发布会·中文精读

马斯克谈 Starlink 增长:50 亿至 100 亿美元投资(MWC 2021 演讲)

Elon Musk talks Starlink Growth, $5-10 Billion Investment! (MWC 2021 Presentation)

中文精读由 AI 基于原始视频与文稿改写,非官方翻译,可能存在偏差或错误;表述以原视频为准。说明

2021 年 6 月,Starlink 的活跃用户数跨过了一个「具有战略意义的数字」:69,420。

马斯克在 MWC 的远程连线里把这个数字念了出来,语气平静,像在汇报一个普通的工程指标。会议室里的人应该都听懂了。一个造火箭的人,用这种方式给一家卫星宽带公司报用户数。

同一场对话里,他回忆起几年前第一次公开谈 Starlink 时被问到的那个问题:你的目标是什么。他当时的回答是,我们的目标是不破产。

把这两句话放在一起看,这场访谈的骨架就出来了。一边是 1500 颗在轨卫星、5 兆瓦太阳能、每秒 30TB 量级的数据吞吐;另一边是四十年来所有低轨通信星座无一例外的破产记录。马斯克在 MWC 上讲的,其实是同一件事的两面:一个技术上极其激进的系统,怎么用最保守的财务纪律活下来。

星链填的是 3% 到 5% 的缝,从来没打算取代 5G

主持人开场就把 Starlink 定性了:一个给全世界提供宽带覆盖的卫星计划,马斯克自己说过这相当于在太空重建互联网。

马斯克的回答很快把范围收窄了。他说有些地方现在根本没有连接,有些地方连接非常有限或者非常贵,还有些地方根本不可能铺设。Starlink 的定位是填补 5G 和光纤之间的空隙,真正到达那些最难覆盖的区域,大概是 3%,也可能是 5% 的人口。

这个比例低得有点反直觉。一家公司投入上百亿美元,只为了地球上 3% 到 5% 的人,商业上怎么可能成立。答案藏在后面那段话里:Starlink 的点波束尺寸按蜂窝通信的说法算是相当大的,所以它适合低密度到中等密度地区,高密度地区只会服务数量有限的客户。

翻译过来就是,城市里它没优势,也不打算有优势。人口密度决定了地面网络的经济性,一公里光纤铺过去接一百户人家和接一户人家,单户成本差着两个数量级。Starlink 反过来,卫星的覆盖范围是固定的,用户密度越低,单位成本摊得越开。

它做的是一门地面网络算不过账的生意,而且只做这一门。

这个边界划得很清楚。马斯克反复用「补充」这个词,说 Starlink 和光纤、5G 是很好的互补技术。他没说要做下一代互联网基础设施,也没说要颠覆运营商。

还有一层生意他没细讲,但提了。很多运营商在偏远地区有蜂窝基站,这些基站回传数据用的是昂贵的老方案。用 Starlink 做回传(data backhaul),成本可能低得多。他没有点名任何合作伙伴,只说对方还没准备好公布。这是典型的 B2B 思路:先把面向消费者的 3% 做着,顺手把运营商的回传生意也接过来。

从生意的角度看,这两条腿的分量并不一样。消费市场要教育用户,要让普通人相信一个屋顶上的圆盘能替代宽带,还要面对各国监管和本地运营商的敌意。运营商回传市场不需要教育,采购方是工程师,他们只看每兆比特的成本和可用性。星链早期真正能快速起量的收入,很可能来自后者。

1500 颗卫星、5 兆瓦太阳能、每秒 30TB

主持人请马斯克给个进展概览。他报了一串数字,语气像在读体检报告。

已经在轨并处于活跃状态的卫星超过 1500 颗。所有卫星加起来的太阳能功率超过 5 兆瓦。整个星座的数据输出能力大约每秒 30TB。

第三个数字在文字稿里写的是 30 terabytes per second,按行业惯例这指的是 30 Tbps 量级的星座总吞吐,语音转写大概率把 Tbps 记成了 TB/s。这两个单位差着 8 倍,看这类材料得留个心眼。

