Latest / In the Loop AI / Elon's Plan for Starlink to DESTROY Verizon, AT&T, and T-Mobile

Transcript

Elon's Plan for Starlink to DESTROY Verizon, AT&T, and T-Mobile · In the Loop AI

0:03
Josh: SpaceX just announced a $17 billion acquisition to essentially allow them to
0:08
Josh: shoot space lasers down from outer space direct to your cell phone to enable
0:12
Josh: you to have service anywhere in the world.
0:14
Josh: So if you're a user of T-Mobile, AT&T, Verizon, if you've ever experienced a dead zone in your life,
0:20
Josh: those days are soon to be behind you because what SpaceX did is they just purchased
0:23
Josh: mobile spectrum that will allow them to shoot these amazing Starlink satellites
0:27
Josh: that are in space, direct to your cell phone to receive service.
0:30
Josh: But that's not the only thing we are covering today.
0:32
Ejaaz: Yeah, apparently OpenAI is launching a LinkedIn competitor.
0:36
Ejaaz: There's a robotics company that's IPO-ing for $7 billion in China.
0:42
Ejaaz: And also Elon's not the only person that spent $17 billion this week.
0:46
Ejaaz: Microsoft also blew that exact same amount of money on a very niche European
0:51
Ejaaz: startup, which we'll get into later.
0:52
Ejaaz: But Josh, sorry, can we rewind a second into the laser beams coming out from space.
0:57
Josh: Yeah, okay. So SpaceX bought this thing called wireless spectrum.
1:00
Josh: So what is wireless spectrum and why did they pay $17 billion for it?
1:05
Josh: So wireless spectrum, I guess, Ejas, you could imagine it kind of like the air
1:08
Josh: being full of invisible waves, like radio signals or Wi-Fi.
1:11
Josh: We're kind of aware that this exists, but not quite sure how.
1:14
Josh: Spectrum is basically buying a chunk of these waves that companies use to send
1:18
Josh: data, like calls, texts, or the internet.
1:20
Josh: And because this is limited in bandwidth, there's a limited spectrum in which
1:23
Josh: you could send these radio signals.
1:24
Josh: The government actually controls who gets to use which parts so you can imagine
1:28
Josh: a close tie to like cbs or abc broadcast networks where the rights to actually
1:33
Josh: broadcast on a network are very expensive because again spectrum is very pricely
1:36
Josh: what spacex did is they bought the license to this spectrum from a company called
1:41
Josh: echo star that allows them the ability to actually,
1:45
Josh: use part of the spectrum and send cellular service direct to your mobile cell phone.
1:49
Ejaaz: Wow. So just to clarify, every two weeks, I go on a hike, Josh, right?
1:55
Ejaaz: And I mainly do it to get away from my girlfriend because she can't contact
1:58
Ejaaz: me. There's no cellular service.
2:00
Ejaaz: You're saying now with this new satellite service that Elon is launching with
2:04
Ejaaz: this new spectrum thing, I'll be able to get coverage wherever,
2:08
Ejaaz: even on a desolate mountain via a hike.
2:10
Josh: Exactly, yes, because this is leveraging the existing Starlink network that
2:14
Josh: we know and love that people use all over the world.
2:16
Josh: And I actually have a question for you because I found this super interesting,
2:18
Josh: EJS. How much of the world do you think is actually covered in cell service currently?
2:23
Ejaaz: It, I mean, come on, it's 2025. It can't be that much. I'm going to go with 15%.
2:28
Josh: It's over 50%. Over one half of the world at any given moment has no cellular service.
2:34
Josh: And what this deal does is it allows SpaceX to cover 100%. So anywhere in the
2:39
Josh: world, You could be in the middle of the Atlantic, the middle of the Arctic.
2:42
Josh: You will have cell service direct to cell. And I think that's what's noteworthy
2:45
Josh: about this is in the past, SpaceX has had to contract out this ability through
2:50
Josh: companies like T-Mobile.
2:51
Josh: So if you remember a few episodes ago, we spoke about the Starship and how this
2:56
Josh: will allow them to send up version three satellites of this,
2:58
Josh: which will give much more bandwidth to the network.
3:00
Josh: They enabled it through T-Mobile because T-Mobile owns the Spectrum license.
