1
00:00:00,040 --> 00:00:03,880
Elon Musk confirmed SpaceX will 
go public in 2026, reversing a 

2
00:00:03,880 --> 00:00:05,800
position he held for more than a
decade. 

3
00:00:06,160 --> 00:00:10,600
The IPO could raise $30 billion 
in value of the company in 1.5 

4
00:00:10,600 --> 00:00:14,600
trillion, making it the largest 
public offering in history, 

5
00:00:14,760 --> 00:00:20,080
surpassing Saudi Aramco's $29 
billion listing in 2019. 

6
00:00:20,680 --> 00:00:23,720
Musk had long insisted SpaceX 
would remain private until 

7
00:00:23,720 --> 00:00:27,360
rockets were flying regularly to
Mars, arguing that public 

8
00:00:27,360 --> 00:00:31,000
shareholders would prioritize 
profits over his colonization 

9
00:00:31,000 --> 00:00:33,520
dreams. 
So why did he change his mind? 

10
00:00:33,960 --> 00:00:36,280
The answer involves data centers
in space. 

11
00:00:36,640 --> 00:00:40,120
Musk plans to retrofit new 
generation Starling satellites 

12
00:00:40,160 --> 00:00:44,120
into orbital computing platforms
that could power the AI 

13
00:00:44,120 --> 00:00:47,240
revolution, and he needs massive
capital to make it happen. 

14
00:00:47,680 --> 00:00:50,000
Today, we're going to cover the 
technical specs of these space 

15
00:00:50,000 --> 00:00:53,320
data centers, the competitive 
race with Jeff Bezos and Google,

16
00:00:53,520 --> 00:00:56,600
and what this means for Musk's 
Mars timeline. 

17
00:00:57,280 --> 00:01:01,400
Now, SpaceX confirmed the 2026 
IPO after ours Technica reporter

18
00:01:01,640 --> 00:01:04,440
Eric Berger published an 
analysis of why the company 

19
00:01:04,599 --> 00:01:07,920
would finally go public. 
Musk responded on X with four 

20
00:01:07,920 --> 00:01:10,240
words. 
As usual, Eric is accurate. 

21
00:01:10,640 --> 00:01:13,640
The company is currently 
conducting an insider share sale

22
00:01:13,960 --> 00:01:18,320
at an $800 billion valuation, 
exactly double what SpaceX 

23
00:01:18,320 --> 00:01:21,640
commanded at its last capital 
raise in 2025. 

24
00:01:22,040 --> 00:01:25,760
And the core driver behind this 
reversal is a new business line 

25
00:01:26,120 --> 00:01:29,680
that would not exist when Musk 
first swore off public markets. 

26
00:01:30,040 --> 00:01:33,840
SpaceX wants to build orbital 
data centers using its Starlink 

27
00:01:33,920 --> 00:01:37,680
V3 satellites, which feature 
high speed laser links capable 

28
00:01:37,680 --> 00:01:41,040
of 1 terabit per second 
throughput. 

29
00:01:41,880 --> 00:01:45,600
Now, that number is 10 times the
capacity of current Starlink V2 

30
00:01:45,600 --> 00:01:48,040
satellites. 
The V3 satellites require 

31
00:01:48,040 --> 00:01:51,480
Starship to launch because they 
are significantly larger, and 

32
00:01:51,480 --> 00:01:55,760
SpaceX plans to deploy 60 of 
them per flight starting in 

33
00:01:55,760 --> 00:01:58,600
2026. 
Unlike previous Starlink 

34
00:01:58,600 --> 00:02:02,680
generations optimized purely for
communications, the V3 units are

35
00:02:02,680 --> 00:02:05,920
being engineered as compute 
capable infrastructure with 

36
00:02:05,920 --> 00:02:09,080
larger solar arrays, machine 
learning accelerators, and 

37
00:02:09,080 --> 00:02:12,560
radiation shielding designed to 
extend operational life in the 

38
00:02:12,560 --> 00:02:16,280
harsh space environment. 
Each satellite will process data

39
00:02:16,440 --> 00:02:19,880
before transmitting it to Earth,
reducing bandwidth requirements 

40
00:02:19,960 --> 00:02:24,480
and enabling real time AI 
inference at latencies below 10 

41
00:02:24,480 --> 00:02:27,320
milliseconds. 
Now space offers several 

42
00:02:27,320 --> 00:02:30,520
advantages over terrestrial data
centers that make this bet 

43
00:02:30,520 --> 00:02:33,600
rational. 
Solar panels in orbit capture 

44
00:02:33,600 --> 00:02:37,360
roughly 8 times more energy than
ground installations because 

45
00:02:37,360 --> 00:02:40,920
there is no atmosphere, no 
weather, and no night time and 

46
00:02:40,920 --> 00:02:44,120
the sun facing side. 
Cooling comes naturally through 

47
00:02:44,320 --> 00:02:48,160
radiative heat dissipation into 
the vacuum of space, eliminating

48
00:02:48,160 --> 00:02:50,440
the massive water and 
electricity requirements that 

49
00:02:50,440 --> 00:02:53,960
plague Earth based facilities. 
And the International Energy 

50
00:02:53,960 --> 00:02:58,440
Agency reports that data centers
consumed about 415 terawatt 

51
00:02:58,440 --> 00:03:04,360
hours of electricity globally in
2024, roughly 1.5% of total 

52
00:03:04,360 --> 00:03:07,280
power consumption. 
That figure is climbing rapidly 

53
00:03:07,280 --> 00:03:12,080
as AI training runs demand more 
computational resources now. 

