近地轨道上有九千颗星链(Starlink)卫星。这几年它们做过无数次轨道调整,避让太空垃圾、避让交叉的轨道、避让别国的航天器。但没有一次,是为了躲开一艘外星飞船。
这是马斯克 2026 年 2 月在达沃斯给出的证据。他说这话时语气很轻,结论一点也不轻。
如果有人能知道我们中间有没有外星人,那应该是我。(马斯克)
后半句他自己先笑了:因为"我就是那个外星人,只是他们不信"。
这场对话发生在世界经济论坛的会场里,对谈的另一方是贝莱德(BlackRock)CEO 拉里·芬克(Larry Fink)。上台前两人还在开玩笑,芬克问这场能产出几个金句,马斯克说五个。开场马斯克还拿当天的和平协议打岔,问是不是在分格陵兰岛的一块、委内瑞拉的一块,"我们只要和平"。
玩笑归玩笑,接下来的四十分钟,马斯克把 AI、机器人、能源、太空、衰老这几件事串成了一条线。串起来看,逻辑比他任何一场发布会都清楚。
达沃斯的第一个数字
芬克开场用了两个数字。
贝莱德自上市以来给股东的复合年化回报是 21%。特斯拉自 IPO 以来的数字是 43%。
"这是给欧洲人的一则广告,"芬克说,"更多公民应该投资增长,投资自己的国家。"他补了一句更有杀伤力的:如果当年有一批养老基金在特斯拉上市时买入,今天会是什么光景。
这话放在达沃斯的会场里有点扎人。欧洲的养老金体系以保守著称,大量资金压在债券和成熟资产上。芬克用两个复利数字说明一件事:错过增长型公司的代价,二十年后会以另一种方式出现在老龄化的账单上。
马斯克的回应短得近乎敷衍:特斯拉有一支了不起的团队,复利来自这里。他把功劳全部推了出去。
然后是正题。芬克把话题拉回技术本身:AI、机器人、能源、太空,这些方向最后都归结到工程纪律、规模化和执行力。真正同时在几条线上做执行的,全世界找不出第二个。他想知道,从工程角度看,这些事有什么共同点。
一句使命,管着五家公司
马斯克给这些公司找了一个公共的分母。
我所有公司的总目标,是最大化文明的未来,也就是最大化文明拥有美好未来的概率,并且把意识扩展到地球之外。(马斯克)
这句话拆开是两件事。第一件是保住现有的,第二件是做备份。
SpaceX 负责第二件。火箭技术的终点,是让生命和意识延伸到月球、火星,最终延伸到其他恒星系。
为什么要备份,他用了一个我很喜欢的比喻。
我脑子里的画面,是广阔黑暗里的一根小蜡烛。意识的小蜡烛,它很容易熄灭。(马斯克)
这个判断建立在一个前提上:就我们目前所知,宇宙里没有第二处生命。九千颗卫星在轨运行,没有撞上过任何非人造物体。马斯克说,既然是这样,就该假设生命和意识极其稀有,可能只有我们这一份。
费米悖论问了七十多年,答案无非两种。要么别处有生命,要么没有。马斯克的立场是:在证据出现之前,按"只有我们"来做工程决策。因为如果只有我们,熄灭就是永久的。
火星计划从这个角度看,不是浪漫主义,是风险管理。自然灾害、人造灾难、一场足够大的战争,任何一种都可能让这根蜡烛灭掉。让生命成为多行星的,是唯一能把风险分散到地球之外的办法。
特斯拉的使命多了后半句
特斯拉原来的使命是加速世界向可持续能源转型。马斯克在达沃斯说,现在这个使命加上了后半句。
特斯拉显然是关于可持续技术的。到现在,我们在使命里加上了"可持续的丰裕"(sustainable abundance),靠的是机器人和 AI。这才是通往全民丰裕的道路。(马斯克)
加了这四个字,公司的边界就变了。能源公司解决的是"怎么少烧一点",丰裕解决的是"怎么让所有人都过得很好"。前者是效率问题,后者是产能问题。
他把这个目标说得很绝对。