3:04
Josh: What they did today is they actually bought the rights to a license that allows
3:07
Josh: them to go around cellular companies. So they don't need to go through T-Mobile.
3:11
Josh: They can issue their own SpaceX network. So you can imagine a world in which,
3:16
Josh: Verizon and T-Mobile and AT&T,
3:18
Josh: well, they kind of suck. They have over 50% of the world as a dead zone.
3:21
Josh: But SpaceX, who will probably offer a service directly similar to a cell provider,
3:26
Josh: will have zero dead zones.
3:27
Josh: And what they bought is the rights to enable that to happen.
3:29
Ejaaz: You know, I'm just thinking about this. And with all the technological progress
3:33
Ejaaz: that we've made over decades, Facebook, social media, all that kind of stuff,
3:37
Ejaaz: I've never actually heard of someone contending with AT&T, T-Mobile,
3:41
Ejaaz: and all these other major carriers.
3:43
Ejaaz: Which brings me kind of to my next question. $17 billion is a lot, Josh.
3:49
Ejaaz: And I was looking at this tweet that said, you said half of it was in SpaceX stock, right?
3:55
Ejaaz: So that's around $8.5 billion. And SpaceX stock was, according to this agreement,
4:01
Ejaaz: was granted at $212 per share, which would value SpaceX at $400 billion.
4:06
Ejaaz: So my thought is like, do these figures seem right to you to get access to this spectrum?
4:12
Josh: Yeah, well, I would think you could probably imagine this as a relative valuation
4:16
Josh: compared, in the case of Starlink, so SpaceX is much more than just Starlink,
4:19
Josh: but if you're evaluating the Starlink network as a whole, which is owned by SpaceX,
4:23
Josh: you could use relative valuations based on actual cellular providers because
4:27
Josh: now this is direct competition.
4:28
Josh: So you could look at the comps being Verizon, AT&T, T-Mobile,
4:31
Josh: and that being just a small subset of the actual SpaceX business.
4:35
Josh: So think about what the market cap of these companies are, maybe even combine
4:38
Josh: that because I mean, now you have true global access.
4:41
Josh: So right now I'm traveling in Greece and Verizon doesn't exist here.
4:45
Josh: T-Mobile doesn't exist here. I'm using foreign network lines,
4:48
Josh: but in the world where SpaceX actually does offer direct-to-sell,
4:51
Josh: not only are the comps Verizon, but they are all the international cell providers as well.
4:55
Josh: So the global cellular service provider marketplace would be the comparable,
5:00
Josh: I guess, comparison to SpaceX now that they have this direct-to-sell capabilities.
5:05
Ejaaz: Okay, it sounds huge. And then where my mind naturally goes to next,
5:09
Ejaaz: Josh, is Elon's got to be making a cell phone right? Surely that's the next step.
5:16
Josh: I would bet no. I would like bet pretty strongly the answer is no.
5:19
Josh: Cell phones are very difficult to make at scale, nor do they contribute to the
5:23
Josh: mission of any of the companies.
5:25
Josh: So the idea of Starlink and SpaceX in general is to get people to Mars,
5:29
Josh: to make life interplanetary.
5:31
Josh: What Starlink is doing is funding this mission directly because it's adding
5:35
Josh: a lot of value to the company. It's adding a lot of value back to the world.
5:37
Josh: For them to build an actual smartphone, not only is it a logistics nightmare
5:41
Josh: to have to capture the entire supply chain and ship many of these,
5:44
Josh: but it just doesn't feel like a requirement when current hardware is so good.
5:49
Josh: I think I'd much prefer SpaceX give my iPhone incredible service than replace
5:54
Josh: my iPhone with something that hasn't been refined over two decades of time.
5:58
Ejaaz: Okay, I'm definitely going to take the other side of that. I've got this tweet
6:01
Ejaaz: pulled up here, which says, seems like we're trending ever closer to Elon making a phone.
6:07
Ejaaz: Battery supply chain synergy with Tesla. So that addresses your supply chain thing.
6:12
Ejaaz: AI-powered OS synergy with XAI. It's rumored to be working on it behind the
6:17
Ejaaz: scenes. XAI now officially owns X and all these other kinds of things.
6:20
Ejaaz: Inference chips locked down with the new AI6 with Tesla. And now telecommunications
6:26
Ejaaz: locked down with Starlink.