覆盖时间表给得很具体:2021 年 8 月开始,除了南北极,全球都能连上。他在现场还纠正了一下自己,先说「几乎就是下个月」,然后改口成「应该说是 8 月」。

运营范围是大约 12 个国家,每个月还在增加。用户数就是开头那个 69,420,他预计 12 个月内能到几十万,甚至超过 50 万。

这些数字放在一起可以看出节奏。1500 颗卫星对应 7 万用户,平均一颗卫星服务不到 50 个人。这个比例在任何通信网络里都难看得要命。但马斯克要的就是这个阶段:先把星座铺好,把覆盖打通,用户数可以后来追,轨道位置和频率资源不行。

他还提到一件容易被忽略的事,SpaceX 在持续改进用户终端、卫星、地面站、网关和接入点(point of presence)。这五样东西同时在迭代。多数卫星公司的商业模型是造一批卫星,打上去,用十五年,中途换一代。SpaceX 的做法接近软件公司,硬件版本号一直在往上走。

这种迭代节奏有一个隐含前提:每一次改动都必须能装进下一批卫星,而不是等到下一颗大卫星。这意味着团队得接受大量在轨设备是「不完全体」,也意味着地面段必须能同时兼容好几个硬件版本。工程管理的难度在这儿被放大了一档。

500 公里对 36000 公里,延迟的上限由光速锁定

主持人问的是技术差异,马斯克给了一个所有人都能听懂的解释。

Starlink 的轨道高度大约 500 公里,同步轨道卫星(GEO)在 36000 公里。这个差距直接决定了延迟。

光速是每秒 30 万公里。信号从地面到 36000 公里外的同步卫星,再回到地面,单跳往返就是 72000 公里,光走完这段路要 240 毫秒。实际使用中还要加上地面段处理和路由,用户感受到的是 500 到 600 毫秒。这个数字意味着视频通话会互相打断,在线游戏基本没法玩。

500 公里的低轨,单跳往返 1000 公里,光走完只要 3 毫秒多一点。剩下的延迟来自处理、路由和排队,这部分可以优化。

马斯克说 Starlink 的目标是把延迟压到 20 毫秒以下,感觉上就是没有卡顿,可以打竞技类视频游戏。

轨道高度一选定,商业模型的大半就定了。同步轨道的省钱之处在卫星数量,三颗就能覆盖大半个地球,可延迟永远救不回来。低轨必须用几千颗卫星组成星座才能连续覆盖,代价是发射成本和地面运维,换来的是接近光纤的体验。这是两种完全不同的生意,并非同一件事的两个版本。

马斯克自己把话说得很直:低轨的延迟和地面光纤、5G 相当。

有个数字对照很有意思。20 毫秒的延迟,大致相当于地面网络跨一个大洲的往返时间。光纤在玻璃里的传播速度约为真空光速的三分之二,绕地球半圈实际走的距离远大于卫星的直线路径。未来加上激光星间链路之后,长距离传输的延迟甚至可能低于地面光纤。这件事在 2021 年还只是理论上的优势,但它给星链留出了一个天花板很高的想象空间。

低轨通信星座的历史是一部破产史

主持人没有放过这个问题。过去几十年,卫星宽带的尝试失败了一大堆,九十年代那些响当当的名字今天都不在了。现在亚马逊的 Kuiper 和 OneWeb 也在做同样的事,你凭什么不一样。

马斯克的回答从一句自嘲开始。

几年前我讲 Starlink 的时候,有人问我目标是什么。我说我们的目标是不破产。因为正如你指出的,有史以来每一个低轨通信星座都破产了。(马斯克)