人们常说要解决全球贫困,要让每个人都有很高的生活水平。我认为唯一的办法是 AI 和机器人。(马斯克)
听上去像技术乐观主义的口号。背后有一道算术:八十亿人如果都达到发达国家的生活水平,按现有技术路径需要的资源和人力是算不平的。稀缺的劳动力、有限的能源、昂贵的护理,这些缺口不会靠再分配消失,只能靠供给侧的产能跃升。
他自己也清楚这话的另一面。
这不意味着没有问题。我们对 AI 要非常小心,对机器人也要非常小心。我们不想发现自己活在詹姆斯·卡梅隆的电影里。(马斯克)
他补了一句"我喜欢他的电影",接着说"但我们不想成为终结者"。这个警告他在几乎每场公开对话里都会重复。重复次数多了容易被当成客套话,但结合他后面给出的机器人数量预测,这句话的分量完全不同。他一边在预测机器人会比人多,一边在提醒这东西可能失控,两个判断之间没有任何缓冲。
经济产出的公式
芬克问了一个机制问题:这种增长会惠及所有人,还是只落在少数人手里?
马斯克的回答可以写成一个乘法式。
如果你有大量人形机器人,经济产出就等于每台机器人的平均生产力,乘以机器人的数量。(马斯克)
这个公式朴素得近乎粗暴,但它把讨论从"分配"拉回了"产能"。人均生产力受生物学限制,一个成年人再怎么训练,产出都有天花板。机器人数量没有这个天花板,只要工厂能造、矿能挖、电够用,数字可以一直往上加。
他给出的预测相当激进。
在未来的良性情景里,我们会造出如此多的机器人和 AI,以至于它们会饱和人类所有的需求。到某个时刻,你甚至想不出还能向机器人要求什么。(马斯克)
然后是更直接的一句:机器人的数量会超过人的数量。
再补一层:地球上每个人都会有一台,而且都会想要一台。这个"想要"很重要,它指向的不是被分配,是被需要。
他举的例子很生活。看孩子、照顾宠物、照看年迈的父母。
我很多朋友都说,他们有年迈的父母,照顾起来非常困难。(马斯克)
芬克接了一句"而且很贵"。马斯克同意,然后说出了问题的实质。
对,很贵。而且根本没有足够的人手来照顾老人,没有足够的年轻人去照顾老年人。(马斯克)
这句话值得停一下。护理是劳动密集型行业,老龄化社会的护理缺口没有任何一种再分配方案能填上,因为缺的是人手,不是预算。养老金可以印,护工不能印。
"如果有一台机器人能照顾和保护你的父母,那会是很棒的东西。"他说这话的时候,语气比谈火星的时候软得多。
有人必须做的工作,就没有全民丰裕
芬克没有放过这个逻辑的另一头:如果机器人把活都干了,人的意义在哪?
马斯克的回答比通常那句"人可以去追求艺术"要硬得多。
没有什么是完美的。但你不能两者兼得。你不能同时拥有必须有人做的工作,和全民的惊人丰裕。(马斯克)
这句话是整个对话里最锋利的一刀。
它的意思是:只要还有"必须有人做"的活,丰裕就是窄的。因为必须做的活意味着门槛,门槛意味着只有一部分人能进入高产出部门,剩下的人被锁在低产出部门。芬克追问"那就是窄的?"马斯克答"对,正是"。
反过来,如果活由机器人干,产出就不依赖人的准入资格,丰裕才可能铺开。
这个论证绕开了一个常见的争论。人们习惯问"机器人会不会抢走工作",马斯克把问题换了方向:如果活必须由人干,那丰裕注定只能覆盖一部分人。工作保护和普遍丰裕,在逻辑上没法同时成立。你可以选择保住岗位,也可以选择让所有人都不用为生计工作,但不能两个都要。
至于人的意义怎么办,他承认这不是工程能解决的问题。"没有什么是完美的。"这是全场他最少见的一次让步,也是他没有答案的地方。
三十五万亿个细胞在用同一个时钟
芬克问了个私人问题:我们能不能在这辈子看到衰老被逆转?