6:27
Ejaaz: Also, I remember Elon mentioning about a year and a half ago,
6:31
Ejaaz: should X basically create their own phone? So I think he's down for it.
6:35
Ejaaz: And then you're pointing around like, you know, this being like a massive slog
6:38
Ejaaz: for like hardware supply chain manufacturer. We just watched Elon pump in how
6:42
Ejaaz: many billions into this new Colossus 2?
6:45
Ejaaz: I think if there's any man to take over it, it's going to be him.
6:49
Ejaaz: I agree that it might not be a cell phone, but whatever the next major consumer
6:53
Ejaaz: hardware thing is, and that might be a version of a cell phone,
6:57
Ejaaz: I think Elon might take a shot at it. So I don't know. I'll take the other side.
7:01
Josh: Okay, because I was going to bet you my entire SpaceX stock that they would
7:04
Josh: not release a cell phone. I feel like very confident in the idea that they will
7:07
Josh: not release a cell phone.
7:08
Ejaaz: I will bet you dinner at the Tesla
7:11
Ejaaz: diner. I won't bet you my entire stock portfolio, but I'll bet you that.
7:16
Josh: Okay. All right. Well, I guess we'll leave it at that. Tesla is doing really cool things.
7:21
Josh: And what this means for the future, basically, is just like,
7:23
Josh: yeah, the entire world has cell service. And that's pretty freaking exciting.
7:26
Josh: So that is not the only news this week, right?
7:28
Josh: I'm actually away. I'm on vacation. I've been mostly offline this week.
7:31
Josh: So I am leaning on you to guide us through what is going on.
7:35
Josh: I'm going to learn kind of as the audience member, I think, for most of this
7:37
Josh: episode. So fill me in what we have next.
7:39
Ejaaz: Okay, so what you're seeing on my screen is a tweet from someone called Fiji
7:43
Ejaaz: Simo. And if that name sounds familiar to you.
7:46
Ejaaz: Yeah, yeah, she is the CEO of OpenAI Applications.
7:52
Ejaaz: Not to be confused with the CEO of OpenAI. Okay.
7:56
Ejaaz: And just a bit of background. And Fidji is a tenured employee at Meta where
8:01
Ejaaz: she actually helped build a bunch of the consumer applications at Meta.
8:04
Ejaaz: And then she recently left and joined OpenAI where she's leading basically the
8:08
Ejaaz: app build out, the app store build out, all the cool different AI apps that you can get access to.
8:13
Ejaaz: And she's just announced the major flagship app that OpenAI is going with,
8:18
Ejaaz: which is AI LinkedIn. in.
8:22
Ejaaz: It's essentially a new platform that will allow businesses or anyone,
8:27
Ejaaz: quite frankly, to connect with AI proficient individuals.
8:32
Ejaaz: So what I mean by AI proficient individuals is, Josh, you and I use ChatGPT, right?
8:36
Ejaaz: And I would say we're fairly good at it. We kind of know how to prompt certain
8:40
Ejaaz: ways and know how to ask it for certain things. We kind of get the vibe.
8:44
Ejaaz: But I'm talking about a layer deeper where there's people that actually learn
8:48
Ejaaz: how to structure prompts for various different situations and tasks that literally
8:52
Ejaaz: dedicate 50 to 70 to 80% of their time every day learning and trying to become
8:58
Ejaaz: proficient in these things, kind of like trying to cut down
9:01
Ejaaz: Waste at companies and make it more efficient.
9:03
Ejaaz: All the type of people that create startups from zero to one that create millions
9:08
Ejaaz: and millions of dollars per month like we're seeing all over social media.
9:13
Ejaaz: So this new AI LinkedIn is an interesting concept because it's called the OpenAI Jobs Platform.
9:19
Ejaaz: And it's basically, like I said, it's a place where businesses can connect with
9:23
Ejaaz: certain different kind of workers. And the main problem that they're trying to fix here, Josh, is
9:28
Ejaaz: Everyone has AI tagged onto their resume at some point right now.
9:32
Ejaaz: But it's very hard to tell whether someone is actually kind of like proficient in that thing.