54
00:03:12,080 --> 00:03:16,080
Goldman Sachs forecasts that 
global power demand from data 

55
00:03:16,080 --> 00:03:22,280
centers will increase 50% by 
2027, as much as 165% by the end

56
00:03:22,280 --> 00:03:24,800
of the decade. 
Orbital facilities avoid the 

57
00:03:24,800 --> 00:03:28,600
land use conflicts and NIMBY 
opposition that have stalled 

58
00:03:28,600 --> 00:03:31,320
terrestrial data center 
construction across the United 

59
00:03:31,320 --> 00:03:34,920
States and Europe. 
Musk has described an even more 

60
00:03:34,920 --> 00:03:37,840
aggressive long term vision. 
He wants to build satellite 

61
00:03:37,840 --> 00:03:41,360
factories on the moon and use an
electromagnetic rail gun to 

62
00:03:41,360 --> 00:03:45,720
launch EI satellites at lunar 
escape velocity without rockets.

63
00:03:46,320 --> 00:03:50,600
In his estimation, his approach 
could eventually deliver 100 

64
00:03:50,600 --> 00:03:53,840
terawatts of AI computing 
capacity per year into orbit. 

65
00:03:54,240 --> 00:03:57,720
Figure sounds absurd until you 
remember that Musk has a track 

66
00:03:57,720 --> 00:04:01,400
record of building things other 
people called impossible. 

67
00:04:02,120 --> 00:04:05,840
Now SpaceX is not pursuing this 
opportunity in isolation. 

68
00:04:06,160 --> 00:04:08,760
Jeff Bezos announced in October 
that Blue Origin has been 

69
00:04:08,760 --> 00:04:12,080
developing orbital data center 
technology for more than a year.

70
00:04:12,440 --> 00:04:17,240
He predicted GW scale space data
centers without 10 to 20 years 

71
00:04:17,480 --> 00:04:20,120
and claimed they would 
eventually beat the cost of 

72
00:04:20,120 --> 00:04:23,560
terrestrial facilities. 
Google launched Project 

73
00:04:23,560 --> 00:04:27,840
Suncatcher in November of 2025, 
partnering with Planet Labs to 

74
00:04:27,840 --> 00:04:31,640
put TPU chips on satellites 
launching in early 2027. 

75
00:04:32,120 --> 00:04:35,400
And Google's research paper 
envisions 81 satellite clusters 

76
00:04:35,640 --> 00:04:39,200
arranged in a 1 kilometer radius
formation, communicating through

77
00:04:39,200 --> 00:04:43,160
free space optical links to 
distribute AI workloads across 

78
00:04:43,160 --> 00:04:46,320
the constellation. 
The NVIDIA backs startup Star 

79
00:04:46,320 --> 00:04:50,120
Cloud already has a satellite in
orbit running an H-100 GPU, 

80
00:04:50,120 --> 00:04:55,120
successfully training Google's 
Jemma A1 model in space. 

81
00:04:55,560 --> 00:04:58,680
Now. 
Star Cloud CEO told CNBC the 

82
00:04:58,680 --> 00:05:02,760
orbital data centers would have 
10 times lower energy costs than

83
00:05:02,760 --> 00:05:06,080
terrestrial facilities. 
Company plans to build a 5 GW 

84
00:05:06,080 --> 00:05:09,600
orbital data center with solar 
and cooling panels measuring 

85
00:05:09,600 --> 00:05:12,200
roughly 4 kilometers in both 
width and height. 

86
00:05:13,360 --> 00:05:16,560
Now China has entered the race 
with its three body computing 

87
00:05:16,560 --> 00:05:20,560
constellation, which launched 12
satellites in May 2025 through a

88
00:05:20,560 --> 00:05:24,760
partnership between Zhengjiang 
Lab and Alibaba Group. 

89
00:05:25,400 --> 00:05:29,360
The financial picture for SpaceX
makes the IPO timing very 

90
00:05:29,360 --> 00:05:32,120
logical right now. 
Company generated roughly $15 

91
00:05:32,120 --> 00:05:36,360
billion in revenue in 2025, is 
projected to hit about 24 

92
00:05:36,360 --> 00:05:40,640
billion in 2026, putting it 
roughly on par with NASA's 

93
00:05:40,640 --> 00:05:43,600
annual budget. 
And Starlink alone has grown to 

94
00:05:43,600 --> 00:05:47,040
over 9 million subscribers from 
4 million just 15 months 

95
00:05:47,040 --> 00:05:50,400
earlier. 
Musk stated publicly the less 

96
00:05:50,400 --> 00:05:53,920
than 5% of SpaceX revenue now 
comes from NASA, meaning the 

97
00:05:53,920 --> 00:05:56,920
future of the company depends 
almost entirely on Starlink. 