马斯克先说自己没在抗衰老上花多少时间,然后给了一个判断:这是个可解的问题,而且答案会非常明显。
他的推理是这样。
你身体里所有的细胞,基本上以同样的速度老化。我这辈子没见过谁左胳膊老、右胳膊年轻。(马斯克)
既然三十五万亿个细胞同步老化,那一定存在一个同步时钟在协调它们。
这意味着一定存在一个时钟,一个在你体内三十五万亿个细胞之间同步的时钟。(马斯克)
找到这个时钟,就找到了干预的入口。"当我们搞清楚衰老的原因,会发现它明显得不可思议。这不是一个微妙的问题。"
这段话的信息量在方法论上:先找一个全局同步机制,而不是逐个细胞去修。生物医学界这几年的方向也压在这个判断上,表观遗传时钟、系统性炎症信号、血液里的循环因子,全都是"同步器"的候选。如果衰老真是由一个统一信号驱动的,那它就是一个可以被单点打断的过程,而不是三十五万亿次独立的磨损。
然后他说了一句不太符合期待的话。
死亡有它的好处。如果人真的能活很久很久,社会有僵化的风险,事情会被锁死在原地,失去活力。(马斯克)
文字稿这里把 ossification 识别成了 oification,按上下文应该是"僵化、骨化"的意思。
马斯克一边在造能飞到火星的火箭,一边承认永生可能让文明停止更新。这个矛盾他没打算解决,只是把它说出来了。一个把"扩展意识"当作最高目标的人,同时承认意识需要有限的生命来保持活力,这种自我冲突在他身上不常见。
最后的判断留了余地,但方向明确:延长寿命,甚至逆转衰老,"可能性很大"。
AI 的价格一个月一个价
话题转到瓶颈。芬克列了一串:算力、芯片、晶圆厂、供电,每一项都是巨大的资本开支。
马斯克先回答了"会不会变窄"。答案是不会,理由很市场:AI 公司会去找尽可能多的客户,这是它们的本能。而 AI 的成本已经很低,还在往下掉。
AI 的成本几乎每个月都在发生有意义的变化。(马斯克)
"每个月"这个频率不太寻常。多数技术成本的下降曲线是按年看的,按月变意味着这一层的竞争已经白热化到价格随时被重写的程度。
他顺手提了开源模型的处境。
开源模型大概只落后闭源模型一年。(马斯克)
这个"一年"值得琢磨。它同时说明两件事。
第一,任何闭源模型的能力优势都有大约十二个月的保鲜期。今天最强的东西,一年后会在开源侧以接近免费的价格出现。
第二,价格战的底线由开源侧决定。开源模型的推理成本是公开的、可复制的,任何人都能拿去压价,闭源厂商的定价权因此被牢牢按住。
一年在技术史上很短,在商业竞争里很长。对做应用的人来说,这个时间差就是你选技术栈的窗口:押注某家闭源模型持续领先一年以上,等于押注它的护城河比开源社区快。
卡住 AI 的不是芯片,是电
然后是这场对话里最实在的一句判断。
AI 部署的限制因素,从根本上说是电力。(马斯克)
不是 GPU,不是先进制程,不是封装产能。是电。
这句话放在 2026 年说出来已经不算惊人之语,但它经得起反复验证。数据中心是电力密集型的,而且要求稳定、连续、便宜。训练集群的功率密度一代比一代高,电网扩容的速度却按年计,大型变压器的交付周期一度拉到两年以上。芯片可以加班造,电网不能加班建。
文字稿在这里中断了,后面的部分应该是关于太阳能的具体讨论。这场对话的标题已经把那层意思挑明:中国领跑,一千吉瓦的光伏推进计划。
一千吉瓦这个数字
一千吉瓦是什么概念。
中国 2025 年的光伏累计装机已经跨过了这个门槛,相当于几十个三峡电站的装机容量,而且是集中式和分布式一起铺开的。更关键的是产业链:硅料、硅片、电池片、组件四个主要环节,中国企业的全球份额都在八成以上。
