Latest / In The Loop / Elon Musk Merged xAI With SpaceX (And Filed To Put A Million Data Centers In Orbit)

Transcript

Elon Musk Merged xAI With SpaceX (And Filed To Put A Million Data Centers In Orbit) · In The Loop

• 0:00
  Elon Musk just merged his AI company with his rocket company,
• 0:03
  creating a $1.25 trillion company that.
• 0:08
  Is aiming to put a million data centre satellites into.
• 0:11
  Space in this. Episode I'm going to dig into
• 0:14
  the financials behind this deal, the skeptics who think it's a
• 0:17
  bailout. Dressed up as a big vision and
• 0:20
  the. People.
• 0:20
  Who really believe in this and believe it's going to be
• 0:22
  completely transformative now what I really.
• 0:25
  Want to do with this is use it as a topic to explore I guess
• 0:29
  the historical patterns that this move highlights and how
• 0:33
  industry and production. Always follow the cheapest
• 0:36
  energy source, from textile mills moving to coal fields to
• 0:39
  data. Centres, even hydroelectric.
• 0:42
  Dams today. We're going to look at the
• 0:43
  Carter Chef scale, the idea behind the Dyson.
• 0:47
  Sphere. And the concept of a
• 0:49
  civilization harvesting its star and whether.
• 0:52
  AI is growing energy, appetite is the thing that's going to
• 0:54
  finally push us. To move.
• 0:57
  Infrastructure off planet. This is in the Loop with Jack
• 1:00
  Halton. Hope you enjoyed the show.
• 1:18
  So I think let's start with the merger and context.
• 1:21
  So many of you will have heard about these companies on one.
• 1:24
  Hand you've got XAI, which is obviously the AI company that
• 1:27
  Elon Musk started. Which is basically ChatGPT but.
• 1:31
  Accessible via. Twitter or X?
• 1:33
  And API. So it's the one behind many of
• 1:35
  the scandals that are enabling people to put in a picture and
• 1:38
  say, turn this person into the same person but wearing
• 1:41
  underwear. And and of.
• 1:43
  Course, everybody knows. SpaceX is.
• 1:45
  The Rocket Company and this. Really sets the scene behind
• 1:48
  this. Big merger.
• 1:49
  Because a single query to a. Large language model consumes
• 1:52
  roughly. About 10.
• 1:54
  Times the energy of a single Google search.
• 1:57
  So if you scale that to say. Billions and billions of queries
• 2:00
  happening every day and. The picture obviously from an
• 2:03
  environmental. Perspective could get quite.
• 2:05
  Uncomfortable. Quite fast.
• 2:07
  The International Energy Agency actually found that global.
• 2:10
  Data centre. Energy consumption.
• 2:13
  Hit about 400. And 15 terawatt hours in 2024
• 2:16
  which? To put that into context, is
• 2:18
  about. 1.5% of. All electricity.
• 2:21
  Generated on Earth. And by 20-30.
• 2:24
  This same agency, the International Energy Agency.
• 2:27
  Actually believes this number will be more than double.
• 2:30
  To nearly 1000 terawatt hours, which is slightly more than.
• 2:33
  Japan's entire. National electricity consumption
• 2:37
  as it is today. And Goldman Sachs actually
• 2:40
  estimates that power demand is going to.
• 2:42
  Surge equivalent to adding. Another top 10.
• 2:45
  Energy consuming country to the planet.
• 2:47
  And that's before you. Account for the growth
• 2:49
  trajectories. Everybody keeps revising
• 2:52
  upwards. So it's in this context that
• 2:54
  about 10 days ago. That Elon Musk announced the
• 2:56
  merger. Of SpaceX and.
• 2:58
  Xai creating. Well, the company called one of
• 3:01
  the most ambitious virtually. Integrated innovation engines on
• 3:05
  and importantly. Off Earth, the combined company.
• 3:08
  Is valued at about $1.25 trillion.
• 3:13
  With SpaceX at roughly a trillion and XAI about 200.
• 3:15
  And $50 billion. So this is, by several measures,
• 3:18
  the largest ever merger in history.
• 3:20
  And it's all being driven by demand and need for AI and the
• 3:24
  financials underneath. All of this.
• 3:26
  To be honest, is it bonkers? SpaceX generated about. $15
• 3:29
  billion in revenue last year, with an estimated $8 billion in
• 3:33
  profit, and Bloomberg has reported, the XAI recorded.
• 3:37
  A loss of a 1.46. Billion dollars in September
• 3:41
  alone and they're only making about $107 million in revenue
• 3:44
  and that. Revenue.
• 3:45
  Itself has been questioned because.
