Latest / Elon Musk Thinking / Latest: Elon Musk Speech, Talked about the SpaceX Super Intelligence Computer!!!

Transcript

Latest: Elon Musk Speech, Talked about the SpaceX Super Intelligence Computer!!! · Elon Musk Thinking

• 0:00
  Speaker 1: Latest interview of Elon Musk.
• 0:03
  Speaker 2: I mean, in a nutshell, what this implies for SpaceX
• 0:10
  Speaker 2: annual recurring revenue by the end of 27 is somewhere
• 0:15
  Speaker 2: between 500 billion and a trillion.
• 0:18
  Speaker 1: On that crazy note, I.
• 0:19
  Speaker 2: Don't want to give Brad a heart attack. So we
• 0:25
  Speaker 2: are in an inflection moment for SpaceX across launch connectivity
• 0:28
  Speaker 2: and and superintelligence. Starting with Starship, Flight 14 was a
• 0:33
  Speaker 2: critical milestone. Starship made it to orbit for the first
• 0:36
  Speaker 2: time and deployed operational Starlink V3 satellites. For Flight 15,
• 0:42
  Speaker 2: which we expect to happen next month, we're going to
• 0:45
  Speaker 2: attempt to catch both the booster and the ship. If
• 0:50
  Speaker 2: we succeed, it will be the first time that an
• 0:54
  Speaker 2: orbital launch vehicle has been fully recovered. This is a
• 1:01
  Speaker 2: massive technology milestone. The only way that we can extend
• 1:07
  Speaker 2: consciousness beyond Earth on a sustained basis is with a
• 1:10
  Speaker 2: fully reusable orbital launch vehicle. I think this is a
• 1:17
  Speaker 2: profound technology milestone in the history of civilization. Success is
• 1:23
  Speaker 2: not guaranteed on these first flights, but I'd say it's
• 1:27
  Speaker 2: more likely than not that this succeeds next flight. We
• 1:30
  Speaker 2: expect Starship to fly weekly at some point in the
• 1:35
  Speaker 2: first half of next year and then second half of
• 1:37
  Speaker 2: next year twice weekly or maybe even more frequently than
• 1:40
  Speaker 2: that by the end of next year. The rate at
• 1:42
  Speaker 2: which we launch Starship will increase dramatically as we solve
• 1:45
  Speaker 2: the rapid and complete reusability.
• 1:48
  Speaker 1: So we'll have complete reusability.
• 1:50
  Speaker 2: Then at first there will be a little bit of
• 1:51
  Speaker 2: refurbishment that's needed between each flight. and will iron out
• 1:56
  Speaker 2: the need for any refurbishment such that you can immediately
• 1:59
  Speaker 2: land the rocket and launch it again just like an airplane.
• 2:03
  Speaker 2: As I said, this is a profound technology milestone that
• 2:09
  Speaker 2: most people in the world don't realize how important it is,
• 2:12
  Speaker 2: but it will be.
• 2:12
  Speaker 1: Obvious in hindsight. Starlink is also just getting started.
• 2:16
  Speaker 2: Flight 14 deployed 26 terabits per second of Starlink bandwidth,
• 2:23
  Speaker 2: that's equivalent to 10 Falcon 9 launches. So a single
• 2:28
  Speaker 2: Starship launch, which wasn't even fully loaded, it was about
• 2:33
  Speaker 2: maybe half loaded, and it did the equivalent of 10
• 2:37
  Speaker 2: Falcon 9 launches. So what I expect over time is,
• 2:42
  Speaker 2: not in the distant future, probably within three years, is
• 2:48
  Speaker 2: that we will have about 100 times the bandwidth of
• 2:52
  Speaker 2: the current Starlink constellation.
• 2:55
  Speaker 1: So because we will have launched.
• 2:59
  Speaker 2: Satellites with an order of magnitude more bandwidth and we
• 3:03
  Speaker 2: will place into orbit an order of magnitude more Starlink satellites.
• 3:08
  Speaker 2: So probably something on the order of 100,000 satellites, each
• 3:12
  Speaker 2: of which have more than 10 times the bandwidth of
• 3:15
  Speaker 2: the current Starlink system. And I think the demand for
• 3:19
  Speaker 2: bandwidth is going to be dramatically higher than it is today.