9:37
Ejaaz: And so the other way that this platform is trying to kind of like surface high
9:41
Ejaaz: quality individuals and this problem that OpenAI is trying to fix is they're
9:45
Ejaaz: linking this in with their OpenAI Academy, which is kind of like a Khan Academy.
9:51
Ejaaz: If you guys have heard of it, it's kind of like an online academy where you
9:53
Ejaaz: can kind of like learn various kinds of things, software engineering,
9:56
Ejaaz: maths courses, all that kind of stuff. but dedicated towards AI.
9:59
Ejaaz: And why I think this holds such esteem for OpenAI in particular is they're putting
10:04
Ejaaz: their full brand and weight behind it, Josh.
10:06
Ejaaz: Like I said, this is the first application that they're launching.
10:10
Ejaaz: And the final point I'll make is when I zoom out, right, we have said so many
10:15
Ejaaz: times on this show that AI is going to replace everything.
10:19
Ejaaz: But we don't specifically know how. This is kind of like this big ambiguous
10:22
Ejaaz: vision that, you know, hundreds and hundreds of billions of dollars is being invested in.
10:27
Ejaaz: We don't know quite how it materializes just yet in each different industry.
10:31
Ejaaz: We're starting to see it.
10:32
Ejaaz: I think OpenAI's job platform is going to be one of the main ways that facilitate this.
10:38
Ejaaz: And I was thinking about the kind of like power that they're going to hold, right?
10:41
Ejaaz: They're going to have basically all the highest influential experienced AI engineers,
10:48
Ejaaz: researchers, product people, all within one platform. And I can't help but think
10:51
Ejaaz: that that data is very useful.
10:52
Ejaaz: But I'm going to pause for a second. Josh.
10:55
Ejaaz: Hot take, initial take.
10:57
Josh: I think it's interesting. The most exciting thing to me is the last line of
11:02
Josh: this, actually, and OpenAI certified for workers to learn and demonstrate their AI skills.
11:06
Josh: I think oftentimes a lot of companies and a lot of employees,
11:09
Josh: they might think they know how to use AI, but there's actually a lot of ways
11:13
Josh: to use AI and creating some sort of a benchmarking system for individuals relative to models.
11:18
Josh: So the way we compare benchmarks with normal models, or the way we compare normal
11:22
Josh: models is using benchmarks. Well,
11:24
Josh: we don't really have those for actual employment productivity skills.
11:29
Josh: Like if you've ever done this back in high school or early days of college,
11:33
Josh: you can get certified in like Microsoft Excel or Microsoft Word,
11:36
Josh: just kind of prove that you know it, which felt very lame.
11:38
Josh: I mean, this is not what I want to see open-air working on, but I guess I'm
11:42
Josh: kind of torn because it feels like it solves an interesting problem,
11:44
Josh: right? Because we do have this employment problem.
11:46
Josh: We do have the existential idea that AI is going to replace jobs.
11:50
Josh: The reality is it probably creates a lot more jobs. And in that world,
11:54
Josh: does having this job platform make sense?
11:57
Josh: Probably. I think so. Because it will, I mean, they're the closest to the metal
12:01
Josh: in the sense that they are the closest to the actual model.
12:03
Josh: They have all of the user data. They can certify who is valuable.
12:07
Josh: It's an interesting first example, and I'd love to hear more about why they chose to do it.
12:13
Ejaaz: I agree. So if I were to summarize my thoughts, the bare case is that it seems pretty uninspired.
12:19
Ejaaz: Like LinkedIn is kind of whatever to me, at least.
12:23
Ejaaz: And so I thought they might come up with something novel and AI native that
12:26
Ejaaz: would wow people, like something magical, right? Like ChatGPT was.
12:30
Ejaaz: But the bull case, I think, is something that you just mentioned,
12:34
Ejaaz: which is to date, benchmarks have been kind of very binary and very gameable.
12:40
Ejaaz: And the only real way to see how AI can impact the real world is through humans.
12:46
Ejaaz: That's why AI agents haven't really exploded just yet because you need the human component.
12:51
Ejaaz: You need the intuitive component. And the only way to do that is via humans.
12:55
Ejaaz: So I think this is one step of many. Potentially, like, you know,
12:59
Ejaaz: create this platform, create a bunch of different new job roles and job types.
13:02
Ejaaz: And then industries at large start to understand how they can effectuate change using AI.