• 3:47
  Revenue did double last quarter, but it would need.
• 3:50
  To double. About four more times to catch
• 3:53
  up with their. Burn rate and.
• 3:55
  The quality of that revenue is again questionable because a lot
• 3:59
  of people are complaining. That actually.
• 4:01
  What they're calling. LM revenue is just advertising
• 4:03
  revenue on Twitter. Days before that merger
• 4:06
  announcement, SpaceX filed an. FCC application form which
• 4:10
  basically allowed them to apply to PUT. 1,000,000.
• 4:13
  AI satellites into low Earth orbit.
• 4:16
  And just. To really understand what that
• 4:18
  means, consider that there's about. 14 1/2 thousand active
• 4:23
  satellites around Earth right now.
• 4:25
  So obviously that is a huge leap in.
• 4:28
  The number of. Satellites in low Earth orbit
• 4:31
  and what the actual application described is a first step
• 4:35
  towards becoming a Kardashev. Level 2 Civilization.
• 4:39
  One that can. Eventually, harness the power of
• 4:42
  the sun to fuel energy demand. And SpaceX has also argued that
• 4:46
  orbital data centers. Would be cheaper and more
• 4:48
  environmentally friendly than those on earth.
• 4:51
  With satellites. Venting heat into.
• 4:53
  Obviously. Space which is extremely cold
• 4:55
  and harvesting solar. Energy without any.
• 4:58
  Clouds or atmosphere in the way. And of course, Musk's claims.
• 5:01
  Go even. Further, he said the and I quote
• 5:02
  my estimate is that within two to three years, the lowest cost
• 5:06
  way to. Generate AI compute will be in
• 5:08
  space, so. When I say compute by.
• 5:09
  The way or people say that in the news.
• 5:11
  It's basically the way we generate energy to produce
• 5:15
  computing. Power and looking.
• 5:17
  Long term, he described. Lunar manufacturing as.
• 5:20
  A way to deploy about. 500 to 1000 terawatts of AI satellites
• 5:24
  into deep space orbit every single year.
• 5:26
  And again, remember the needs of say, Japan is about 900
• 5:31
  terawatts of energy, and if they managed to do this, it would be.
• 5:34
  Thousands more gigawatts of energy than exists.
• 5:37
  On Earth now. Obviously not everybody's buying
• 5:39
  this, you know, there's people like investor Ross, Gerber
• 5:43
  offered. What became one of the most
• 5:45
  shared tweets or. Reactions on Twitter.
• 5:49
  Saying basically extras. Out of money S it merged with
• 5:52
  XAI. XAI was then out of money, so it
• 5:55
  merges SpaceX. SpaceX is going to run out of
• 5:57
  money, so it merges with Tesla. And when they're all out of
• 6:00
  money, they're going to go to the public stock market for a
• 6:02
  huge fundraiser. And it's interesting because
• 6:05
  SpaceX didn't explain how XAI would actually contribute.
• 6:08
  To orbital. Data centres beyond being just a
• 6:11
  customer arguing that. Really, the move was quite
• 6:13
  financially motivated. Because ultimately, Musk with
• 6:16
  XAI faces the same funding. Challenges as any startup leader
• 6:21
  in the AI space. You're competing against.
• 6:23
  Meta and Google and their advertising business model.
• 6:26
  Because ultimately, what? Gary Marcus said in his sub
• 6:29
  stack was. By trading XA say thousands
• 6:32
  times their actual. Revenue when they're in an
• 6:34
  increasingly crowded field where everyone is forced really into a
• 6:38
  price war seems crazy and. Behind a lot of this, it's the.
• 6:42
  IPO question because. SpaceX is reportedly targeting a
• 6:45
  mid June 2020. Six IPO.
• 6:47
  And it's supposed to be probably 1 of the largest.
• 6:50
  IP OS ever in history? And so many people are saying
• 6:52
  that this is less about orbital. Space centres and more just
• 6:55
  about trying to create a very attractive entity essentially
• 7:00
  for that public listing to generate the most cash.
• 7:03
  But I mean, obviously optimists have arguments too.
• 7:05
  You know, Tom Nash pointed out that when SpaceX talks about.
• 7:08
  Infrastructure history suggests that everybody should really be
• 7:11
  listening. Even if it starts with, what is?
• 7:14
  Very uncomfortable, unrealistic ideas at first.
• 7:17
  And a good example of this is. Starlink when?
• 7:18
  SpaceX. Began launching.
• 7:21
  Internet into space. The economics, obviously.
• 7:23
  Didn't work. But since?
• 7:24
  Then the. Cost to launch has dropped from
• 7:27
  20 fold to. Very, very affordable.