他接着把这段历史讲得更细。有些星座是在部署完成之后破产的,更多的连星座都没铺完就破产了。有几家后来从破产里爬了出来,原始股东却没有从中受益。铱星(Iridium)和 Orbcomm 今天活得还算不错,受益的是破产重组之后的买家。

这段回答里有个细节值得琢磨:主持人提到了亚马逊和 OneWeb,马斯克一次都没有接。他没有评价对手的技术路线,没有比较卫星数量,也没有谈频率竞争。他只是把话题拉回自己的清单:第一件事是别破产。

这个反应背后有一套清晰的优先级。在一个所有先行者都死掉过的行业里,最大的风险从来都是自己先烧完钱。OneWeb 在 2020 年 3 月破产重组,这个时间点距离这场访谈只有一年多。马斯克不需要点名。

他还给了一个隐含的成本结构判断:这些星座失败的原因不完全是需求不存在,单位成本压不下来才是主线。卫星要造,要发,要在地面建网关,要养用户终端。每一项都是重资产前置投入。谁能把这四项里的任何一项成本砍掉一半,谁就能活到别人死掉之后。

把时间线排一下会更清楚。九十年代的铱星计划,66 颗卫星,单颗造价几千万美元,终端售价三千美元,通话费每分钟几美元。需求方是商务差旅人士,结果蜂窝网络迅速铺开,把这批人全抢走了。这是一场时间差上的失败:产品造完的时候,市场已经被另一种技术吃掉了。

星链要避免的是同一种失败。它给出的解法是让成本下降得比市场被抢走的速度更快。猎鹰 9 复用把发射成本压到每公斤几千美元,卫星批量生产把单颗成本压到几十万美元量级,用户终端从上千美元往几百美元走。三条曲线同时向下,才有活路。

航天业的惯例是把落后一代的技术送上天

这是整场访谈里信息密度最高的一段,也最能解释 Starlink 为什么可能活下来。

马斯克说,航天领域的惯例很讽刺:上天的技术往往比地面上的技术更老。

原因在于决策逻辑。公司会说,我们要用在地面上已经被证明可行的技术,然后再花四年时间测试它在真空、高辐射环境下能不能工作。等这一圈验证走完,这项技术在地面上已经过时了。

SpaceX 走了反方向。马斯克的原话是他们采取了一种冒险的做法:要做的技术在某种程度上比地面上的还先进,然后赌一把。

赌出来的东西是什么?他用的词是迄今为止最复杂、最先进的相控阵(phased array)技术,而且按他所知,这个水平的相控阵系统连军用级别都没有达到过。

这句话可以两听。一半是营销,一半是事实。Starlink 的用户终端是一片扁平的相控阵天线,成本被压到了几百美元的量级,这在五年前不可想象。军用相控阵雷达的价格是它的几十倍上百倍,因为军用的产量是几十套,Starlink 的产量是几十万套。产量本身就是一种技术。

马斯克还讲了这个选择为什么对低轨星座特别重要:低轨卫星在天上跑得飞快,90 分钟左右绕地球一圈,用户头顶的卫星几分钟就换一颗。如果天线是机械转动的,切换过程会有明显的停顿;如果天线是相控阵,切换可以在微秒级完成。

这里有一个被很多人忽略的工程细节。机械转动的天线有一个物理惯性,转到位需要时间,误差需要校准,长期使用还会磨损。相控阵没有活动部件,波束靠电磁波的相位差合成,指向由算法决定。可靠性上的差别在需要连续运行十几年的设备上会被放大成真金白银。

相控阵让卫星切换发生在微秒之间

用户终端是相控阵,卫星上的天线也是相控阵。两边都能电子扫描,不靠机械转动。

马斯克描述了切换的体验:从一颗快速掠过头顶的卫星切到另一颗,在微秒级别完成,用户察觉不到延迟或数据的变化。一颗卫星可以在地面投射多个蜂窝式的点波束,服务不同位置的用户。