98
00:05:57,240 --> 00:06:01,480
SpaceX launched more than 160 
Falcon 9 rockets in 2025, which 

99
00:06:01,480 --> 00:06:04,240
is more than the rest of the 
world combined, and put over 

100
00:06:04,240 --> 00:06:07,720
3200 Sterling satellites into 
orbit during the calendar year. 

101
00:06:08,200 --> 00:06:12,640
A single Falcon booster, 
designated B1O67 has now 

102
00:06:12,640 --> 00:06:16,120
launched and landed 32 times, 
with turn around windows as 

103
00:06:16,120 --> 00:06:19,880
short as three weeks. 
The company hit successful 500 

104
00:06:19,880 --> 00:06:23,560
landings in October 2025, more 
than every other company in 

105
00:06:23,560 --> 00:06:27,320
nation state combined. 
Blue Origin, in second place, 

106
00:06:27,360 --> 00:06:33,600
has landed exactly 1. 
Now Mars, it remains on the 

107
00:06:33,600 --> 00:06:36,240
forefront. 
It's the elephant of the room 

108
00:06:36,240 --> 00:06:39,440
for long time SpaceX fans. 
Musk founded the company 

109
00:06:39,440 --> 00:06:42,480
specifically to make humanity 
multiplanetary, and he spent 

110
00:06:42,680 --> 00:06:45,000
years arguing the public 
shareholders would never 

111
00:06:45,000 --> 00:06:48,360
tolerate that expense. 
Now he said it would be 

112
00:06:48,360 --> 00:06:51,200
unprofitable work of colonizing 
another planet. 

113
00:06:51,640 --> 00:06:54,600
In 2023, he told Twitter 
followers that SpaceX would only

114
00:06:54,600 --> 00:06:59,480
go public when Mars colonial 
transport, Now SpaceX President 

115
00:06:59,480 --> 00:07:02,400
Gwyn Shotwell reinforced his 
position in 2018, saying the 

116
00:07:02,400 --> 00:07:05,560
company would remain private 
until rockets were flying 

117
00:07:05,560 --> 00:07:08,960
regularly to the Red Planet. 
That's not happened yet. 

118
00:07:08,960 --> 00:07:11,720
They haven't even gotten a 
Starship to orbit yet. 

119
00:07:12,080 --> 00:07:15,560
There's no in orbit refuelling, 
and they haven't landed on 

120
00:07:15,560 --> 00:07:19,240
another celestial body, let 
alone on Mars. 

121
00:07:19,560 --> 00:07:22,720
Starship has not completed a 
successful orbital mission after

122
00:07:22,720 --> 00:07:25,840
several explosive tests, and 
unaccrued flights to Mars remain

123
00:07:25,840 --> 00:07:30,280
at least a year away by the most
optimistic projections. 

124
00:07:30,280 --> 00:07:31,960
But that's probably not going to
happen as well. 

125
00:07:31,960 --> 00:07:36,040
Elana stated on the record that 
Mars is the distraction now. 

126
00:07:36,880 --> 00:07:40,880
Justice Palmer, CEO of Fortuna 
Investments and venture capital 

127
00:07:40,880 --> 00:07:45,480
firm invested in SpaceX, said he
expected an IPO to happen after 

128
00:07:45,480 --> 00:07:49,160
Starship reaches Mars because it
would remove a massive risk 

129
00:07:49,160 --> 00:07:52,480
factor. 
A failed Mars attempt as a 

130
00:07:52,480 --> 00:07:56,520
private company causes no risk, 
no stock fluctuation, but the 

131
00:07:56,520 --> 00:07:59,960
same failure as a public company
could hammer the share price. 

132
00:08:01,720 --> 00:08:05,040
Now the technical challenges are
substantial, but not 

133
00:08:05,040 --> 00:08:07,920
insurmountable. 
Radiation in space can damage 

134
00:08:07,920 --> 00:08:11,600
computer chips, causing bit 
flips and burnout that corrupt 

135
00:08:11,600 --> 00:08:15,120
AI calculations. 
Space rated processors currently

136
00:08:15,120 --> 00:08:18,000
lag three to four orders of 
magnitude behind commercial 

137
00:08:18,000 --> 00:08:21,400
hardware and raw performance, 
but Google estimates that space 

138
00:08:21,400 --> 00:08:25,880
based AI computing will only 
become economically viable by 

139
00:08:25,880 --> 00:08:31,480
2035 if launch cost dropped to 
$200 per kilogram from the 

140
00:08:31,480 --> 00:08:36,880
current 1400 to $2900 range. 
Starship is the key to hitting 

141
00:08:36,880 --> 00:08:41,080
that target, which puts SpaceX 
in a stronger position than any 

142
00:08:41,240 --> 00:08:42,240
other competitor.