把这两组事实和马斯克那句"限制因素是电力"接起来,逻辑就闭合了。如果 AI 的瓶颈是电,那么电的瓶颈就是光伏的产能和装机速度。谁在这条链上占住份额,谁就在 AI 的基础设施上握着一张牌。
光伏的另一个特性经常被忽略:它的建设周期短。一座地面电站从开工到并网可以压缩到几个月,而一座核电站是十几年,一条特高压线路是五到八年。当电力需求突然出现指数级增长时,能在两三年内补上缺口的只有光伏和储能这一组。这解释了为什么 AI 算力的扩张速度,最终会被光伏组件的出货量牵引。
马斯克本人对太阳能的态度一贯明确。他早年就讲过太阳能加电池是终局答案,特斯拉的能源业务这些年也一直往这个方向铺。这场对话的标题把中国的光伏规模摆在前面,指向的正是他刚刚说的那句话:限制因素是电力。
对做 AI 的人来说,这条传导链很短:模型能力的上限由算力定,算力的上限由电力定,电力的上限由光伏和储能的产能定。任何一环卡住,前面所有乐观的预测都要打折。
整场对话里马斯克反复用的词是"丰裕"。他没说的是,丰裕不是天上掉下来的。它需要电、需要矿、需要工厂、需要一个能把东西大规模造出来的工业体系。他讲的是终局,中国在做的是中间那段。
回到开场那根蜡烛。马斯克说它很小,很容易熄灭。他的解法是把它复制一份放到另一颗行星上。而让这根蜡烛烧得更亮的办法,是给它接上足够多的电。
两件事他都在做,只是第二件比第一件更快见效。
- 把电力当成 AI 战略的一等变量。评估一个算力项目时,先问三件事:电从哪来、能不能稳定供五年、扩容路径是什么。芯片的供应会在两年内缓解,电网的约束会持续十年。
- 算清楚开源和闭源的一年时间差。不要在架构里把某一家闭源模型写死。设计上假设模型层每年会被替换一次,把切换成本压到最低,这比追最新榜单更值钱。
- 跟踪人形机器人的单台成本曲线,而不是演示视频。马斯克的经济公式里,机器人数量是乘数。乘数什么时候起飞,取决于单台成本什么时候跌破一年的护工工资。这个拐点比任何一次发布会都重要。
- 护理是机器人最先落地的场景。需求刚性、支付方明确、任务边界清晰、失败代价可控。要做机器人应用,从养老和托育切进去,比从通用家务切进去容易十倍。
- 读一读衰老的"同步时钟"假说。表观遗传时钟、循环因子、系统性炎症,这几个方向的论文值得花一个周末。马斯克的判断未必对,但他提的问题是对的:三十五万亿个细胞为什么同步老化。
- 在能源侧看光伏和储能的出货量,而不是装机公告。组件出货和储能招标是领先指标,装机统计是滞后指标。AI 数据中心的电力缺口,会先出现在这两张表上。
- 对"机器人饱和人类需求"这类预测保持怀疑,但别无视它。把十年期的职业规划做一次压力测试:如果你的工作产出的价值可以用"每台机器人平均生产力乘以数量"来描述,那它就在被替代的射程内。
0:00A large applause. Start again. [applause and cheering]
0:09That's better. Thank you.
0:11Yeah. We're going to make this interesting. [laughter] How many how many quotes are you going to want that are going to be after this session?
0:22I don't know. I mean, five.