13:08
Ejaaz: It becomes much more realistic. Like, oh, this is what an AI product manager
13:12
Ejaaz: does or, oh, this is what an AI-enabled engineer can do and this is how much they're worth.
13:16
Josh: It's exciting for the, I guess, the business world. If you are a business owner,
13:20
Josh: if you're someone in the job market, this is probably interesting to you.
13:23
Josh: It probably takes time to get seated and actually become valuable.
13:27
Josh: But selfishly, I do wish the product team at OpenAI was releasing more products
13:32
Josh: that we could use as customers.
13:34
Josh: I want some, like you mentioned, I want some improvements. I want some new tools.
13:37
Josh: I want to be able to further leverage my skills instead of prove that I have skills.
13:42
Josh: So I guess this one falls a little flat for me, but I could see why it's very important.
13:45
Josh: And I hope that the future will lead it to be much more important than I hope for.
13:50
Josh: But there's more, right? There's more ChatGPT news this week or OpenAI news in general.
13:54
Ejaaz: Yeah, so I think this one's going to make you happy, Josh. OpenAI revealed, by popular request,
14:02
Ejaaz: you can branch conversations in ChatGPT now, letting you more easily explore
14:07
Ejaaz: different directions without losing your original thread. it.
14:10
Ejaaz: So for the listeners of this show, imagine you open up a conversation with ChatGPT
14:15
Ejaaz: and you're asking about, you know, travel plans to go to, you know,
14:20
Ejaaz: this option, that option, or maybe asking to write an essay for you, right?
14:24
Ejaaz: For high school, which I know a lot of you guys are doing.
14:28
Ejaaz: And you kind of think, okay, I wonder how this essay might have been written
14:33
Ejaaz: if I had prompted it differently, or if I had asked it to take a particular
14:38
Ejaaz: different tone for this essay.
14:40
Ejaaz: Well, now you can scroll back up to where you asked the initial prompt or wherever and click branch.
14:46
Ejaaz: And it opens up a new tab with a brand new conversation, but with all the same
14:51
Ejaaz: context that you've already fed ChatGPT.
14:53
Ejaaz: Now, for those Marvel fans out there, specifically Marvel Avengers,
14:58
Ejaaz: you know or are familiar with parallel universes, right? Many different timelines
15:03
Ejaaz: for many different possibilities.
15:05
Ejaaz: This is effectively what OpenAI has just enabled for your conversations.
15:09
Ejaaz: And I love this personally because I can now run many different experiments
15:13
Ejaaz: and see which one comes to the best answer. Josh, what do you think?
15:18
Josh: Yeah, this is great. See, this gets me excited. This is something you could
15:21
Josh: go into ChatGPT today and use and it will improve your experience.
15:24
Josh: It'll improve your quality of life. I think it's great.
15:27
Josh: What's the funny thing that always drives me crazy is that you can use the same
15:30
Josh: model and share the same prompt and it will spit you out a different answer
15:33
Josh: every single time without fail.
15:36
Josh: And the ability to now fork off the conversation and to try something different
15:39
Josh: while maintaining that stream of thought and more importantly,
15:42
Josh: the context prior in the conversation, that to me is a really cool use case.
15:46
Josh: Because there are oftentimes where I'm exploring an idea or I'm doing some research
15:50
Josh: and I want to go off in one direction.
15:52
Josh: But by adding the additional context of the future answers, it throws off the
15:56
Josh: further streams of directions that I want to go.
15:58
Josh: So being able to cut it, fork it, and then try multiple times, that's fun.
16:03
Josh: I like that. This is good news. This is good. OpenAI, thumbs up.
16:07
Ejaaz: Agreed. Well, there is another big up that happened this week,
16:12
Ejaaz: and that is Sam Altman casually announced that there's been a change,
16:19
Ejaaz: Josh. There's been a slight change.
16:21
Josh: Oh, a change?
16:22
Ejaaz: That's new. Initially, OpenAI had planned to spend
16:27
Ejaaz: along the lines of maybe $30 to $40 billion on compute over the next couple
16:33
Ejaaz: of years, which for a single company, by the way, is a staggering amount of money.
16:38
Ejaaz: That is like, whatever, That is like something like 40% of Meta's cash reserves.
16:44
Ejaaz: And Meta is a very mature public company that has produced many valuable things for the world.