• 7:30
  And now Starlink has over 9 million global.
• 7:33
  Subscribers generating about $10 billion in revenue.
• 7:36
  So the thing that everyone said was impossible became possible.
• 7:48
  And this is where I think the really interesting historical.
• 7:51
  Comparison can be made. And as you know if.
• 7:53
  You've been listening to the podcast on a regular basis.
• 7:55
  Any opportunity to make historical power that parallels
• 7:58
  is always quite fun, because it does tell us whether this is
• 8:02
  likely to succeed, to be honest. Or likely to gain more and more.
• 8:06
  Momentum because. Before we say.
• 8:08
  Anything I'll mention? Orbital data centers.
• 8:11
  The pattern really is held for 250 years, that says.
• 8:14
  Yes, we will end up in space to generate energy and computing
• 8:17
  power, you know. Look at the late 18th century.
• 8:19
  So textile. Mills in England started out at
• 8:22
  rivers. Or just around rivers, they had
• 8:24
  access. To water wheels where?
• 8:26
  That was the only real source of revenue, and the Pennine values
• 8:30
  north and. East of.
• 8:31
  Manchester were fantastic for this, you know, had lots of free
• 8:34
  flowing streams that drove. Cotton production.
• 8:38
  Then we saw. The Steam.
• 8:39
  Engine change everything. Suddenly, mills didn't need
• 8:42
  rivers, they needed coal. And so they moved by 18-O2.
• 8:46
  Manchester and Place. That Salford had more than. 50
• 8:50
  steam powered mill was. Drawn there.
• 8:52
  By. Ready access to.
• 8:54
  Coal fields in Lancashire. The same pattern played out in
• 8:57
  the. 20th century with aluminium smelting you look at data
• 8:59
  centers Microsoft has operated. A huge.
• 9:02
  Campus in Quincy in Washington since about. 2000 and. 6.
• 9:06
  And this place has no particular reason to.
• 9:08
  Exist as a. Technology hub, except that it
• 9:12
  sits downstream. Of a huge dam where obviously
• 9:15
  hydroelectric power is among the cheapest in the country and
• 9:18
  Intuit. Yahoo.
• 9:20
  Vantage has followed Microsoft. To the exact.
• 9:22
  Same place. So the need.
• 9:24
  To generate energy is getting crazier and crazier, and in
• 9:26
  2024, Microsoft even. Signed a 20.
• 9:30
  Year power agreement to restart a nuclear power plant.
• 9:34
  And Amazon and Google are doing the exact same.
• 9:36
  Trillion dollar companies doing what governments used to do just
• 9:40
  to generate energy for data centers.
• 9:42
  And so obviously this is the context that makes.
• 9:44
  This space argument really interesting and attractive.
• 9:46
  To. Investors and and.
• 9:48
  SpaceX and just the entire. World, quite frankly, because
• 9:52
  solar energy in orbit. Is about 8 times more productive
• 9:55
  per unit of panels. Of solar panel than on earth.
• 9:58
  Because there's no. Atmosphere absorbing radiation,
• 10:01
  No night cycle, no weather, and no seasonal variation.
• 10:05
  But obviously that doesn't mean that this is easy.
• 10:06
  Because the engineering. Challenges are huge.
• 10:09
  Like latency is real. Even in low Earth orbit.
• 10:12
  You're adding about 20 to 50 milliseconds of a round trip of
• 10:16
  sending energy and information, which obviously rules out any
• 10:20
  real. Time applications and.
• 10:22
  Radiation also degrades electronics, so maintenance
• 10:25
  means sending robots to space, not technicians.
• 10:28
  And the manufacturing scale to produce 1,000,000 satellites
• 10:32
  doesn't even. Exist on the world.
• 10:34
  Yet, and to put that into perspective, SpaceX currently.
• 10:37
  Launched about 2000 satellites per year.
• 10:40
  So just imagine what? Sending a million satellites to
• 10:42
  space would require in terms of production and then launching
• 10:46
  it. It's it's close to science
• 10:47
  fiction, but it's also believable as a vision.
• 10:50
  So yeah, I think this pattern. Holds.
• 10:52
  Industry has always. Followed the cheap energy every
• 10:54
  single time. Without exception.
• 10:56
  And the question isn't. Whether that continues, to me
• 10:59
  it's it will. Continue the.
• 11:00
  Question is whether and when that.
• 11:02
  Next migration is going to. Be upward space.
• 11:04
  I mentioned it before. Personally, I believe that if
• 11:07
  we're to survive. As a society and exist in the
• 11:09
  future, we. Need to do.