他还强调了一个词:数字相控阵。这意味着波束的形状和指向可以靠软件重新编程,随着算法改进,同一个硬件的效率还能往上走。

这一点在卫星行业里不常见。传统卫星的设计寿命是十五年,发射上去那一刻性能就固定了,后面十几年只能眼睁睁看着地面技术超过自己。数字相控阵把一部分性能留在了软件里,硬件打上去之后还能继续变好。

这套设计的真正价值藏在「升级」这两个字里,峰值速度只是表面。硬件打上去之后不用干等十五年,算法迭代一次,在轨的几千颗卫星同时变强。这种「发射后仍可变好」的能力,在重资产行业里几乎算作弊。

再往深一层想,数字相控阵还带来一个运营上的好处:频谱的分配可以动态调整。不同地区的用户密度、干扰情况、监管要求都不一样,波束的形状和频率用软件配置,不用为每个国家定制硬件。这一点在国际化扩张时省下的是时间和审批成本。

真正的护城河在发射环节:一年送上去全球 80% 的入轨载荷

技术讲完,马斯克转而讲 SpaceX 手里另一张牌,这张牌对手短期内抄不了。

他的说法是,SpaceX 有发射频率极高的发射系统,能把很大的质量送到轨道上。2020 年,SpaceX 送了全球入轨载荷的大约三分之二。2021 年这个比例可能接近 80%。

剩下的部分他也拆了:大约 12% 是中国,8% 是其他所有国家加起来。

80% 这个数字需要一点背景才看得懂。它算的是质量(payload mass),并非发射次数。星链卫星一颗 260 公斤左右,一次猎鹰 9 能打 60 颗,一发就是十几吨。这种批量发射把统计口径整个拉偏了。

成本为什么能压下来,马斯克说得很直接:助推器可重复使用,整流罩也可重复使用,所以每次只需要造二级火箭。这在历史上从未有过。入轨成本因此降到了历史最低。

他把这件事定义成一个闭环:星链和猎鹰 9 互相喂养。星链给猎鹰 9 提供稳定的、内部的、不受市场波动影响的订单,猎鹰 9 给星链提供别人拿不到的低价发射。

对照一下就清楚了。OneWeb 早期主要靠联盟号火箭发射,单价高,节奏受制于人。亚马逊的 Kuiper 要等蓝色起源的 New Glenn 成熟,或者去买 ULA 的发射服务。在一个成本由发射决定的行业里,拥有发射能力等于拥有了定价权。

马斯克还顺带提了一句制造端的难度:学会造这些先进卫星,学会批量造这么多颗,学会造用户终端和网关,本身就是一场冒险。这句话听着轻,点出的却是另一层壁垒。实验室里造一颗卫星和一年造一千颗卫星,是两种完全不同的工业能力。

值得多看一眼的是这个闭环的时间差。猎鹰 9 的复用技术成熟在前,星链的规模化需求在後,两者刚好接上。如果星链早三年启动,SpaceX 没有足够的发射能力支撑;如果星链晚三年启动,OneWeb 和 Kuiper 可能已经把频率和轨道位置抢完。窗口期这种东西事后看总是刚好,身处其中的人只能靠赌。

氪离子推进器、激光星间链路和明年的 v2 卫星

讲到卫星本身,马斯克提到一个自研部件:SpaceX 自己开发的氪(krypton)霍尔效应推进器(Hall effect thruster),原理是喷出高速的氪离子。

霍尔推进器的推力很小,比冲却高,能用很少的推进剂维持轨道、调整位置、寿命结束时主动离轨。用氪而不用更常见的氙,原因在价格,两者差好几倍。一个要造上万颗卫星的公司,推进剂的单价会被放大成一个真实的成本项。

马斯克在这里插了一个玩笑:超人当心点,别靠近我们的卫星。氪石是超人唯一的弱点,这个梗他显然很满意。

真正的技术升级在后面。他说 1.5 版卫星已经接近发射,这一版带激光星间链路(laser intersatellite links),主要解决北极和南极地区以及高纬度地区的连续覆盖问题。

