague that you've gotten and then and then buy a different one either new or used and I should mention we're going to really give more prominence to use Tesla's on our website but we're not going to call it pre-owned JIT that is like a bogus name of several BS it was used okay every one of that we're we fill in here so the use Tesla section of the
14:46website is it's going to get a lot more attention and particularly if the car is four years old and it's got a lot of mileage you can buy a Model S
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