• 3:22
  Speaker 2: I think in three to five years, there's probably a
• 3:28
  Speaker 2: demand for a thousand times more bandwidth than there is
• 3:31
  Speaker 2: currently because AI or SI agents will, on a per-agent basis,
• 3:42
  Speaker 2: use far more bandwidth than a person, and there will
• 3:45
  Speaker 2: be far more agents than there are people.
• 3:47
  Speaker 1: So this ends.
• 3:49
  Speaker 2: Up requiring a several order of magnitude increase in communication
• 3:54
  Speaker 2: bandwidth in order to satisfy that communication demand. I can
• 3:57
  Speaker 2: say we do see a clear path to several hundred
• 4:02
  Speaker 2: billion dollars a year in connectivity revenue. So talking about
• 4:05
  Speaker 2: SpaceX superintelligence, SpaceX SI, sexy, whatever. So I think people
• 4:16
  Speaker 2: are starting to realize the constraint on deployment of of
• 4:22
  Speaker 2: superintelligence is more in power and cooling and.
• 4:27
  Speaker 1: Infrastructure than it is in chips.
• 4:32
  Speaker 2: So what we've done at SpaceX is we've taken a
• 4:36
  Speaker 2: team that has been able to achieve a market share
• 4:39
  Speaker 2: in mass to orbit that is almost 90% of all
• 4:44
  Speaker 2: mass to orbit of Earth, where China is about, I think,
• 4:48
  Speaker 2: roughly 6% and the rest of the world combined, including
• 4:52
  Speaker 2: the rest of the U.S., is about 4%. So a
• 4:55
  Speaker 2: team that has been able to achieve results that are
• 4:57
  Speaker 2: an order of magnitude greater than the rest of all
• 5:02
  Speaker 2: competitors in space combined and applied that same team to
• 5:07
  Speaker 2: the deployment of terrestrial compute. So this is a bit
• 5:11
  Speaker 2: like the New York Yankees playing in Little League.
• 5:16
  Speaker 1: It's not fair. But the net result.
• 5:19
  Speaker 2: is being able to deploy power cooling infrastructure at a
• 5:22
  Speaker 2: rate that is vastly faster than anyone else and may
• 5:28
  Speaker 2: at some point be faster than everyone else combined.
• 5:31
  Speaker 1: So I just let that sink in for a minute.
• 5:35
  Speaker 2: I've been spending personally a lot of my time in Memphis,
• 5:38
  Speaker 2: often four days a week, night and day, just getting
• 5:44
  Speaker 2: the Memphis Super Compute Cluster classes to online classes We're
• 5:50
  Speaker 2: tracking to be able to bring roughly 200,000 GB300s online
• 5:56
  Speaker 2: per month for every month for the remainder of this
• 5:59
  Speaker 2: year and continuing that pace next year, with next year
• 6:06
  Speaker 2: being primarily Ruby, very Ruby compute.
• 6:12
  Speaker 1: At this point, I would say I'm.
• 6:16
  Speaker 2: I would say I'm fairly confident of being able to
• 6:20
  Speaker 2: bring an incremental 10 gigawatts of compute online in 2027.
• 6:28
  Speaker 2: That would be in addition to the roughly three gigawatts,
• 6:32
  Speaker 2: maybe two and a half, three gigawatts that we'll have
• 6:34
  Speaker 2: online by the end of this year.
• 6:36
  Speaker 1: So I.
• 6:39
  Speaker 2: Would be surprised if we brought on less than 13
• 6:42
  Speaker 2: gigawatts of compute by the end of next year. So
• 6:47
  Speaker 2: to make sure that that is not a constraint for
• 6:49
  Speaker 2: our build-out, we're actually going to aim for an incremental
• 6:53
  Speaker 2: 20 gigawatts of power cooling infrastructure next year.
• 6:58
  Speaker 1: So there is a.
• 7:00
  Speaker 2: Non-zero possibility, maybe it's a small possibility at this point,
• 7:04
  Speaker 2: but it's not zero, of SpaceX having over 20 gigawatts
• 7:09
  Speaker 2: of super intelligence compute online next year. a small possibility,
• 7:14
  Speaker 2: but it used to be zero. It's not zero anymore.