激光星间链路的意义要展开说。传统卫星通信的路径是:用户、卫星、地面网关、地面骨干网。没有网关的地方就没有服务,海上、沙漠、极地都覆盖不了。有了星间链路,卫星之间可以直接用激光传数据,一颗卫星可以把流量转给另一颗,一路接力到有网关的地方。

这会改变成本结构。全球建几百个地面网关是巨额资本开支,星间链路能把这个数字砍下来。它也会改变延迟:数据在太空里走直线,路径可能比绕地面光纤更短。

马斯克说,2022 年开始发射 v2 版卫星,能力会显著更强。1.5 版之后的所有卫星都会带激光链路。

v2 版具体强在哪儿,他没有展开。按当时公开的信息推测,方向无非几个:更大的天线面积带来更高的单星容量,更多的激光终端带来更密的网状连接,更高效的推进和电源系统带来更长寿命。这些改进单看都不惊艳,叠在一起就是每颗卫星能服务的用户数翻几倍。对一家正在亏损的公司来说,这个数字直接决定盈亏平衡点在哪里。

数据不走弯路:直连主要服务器中心

访谈里有一段很多人会跳过去,对体验的影响却最大。

马斯克说,SpaceX 正在把网关和接入点直接连到主要的数据中心。这样数据从用户终端出来,到服务器中心,走的是尽可能短的路径。

他举了例子:有人在用 YouTube、Netflix、Google 搜索、Xbox、AWS,数据都走最短路径,延迟因此在总体上被压到最低。

过去卫星宽带的糟糕体验,很大一部分出在地面这一段。用户的流量要先进运营商的网关,再绕到骨干网,再到目标服务器,中间的绕路和互联环节加了几十上百毫秒。轨道高度改不了这段路,只有把接入点搬到服务器旁边才行。

马斯克还提了一个附带的好处:即使互联网的大块区域中断,Starlink 仍然能提供连接。这句话没有展开,意思却清楚。地面网络是树状结构,主干断了,下游全断。卫星网络的拓扑不一样,用户直接对着天,不依赖某一条地缆。

星链在偏远地区和应急通信之外,还多了第三个卖点:绕开地面基础设施的脆弱性。

这个卖点在后来的几年里被反复验证。地震、山火、洪灾、战争,凡是地面通信被打断的场合,星链都会出现在新闻里。它成了一个国家通信体系的备份层,而这个定位的付费方往往不是个人用户,是政府和机构。这类客户的黏性远高于零售用户。

50 亿到 100 亿美元只是门票,累计投入可能到 200 到 300 亿

主持人问到钱。这个项目到底要投多少。

马斯克先是难得地表现出一丝犹豫。他说这取决于怎么算,一种算法是看现金流转正之前一共投进去多少钱,以及要不要把猎鹰 9 的研发也算进去。

然后就给了数字:到完全实现正向现金流,可能是至少 50 亿美元,也可能高达 100 亿美元。他说这确实是一大笔钱。

后面的部分更值得注意。主持人接着问,现金流转正之后还要投多少。马斯克说,那个阶段之后还得继续大额投入,原因是不能让自己被地面蜂窝网络的持续进步、蜂窝覆盖的持续扩张,或者成本更低的同步轨道卫星变得无足轻重。

他解释了一句 GEO 的局限:同步轨道卫星往往能覆盖很大一片区域,能提供的总带宽却不高,延迟还高。即便如此,Starlink 也必须能给出和 GEO 相当、最好更低的价格。

所以他的估算是,随着时间推移,总投入会是最初那个数字的若干倍,可能是 200 亿到 300 亿美元。

把这三段连起来看,是一次很少见的坦白。多数基础设施项目在融资时讲的是「投完这些就够了」,马斯克讲的是「投完这些才开始」。他给出的理由也落在防守上:地面网络每年都在变好,你停下来的那天,就是被追上的那天。