0:25Okay. [laughter] So, good afternoon everyone. It's great to see everybody here. it's been an amazing week here in Davos. hopefully everybody saw that we are having conversations here. Hopefully everybody agrees. There are some conversations that we may disagreed. There's many conversations we may have agreed but through those conversations and I think today's result with a peace agreement earlier today the world economic forum is here to have those conversations to have understandings and also resolution. So
1:01it's an important component of who we and what we are and I'm thrilled to have Elon Musk here. he came all the way from California to be here to see all of you. So, thank you, Elon. you're most welcome.
1:20I mean, I heard I heard about heard about the formation of the the peace summit and I was like, is that is that PIC? [laughter] you know, little piece of Greenland, a little piece of Venezuela. [laughter] We we got one. All we want is peace.
1:42Okay. I'm going to as I said, I'm a pretty proud CEO of Black Rockck since we went public. the compounding return of Black Rockck to our shareholders with 21%. since Elon took Tesla public, his compounded return is 43%. This is I just another advertisement for everybody, especially for Europeans.
2:08This is why more citizens should be investing with growth, investing with their countries. Imagine if a lot of pension funds invested with Elon when Tesla went public and how much return with the all the pension funds that invested side by side with Elon and the growth. So, a spectacular return. There's very few companies, well I don't think there's any other company as large as Tesla today that has that compounded return. So congratulations. It's a good measurement.
2:43Well, we have an incredible team at Tesla and that's the reason. So I want to get into the dirt the the meaningful component about technology the possibilities I want to talk about AI and robotics energy space and the progress ultimately coming down to engineering engineering discipline scale execution and few few people if not anyone has the experience and the fortitude to confront these issues head on not just the idea is but the execution across so many different technologies Elon and that's why I thought it was important for us to have this dialogue here in Davos so [clears throat] you're presently building on AI on robotics on space on energy all at the same time
3:34when you look across those efforts what do they have in common from an engineering standpoint
3:42well they're all very difficult technology challenges but the the overall goal of my companies is to maximize the future of civilization like basically maximize the probability that civilization has a great future and to expand consciousness beyond earth. So take SpaceX for example that SpaceX is about build advancing rocket technology to the point where we can extend life and consciousness beyond Earth to the moon to Mars eventually to other star systems and
4:20and I think we should always view consciousness life as we know it as as precarious and delicate because to the best of our knowledge we we we don't know of life anywhere else. You know, I'm often asked are there aliens among us? And I'll say that I am one. But
4:42or you're from the future. They don't believe me. Okay.
4:45so, but I I I [clears throat] think if anyone would know if there are aliens among us, it would be me. and we we have 9,000 satellites up there, and not once have we had to maneuver around an an alien spaceship. So I'm like I don't know. It's bottom line is I think we need to assume that life and consciousness is extremely rare and it might [clears throat] only be us. And if that's the case, then we need to do everything possible to to ensure that the [clears throat] the light of can the light the light of consciousness is not extinguished because we're effectively the way I view it is the image in my mind is of a a tiny candle in a vast
5:33darkness tiny candle of consciousness that could easily go out. and that's why it's important to make life multilanetary such that if there is a natural disaster or a man-made disaster on earth that consciousness continues. That's the purpose of SpaceX. te Tesla is obviously about sustainable technology and and and also at this
5:59point we've we've sort of added to our mission sustainable abundance. though with robotics and AI this this is really the path to abundance for all. If you say you know people often talk about solving global poverty or essentially how how do we make give everyone a very high standard of living? I I think the only way to do this is AI and robotics.
6:30which which doesn't mean that it is without its issues. I mean there we need to be very careful with AI. We need to be very careful with robotics. We don't want to find ourselves in a James Cameron movie. you know Terminator he's great movies. Love his movies but but we don't want to be in Terminator obviously. but but if you have
6:56ubiquitous AI that is essentially free or close to it and ubiquitous robotics then you will have an an explosion in the in the global economy an expansion in the global economy that is truly beyond all precedent.