• 7:18
  Speaker 2: We're casting the blades and vanes for the turbines. So
• 7:24
  Speaker 2: SpaceX already has a very advanced castings capability that was
• 7:30
  Speaker 2: developed for rocket engines. So we have, I think, actually
• 7:34
  Speaker 2: the most advanced casting capability on Earth because the demands
• 7:40
  Speaker 2: for castings for rocket engines exceed the demands for anything
• 7:44
  Speaker 2: else you could possibly imagine. So taking that incredible team
• 7:51
  Speaker 2: that has, I think, objectively the most advanced casting technology
• 7:55
  Speaker 2: in the world and saying, hey, can you guys also
• 7:57
  Speaker 2: make some blades and vanes for terrestrial turbines is much
• 8:03
  Speaker 2: easier than what.
• 8:04
  Speaker 1: They have done in the past.
• 8:07
  Speaker 2: And when we talk to the turbine suppliers, they tell
• 8:09
  Speaker 2: us that the blades and vanes are 12 to 18
• 8:15
  Speaker 2: months behind the rest of the turbine. Meaning they can
• 8:18
  Speaker 2: deliver everything except the blades and vanes 12 to 18
• 8:21
  Speaker 2: months before they can deliver the blades and vanes. So
• 8:26
  Speaker 2: casting is this really difficult bottleneck. And if anything, it
• 8:30
  Speaker 2: seems to be getting worse. So that means that we
• 8:34
  Speaker 2: can bring power online 12 to 18 months faster than
• 8:38
  Speaker 2: anyone else. That's the kind of advantage that's difficult to replicate.
• 8:44
  Speaker 2: Unless someone else is building an advanced casting shop, they
• 8:49
  Speaker 2: will be unable to compete with SpaceX in this regard.
• 8:51
  Speaker 2: But it's what we're aiming for internally, is to make
• 8:55
  Speaker 2: our first sort of test chip functional in approximately May
• 9:00
  Speaker 2: or June of next year. That's just kind of a
• 9:04
  Speaker 2: SRAM chip or something to basically test the line. That's
• 9:07
  Speaker 2: at the research fab in Austin. And we're aiming to,
• 9:11
  Speaker 2: our aspiration is to reach volume production around Q3 of
• 9:15
  Speaker 2: 28 at TerraFab at Givens Creek, the main TerraFab location.
• 9:22
  Speaker 2: So these are obviously like insanely fast time frames for
• 9:26
  Speaker 2: chip production. We do have, I think, some significant potential
• 9:32
  Speaker 2: breakthroughs in in chip fabrication, but I don't think I
• 9:37
  Speaker 2: should discuss that on an earnings call.
• 9:40
  Speaker 1: Let's not give away the shop here.
• 9:44
  Speaker 2: But I am optimistic that we will dramatically accelerate chip
• 9:49
  Speaker 2: production and that TerraFab will ultimately do what the name suggests,
• 9:58
  Speaker 2: which is to make a Terra of super intelligence chips
• 10:03
  Speaker 2: per year. On the GrokBot side, we're seeing the fastest
• 10:08
  Speaker 2: growth of a product that I've ever seen in my
• 10:10
  Speaker 2: life for GrokBot.
• 10:17
  Speaker 1: I've never seen anything grow this fast.
• 10:20
  Speaker 2: I think Rockpot may end up being what uses most
• 10:24
  Speaker 2: of our superintelligence compute. We are also, frankly, hounded relentlessly
• 10:32
  Speaker 2: by every hyperscaler and by dozens of smaller companies to
• 10:41
  Speaker 2: rent more of our compute. So we could rent out
• 10:46
  Speaker 2: our compute several times over if we said yes to everyone.
• 10:53
  Speaker 2: That's just the situation here.
• 10:57
  Speaker 1: That's why I posted that.
• 11:00
  Speaker 2: Video meme of the Monty Python cheese shop sketch, where
• 11:05
  Speaker 2: the guy goes into a cheese shop to get some cheese.
• 11:08
  Speaker 2: And they actually have no cheese. And so he goes,
• 11:11
  Speaker 2: what about this kind of cheese? What about that kind
• 11:13
  Speaker 2: of cheese? And they really just don't have any cheese.