50 亿到 100 亿美元这个区间对应的用户规模是多少,他没说。按 12 个月 50 万用户、长期数百万用户的路径,这门生意要成立,ARPU 必须撑得住,设备成本必须继续往下掉。用户终端从早期的上千美元降到几百美元,这里省下的每一块钱都要乘以百万量级。

还有一个容易被忽略的账:卫星的寿命。低轨卫星的设计寿命大约五年,比同步轨道卫星的十五年短得多。这意味着星座不是一次性投入,是一条持续的替换流水线。每年都有一批卫星要退役、要补发。这笔常年支出在早期的融资模型里经常被低估,却是决定长期现金流的关键项。马斯克说要投 200 到 300 亿美元,很大一部分就花在这条流水线上。

主持人最后想接着问普通用户怎么开通服务,文字稿在这里断了。前面这些内容已经足够画出 Starlink 在 2021 年中的样子:技术上激进到行业里没人敢跟,财务上保守到把「不破产」当成第一目标,商业上把自己限定在 3% 到 5% 的边缘市场,然后用自建的发射能力把成本压到别人跟不动的位置。

四年之后再回看这场访谈,有几个判断被验证了,也有几个数字明显滞后。框架没有变:低轨星座的胜负,最后落在单位经济学上。

  • 评估任何一类新兴基础设施项目,先问它服务的是哪一群人,别急着看技术参数。星链的答案是 3% 到 5% 的边缘人口,这个答案决定了它的成本结构和定价空间。如果一个项目说不清自己服务谁,它的商业模型大概也说不清。
  • 测算一项业务能不能活下来,先算生命周期内的累计投入,再算现金流转正的时间点。星链的门票是 50 亿到 100 亿美元,累计可能到 200 亿到 300 亿美元。融资材料上写「需要 X 亿」的项目,多半只算了第一阶段。
  • 看一家公司的护城河,去找它哪一部分成本是自己控制的。SpaceX 的答案是发射,因为助推器和整流罩都能重复使用,每次只造二级。别人买发射,它自己造发射,这一项决定了谁能撑到最后。
  • 把消耗品的成本也算进模型。氪比氙便宜好几倍,一颗卫星省下的钱不多,一万颗卫星就是一笔真金白银。规模会改写每一个小决定的权重。
  • 行业里被当成铁律的保守做法,值得每过几年重新审一遍。航天业为了可靠性坚持用落后一代的技术,SpaceX 反过来用最先进的相控阵,因为迭代速度本身就是一种可靠性。你的行业里有没有类似的、没人敢动的默认规则?
  • 判断技术方案时,分清哪些约束是物理的,哪些是工程的。光速决定了同步轨道卫星的延迟下限,工程优化救不回来;500 公里的低轨把这个下限降到个位数毫秒,剩下的都可以改。选错物理层,后面所有努力都白费。
  • 如果一个项目的第一目标是「不破产」,这个目标本身值得尊重。低轨通信星座四十年里全军覆没,活下来的人靠的是钱烧得比对手慢,技术领先只是次要因素。任何时候,先算清楚自己能撑多久,再谈能跑多快。
  • 看到「每秒 30TB」这类数字,先去核对单位。语音转写会把 Tbps 记成 TB/s,两个量级差 8 倍。二手材料里的技术参数,凡是不能反推出量级合理性的,都要打问号。
YouTube 平台字幕(自动或人工),可能有识别错误,人名与数字请以视频为准。已按语句合并成段落,并去掉 um、uh 等口头填充词。 文稿来源:YouTube(自动/人工字幕)。

0:00yes [Applause] elon we're going to talk obviously about spacex the wider goals for spacex before we do that i would love to focus on starlink now this is a plan to provide the world with broadband coverage via satellites you've actually compared the project to rebuilding the internet in space