• 11:12
  That we need to take that leap. You know we're killing the
• 11:14
  planet even when COVID. Shut down most of the world.
• 11:17
  We were producing far too much CO2.
• 11:20
  Emissions to even hit the Paris climate agreement targets, and
• 11:24
  so I don't really believe it's possible to reduce.
• 11:26
  Our emissions fast enough and that's not me saying.
• 11:29
  We shouldn't try to or we shouldn't be more efficient.
• 11:32
  You know, I've been a vegetarian for over 10 years.
• 11:34
  I don't own a car. But I do believe we obviously
• 11:37
  need to find solutions. That are extremely ambitious and
• 11:40
  I do. Believe we have to do that fast.
• 11:43
  I. Also believe that AI has a huge.
• 11:45
  Role in that process, and this really is.
• 11:58
  Where that Khadashev? Vision of society comes in and
• 12:01
  why I think it's a very interesting topic.
• 12:03
  And the fact? That it's been referenced by
• 12:05
  SpaceX is even more interesting. So, so let's just discuss that
• 12:07
  just for a second. So in 19. 64 a.
• 12:10
  Soviet astrophysicist named Nikolai Khadashev published a
• 12:14
  paper in which he proposed classifying civilizations by the
• 12:18
  amount of energy they consume, and so a.
• 12:21
  Type. One civilization harnesses.
• 12:22
  All energy on its. Planet A Type 2 will harness the
• 12:26
  entire. Output of its star type 3 will
• 12:29
  take all the energy of its Galaxy.
• 12:31
  And apparently according to Carl Sagan's formula.
• 12:34
  Today we roughly sit at. 0.73 on that scale, so not even type 1
• 12:39
  yet and. People predict that we might
• 12:40
  reach type 1. Officially in about 100 years
• 12:43
  time. And so obviously, what makes
• 12:45
  that SpaceX? Filing very interesting is
• 12:48
  regardless of what you think of the company or of Elon.
• 12:50
  Musk it explicitly. Quotes that framework.
• 12:54
  To move to that. Type 2 civilization and what?
• 12:57
  They're really alluding to in the submission, and the
• 12:59
  application is a Dyson sphere. Which Freeman Dyson had already.
• 13:04
  Described what? A type 2.
• 13:05
  Civilization infrastructure might look like but.
• 13:08
  Essentially, he described this. Technology, which is
• 13:12
  essentially. A swarm.
• 13:13
  Of objects around its star to. Capture its total.
• 13:16
  Energy output, and he later clarified that a.
• 13:19
  Rigid shell, obviously. Was just impossible, but.
• 13:21
  If you had swarms of objects. Travelling an independent orbits
• 13:25
  around the sun, sending energy back.
• 13:27
  That would be. A way of harnessing the sun and.
• 13:30
  Obviously, that's Spacex's vision.
• 13:32
  Here a million solar panelled. AI satellites orbiting.
• 13:35
  Earth and then eventually, of course, as it said, to reach.
• 13:38
  A type. Two civilization so the same but
• 13:41
  for. The sun.
• 13:42
  So yeah, I I do think that this it feels.
• 13:44
  Exciting. It feels right.
• 13:46
  Whether it's overly ambitious, whether it's just for money,
• 13:48
  whether the companies are dying, I don't know.
• 13:50
  It's hard. To tell, but regardless, taking
• 13:53
  a step towards. Unlimited clean energy.
• 13:55
  Is probably. Always a good thing.
• 13:56
  I personally. Believe in that importance of
• 13:58
  becoming a Type 2 civilization. Before we kill.
• 14:01
  Ourselves off. Anyway, let's conclude and wrap.
• 14:13
  Up. Obviously, as we spoke about
• 14:15
  every single query you. Send to an AI model.
• 14:18
  Every image generated, every agent running in the background
• 14:20
  is consuming loads of energy. And collectively these queries
• 14:23
  are. Creating a demand.
• 14:25
  Curve which is going to. Double by 2013 apparently is.
• 14:29
  The same as adding another country.
• 14:31
  To the energy grid, a top 10 country in terms of energy
• 14:34
  consumption. And this SpaceX?
• 14:36
  Filing and the. Merger with XAI speaks to.
• 14:40
  Either a vision of the future or being completely delusional.
• 14:43
  And the honest answer is that we just won't know for years.
• 14:46
  You don't have to like the person proposing it.
• 14:50
  You don't have to like the company.
• 14:51
  I certainly would never own a Tesla, for example.
• 14:54
  But ultimately. I do believe in that future of
• 14:58
  the planet anyway. Thanks for listening I.
• 15:01
  Hope I'll see you next week and I hope you enjoyed today's
• 15:04
  episode.