7:14Elen, can that expansion be broad? Yes.
7:18Or is it narrow? And how can that be created? How can it broaden the global economy?
7:25Yeah, it's I mean [sighs and gasps] I mean a way to think of it is that if you have a large number of humanoid robots the economic output is the average productivity per robot times the number of robots,
7:41right? and and actually my prediction is in the in the benign scenario of the future that we will the robots we will actually make so many robots in AI that they will actually saturate all human needs. Meaning you won't be able to even think of something to ask the robot for at a certain point like like there will be such an abundance of goods and services because the my prediction is that there'll be there'll be more robots than people. So
8:13but how do you then have human purpose in that scenario? Yeah, I mean, you know, there are nothing nothing's perfect, you know, but but I mean I mean it it is a a necessary like you can't have both. You can't have work that has to be done and amazing abundance for all. because if it's if it's work that has to be done then then then you and and only some people can do it then then you you you can't have abundance then it's narrow.
8:48It's narrow. Exactly. So, but if you if you have billions of humanoid robots, I think there will be, I think I think everyone on Earth is going to have one and going to want one. because, you, who wouldn't want a robot to, you know, assuming it's very safe, watch over your kids, take care of your pet. if you have elderly parents, a lot of friends of mine said they have elderly parents.
9:18It's it's very difficult to take care of them. Expensive.
9:20Yeah, it's expensive and it's expensive and there just aren't enough people to take care of the there aren't enough young people to take care of the old people, right?
9:28so if you if they if if you had a robot that could take care of and protect an an elderly parent, I think that would be great. That would be an amazing thing to have. and and that I think we will have those things. So I mean overall I'm I'm very optimistic about the future. I think we're headed for a future of amazing abundance.
9:52which is very cool. and and and definitely we [clears throat] are in the most interesting time in history. I think there's more a more interesting time in history. Can we can you and I reverse aging in this new history or or or are we going to see it?
10:13You know, I haven't I haven't put much time into the aging stuff. I I do think it is a very solvable problem like like you can I think when when we figure out what causes aging,
10:27I think we'll find it's incredibly obvious. It's not a subtle thing. the reason I say it's not a solid thing is because all the cells in your body, you know, with some pretty much age at the same rate. I've never seen someone with with an old left arm and a young right arm ever in my life.
10:46so why is that? It that means that there must be a clock, a synchronizing clock, right,
10:51that is synchronizing across 35 trillion cells in your body. and you know the there is some benefit to death by the way. It's like there's there's a reason why we don't actually have a longer lifespan. because if you if you have if people do live forever for very long time I think there's some risk of an oification of society of of things just getting kind of locked in place. and
11:27you know it it just may become stalifying just not lack lack vibrancy. but but that that said do I think we'll figure out ways to extend life and and maybe even reverse aging? I think that's highly likely.
11:48I'm looking forward to that.
11:51Yeah. So in the future that you talk about AI models, autonomous machines, rockets depends on massive increases of compute, massive increases in energy, expensive energy, manufacturing scale, what are the bottlenecks to to get there? And then once again with all that expenditures again, how can we make sure that it's broadened, not narrow?
12:23[clears throat] I I just think the natural thing is it's going to be very broad because AI companies will seek as many customers as they possibly can and the cost of AI will get is already very low and it's is plummeting every year. I mean you almost the cost of AI is almost ch meaningfully changing on a month-to-month basis.
12:46There's open there's open models now everywhere. Yeah.
12:49Yes. very there's open models [clears throat] and the open models only lag they're maybe a year behind right the the private the sort of closed models
12:59so so I I think the the ai companies will seek as as many customers as possible which means they'll seek they'll provide AI to the world
13:09but the cost of getting to there the compute the chips the fab the powering that to me what are the what are the B you know those are huge
13:23the limiting factor yeah I think the limiting factor for AI deployment is fundamentally electrical power it's This
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