0:30you're an engineer at heart elon what has driven you to take on this new challenge well there's a need for connectivity in places that don't have it right now or or where it is connectivity is very limited or very expensive it's it may be possible like i said that's simply not available so you can think of starlink as filling in the gaps between 5g and fiber and and really getting to the parts of the world that are the the hardest reach the the most difficult to reach three percent possibly five percent and i think it really quite nicely complements fiber and and 5g

1:23nice complementary technologies you're saying and we'll dive a bit deeper into that elon first of all give us a bit of a kind of high level overview as to how far progressed starlink actually is what have you done so far so we've launched and now have active over 1500 satellites let's see the i mean a few interesting stats the the combined power of all the satellites is over five megawatts so they're there's over five megawatts of solar from all the satellites combined it's they're capable of outputting about 30 terabytes per second of data and and starting in actually next month almost almost next month i should say starting in august we should have global connectivity

2:21for everywhere except the poles i know as i said it's it's it's really meant for sparsely populated regions so the our our spot size taking talking in terms of cellular is quite big so we're also to low to medium density areas but but not too high density areas in the high density areas we will serve a limited number of customers

2:55and it is operational we recently passed the strategically notable level of 69 420 active users and we are i think on our way to having a few hundred thousand users possibly over 500 000 users within 12 months so i think so it's quite it's growing rapidly and we're continuing to innovate the user terminal and the satellite and the the ground stations and gateways and points of presence i think we're operational now in about 12 countries and more being added every month so it's like so it's it's a nice compliment to fiverr and to 5g and it's also although we can't talk about about those deals today because our partners are ready to announce them i think it can be quite

4:04useful to a lot of telcos for data back home so just you know if people have cellular stations in remote regions just using starlink for data backhaul to their network can be a very cost effective way of doing data back home and then notably for solving relative to say other satellite communication systems we are at around 500 kilometers whereas the the synchronous satellites are around 36 000 kilometers so latency for a solid system is similar to latency for ground-based fiber and 5g so we're expecting to get latency down under 20 milliseconds so you can still do it it feels very fast like there's no lag and you can you can play for example competitive video games

5:06on a starting system this is a a very tough market elon i mean if we look back over the last couple of decades it has been littered with failure we go as far back as the 1990s there are big names that no longer are with us they've tried to make satellite broadband as success you've also now got strong competition the likes of of amazon and oneweb are also now attempting to launch new satellite broadband projects what is it that that makes starling different to to what's gone on in decades beforehand elon but also with your new competition as well what is it that makes you different yes well when i gave a talk a few years ago about starlink i was asked what's my goal for starting

5:57and i said well our goal is not to go bankrupt because every as you point out every other low earth orbit communications constellation ever done has gone bankrupt now some of them have emerged from bankruptcy but with the let's just say the original owners did not benefit from from those constellations so yeah they've either gone bankrupt after most of them went bankrupt before even deploying the constellation fully and and even the ones that did this deploy their constellation they they still went bankrupt you know they did subsequently emerge from bankruptcy but you know and obviously companies like iridium and opcom are doing reasonably well

6:48today but not for the original owners so step number one for starlink is don't go bankrupt that is and you know then we can if we succeed in not going bankrupt then that'll be great and we can move on from there but i do think from a technology standpoint starlink is quite different from prior leo constellations in that the technology that we're deploying is is very advanced what happens often with space-based technology somewhere ironically is that it tends to be older technology that exists on the ground so companies generally have been very conservative so they've said well we wanna we want technology that's proven to work on the ground and then we have to test it

7:44very you know for four years to make sure it's going to work in the vacuum of space with you know a higher radiation load and that kind of thing and and so the technology that's generally been launched into orbit has been older technology and we took the officer's approach and said we're going to make technology that is in some ways at least more advanced than what is on the ground and we're just going to take a chance and and so we have what's probably fair to say are the most sophisticated but the most advanced phase rate technologies