• 11:18
  Speaker 2: we've actually tried to offload some of our inference compute
• 11:23
  Speaker 2: to other cloud providers so that we can use more
• 11:25
  Speaker 2: compute for training. And we get the cheese shop sketch.
• 11:29
  Speaker 2: You know, they'll say like, well, we'll ask them about,
• 11:31
  Speaker 2: what about some, you know, GB300s? Nope, don't have any
• 11:34
  Speaker 2: of those. What about, you know, some GB200s? Nope, don't
• 11:38
  Speaker 2: have any of those. What about some TPUs?
• 11:40
  Speaker 1: Nope. What about some trainings? Nope.
• 11:41
  Speaker 2: What about some, okay, how about some H100s? And like, well,
• 11:46
  Speaker 2: we might have some, overseas somewhere, and then they'll check
• 11:50
  Speaker 2: and come back and say, nope, we don't have that either.
• 11:52
  Speaker 1: So it's ridiculous. Yeah, so I don't think we're going
• 11:58
  Speaker 1: to have an demand issue.
• 11:59
  Speaker 2: And it's not impossible that we will have a six-month
• 12:03
  Speaker 2: payback period for compute next year. I mean, this is
• 12:06
  Speaker 2: definitely the kind of question where I feel like I'm taft.
• 12:08
  Speaker 1: Dancing through a minefield. So I have to be thoughtful
• 12:14
  Speaker 1: about what I say here.
• 12:18
  Speaker 2: I think... So both SpaceX and Tesla have formed special
• 12:25
  Speaker 2: committees to discuss this idea at the request of shareholders.
• 12:30
  Speaker 2: And these special committees consist of independent directors. And because
• 12:39
  Speaker 2: I'm a conflicted party, I can't actually be in the
• 12:42
  Speaker 2: discussions of the special committees.
• 12:44
  Speaker 1: So it's sort of a.
• 12:47
  Speaker 2: it's an odd kabuki dance, uh, where two, two, two
• 12:51
  Speaker 2: groups of people are talking and I, I can't be
• 12:54
  Speaker 2: in the conversation. It's, it's, I think some people out
• 12:57
  Speaker 2: there assume that a SpaceX and Tesla merger is, is
• 12:59
  Speaker 2: a certainty. And I think that would be overstating it. Uh, the, uh,
• 13:05
  Speaker 2: you know, do, do I think it would be a
• 13:07
  Speaker 2: good idea?
• 13:08
  Speaker 1: Yes.
• 13:09
  Speaker 2: Um, but at the end of the day, uh, we, we, the,
• 13:13
  Speaker 2: the shareholders have to be supportive of this and, um,
• 13:17
  Speaker 2: And particularly on the Tesla side, we will need a
• 13:20
  Speaker 2: 50% vote of all shares, not just those who vote,
• 13:23
  Speaker 2: but of all shareholders. And that's quite a high hurdle
• 13:27
  Speaker 2: to try to achieve. And I can just say at
• 13:35
  Speaker 2: this point that the special committees have not reached agreement
• 13:38
  Speaker 2: on the terms of any transaction. And obviously, there's no
• 13:42
  Speaker 2: guarantee that they will. But I do think there's a
• 13:46
  Speaker 2: lot of merit to combining the companies as we are
• 13:51
  Speaker 2: pretty much in the superintelligence singularity. And there's a lot
• 13:56
  Speaker 2: to be gained from cooperating on developing superintelligence software, chip design,
• 14:04
  Speaker 2: the Terafab, obviously, robotics, mass to orbit with the star
• 14:10
  Speaker 2: mind constellation.
• 14:12
  Speaker 1: Connectivity. Connectivity. So....
• 14:17
  Speaker 2: I think it would be a good idea, but it's
• 14:21
  Speaker 2: really up to the special committees of both companies. They've
• 14:25
  Speaker 2: got to come up with something and then that's got
• 14:27
  Speaker 2: to be put to the shareholders as a vote.
• 14:29
  Speaker 1: If you've ever thought about starting a YouTube channel or
• 14:32
  Speaker 1: you've got one, thanks for listening. See you in the
• 14:34
  Speaker 1: next episode.