8:21that best of knowledge no one has not even it's not even at that sort of military level it has this level of sophistication with base grade technology and i think that's that's sort of quite important for a fast-moving rear constellation because the the terminal which is user terminal that's also phase 3 and the terminal on these the tool and the satellite the antenna on the satellite they're both phased array so you can switch from one satellite that's moving rapidly overhead to another one and do so at the microsecond level so there's no you can't tell as as the system is switching over one satellite to another there's no change in latency or data

9:12from one satellite to another and a single satellite can illuminate many different use of cell spots on the ground so and it's it's and it's a digital phase array system so we can continue to program and reprogram the system to greater efficiency because it is digital so like since the best of my knowledge is the most advanced phase rate system in the world so that's pretty cool we also have a advantage of a of a launch system that is capable of a very high launch rate and capable of putting a lot of mass towards i believe last year spacex delivered about two-thirds of all payload to orbit and this year we may deliver closer to eighty percent of all payloads

10:13orbit of earth just with that position you know so and off the remaining of course remains 16 for this year if we actually put close to 80 percent will payload to it but i've been remaining let's say roughly 20 percent about 12 would be china or their about and then eight percent is everyone else so the shared share master urban capability of the falcon 9 rocket is because it is a reusable booster and the fairing is reusable so we only have to make the upper stage is that has never existed before and the and the cost of mass orbit also is the lowest it has ever been so because of reusability so this is

11:01the two worlds work together quite well but it has been quite an adventure learning how how to make these advanced satellites and learning how to make so many of them and the user terminals of the gateway and the satellites for example use a spacex developed krypton hole effect thruster or which basically shoots out high speed ions of krypton it's kind of cool superman watch out for our satellites don't go close to them that's all that crypto and just in general it's quite a clever satellite and then we're getting close to launching satellite 1.5 which has laser intersatellite links and that will be used for especially for continuous connectivity over the arctic

11:53and antarctic regions or basically the for the high polar regions high the higher latitude regions and then next year we'll start launching a version two of our satellite which will be significantly more capable and then you know all satellites from 1.5 onwards will have major satellite links and then we'll start getting our gateways and our points of presence to really directly to the

12:25the major server centers so the data really goes directly from the user terminal right from user terminal directly to the server center so if somebody's using youtube or netflix or whatever you know google search or xbox whatever the case may be amazon web services the the the the data flows in the shortest path possible thus minimizing latency in general so even if even if big chunks of the internet go down then you still have connectivity

13:06so i'm pretty excited about the future it's looking you know we're get overconfident but it's looking quite positive and like i said just it serves as a as a great natural complement to major telco fiber and 5g talk us through if you can elon the the finances of this huge project how much you investing in starting how much you prepared to invest in starlink

13:37well i think especially depends on how you count investment it is sort of what's the total i mean one way to account investment is like what's the total amount of money invested before selling becomes positive cash flow and and do you how much of that do you include falcon 9 and everything we've done there you know i think probably before we go to you know fully positive cash flow it might be it'll be at least five billion dollars and maybe as much as 10. so it's quite a lot

14:18and yeah and then if you say like well how much are we going to invest even after it goes positive cash flow i think what because you keep a great deal after that point in order to not be made irrelevant by continued improvements in cellular and continued extensions of cellular or or or or lower cost geosynchronous satellites which you know geosynchronized often can conserve a a very large swath of territory but but the total bandwidth they can serve is is not that great and the latency is high but still we we need to be able to offer our service at a you know a comparable or ideally lower rate than gto it's not like connectivity so anyway so it's basically total

15:17investment probably it's like at least five maybe 10 billion dollars and then over time it's going to be some multiple of that i don't know could be 20 or 30 billion over time there's a lot basically yeah it is a lot of money and to recruit that investment obviously you need customers and subscribers elon just in terms of the basics to get myself set up with a

资源与出处

相关视频

同频道、同主题或相邻年份