16:50
Ejaaz: Open AI, still private, still kind of working on their things,
16:54
Ejaaz: still working on the consume applications.
16:55
Ejaaz: Sam Altman announced this week that, yeah, we need to up that by around 80 to $100 billion.
17:02
Ejaaz: Yeah, we need more compute.
17:04
Josh: Whoa. So that puts them, what, over $100 billion of spending now for a company?
17:09
Ejaaz: $150 billion to be specific.
17:12
Josh: For a nonprofit private company. That is incredible.
17:16
Josh: So like, I have a lot of thoughts going through my mind. I'm curious to hear
17:19
Josh: yours, but my first take is like, oh my God, if we don't get AI or AGI with
17:23
Josh: $115 billion of processing power through the end of the decade,
17:27
Josh: we have a serious problem.
17:28
Josh: So this is what we were talking about this briefly with the Colossus project
17:33
Josh: where they're planning to put, the XAI team is planning to put a million GPUs online coherently.
17:37
Josh: We are approaching numbers that are truly astounding.
17:41
Josh: Like $115 billion of spend is millions and millions of GPUs that you assume
17:46
Josh: will be coherent that should produce models that are exponentially better than what we have today.
17:51
Josh: So assuming this works, the future, there's no, there's no way we don't get AGI, right?
17:57
Josh: Do you agree? What do you think? It's a huge amount of money.
18:01
Ejaaz: Yes, it's a huge amount of money. And there are basically two camps of thought around this.
18:07
Ejaaz: I call it the bear camp and the bull camp.
18:10
Ejaaz: The bear camp basically thinks this is all a bubble, specifically something
18:15
Ejaaz: called the AI capex bubble.
18:17
Ejaaz: Basically, people are spending way too much on this AI infrastructure thing,
18:21
Ejaaz: and there's no goods to show.
18:23
Ejaaz: You could also argue that Fiji's announcement of AI LinkedIn for open AI kind of proves that.
18:29
Ejaaz: You could also argue that GPT-5 launching and kind of being, yeah, nothing magical.
18:34
Ejaaz: Maybe I'm desensitized to the magic of AI. You could also use that as a proofing point behind that.
18:40
Ejaaz: The bull camp says that, no, there's going to be actually an over demand for
18:45
Ejaaz: compute and not enough money is actually being spent on it, right?
18:49
Ejaaz: So there's these two camps.
18:51
Ejaaz: $80 billion is a massive swing.
18:54
Ejaaz: Sorry, it's not even a swing. It is a completely new budget, right?
18:59
Ejaaz: It is over like 70% of the originally termed budget.
19:04
Ejaaz: So this could be, we could like speculate on this.
19:07
Ejaaz: Like my view is, I think Sam sees something that we don't, which is going to
19:13
Ejaaz: require a lot more compute.
19:14
Ejaaz: Maybe this is the new hardware consumer device that he's going to launch.
19:17
Ejaaz: Maybe he's realized a new architecture,
19:20
Ejaaz: which is different to transformer architecture to train a new model.
19:23
Ejaaz: Well, maybe he's just realized, but to get to the next step change,
19:27
Ejaaz: AGI, you need this amount of compute.
19:29
Ejaaz: That's my bold case. My bad case is he's kind of pulling this number out of
19:33
Ejaaz: his ass and we're not going to get it anywhere.
19:35
Josh: Yeah, I hope for the latter. I hope that, or I guess the former,
19:39
Josh: which was that OpenAI actually has a lot more left in the tank.
19:43
Josh: And GPT-5 was kind of an anomaly in the sense that they optimized for things
19:46
Josh: that maybe weren't quite popular with the masses, but were very important to the business.
19:50
Josh: So I think retroactively looking back at GPT-5, it was optimized for cost per
19:55
Josh: token, really, and trying to optimize for quality per token by using a router.
19:58
Josh: I think they understood this misstep and now they're going for the raw power.
20:03
Josh: And I'm hopeful that there actually is no scaling wall and that there is a direct
20:08
Josh: correlation between dollars in and intelligence out.
20:12
Josh: Now the question, and I would love comments too for the people who are listening
20:15
Josh: to share, is this actually a CapEx bubble?
20:18
Josh: Is there a world in which they can derive greater than $115 billion of profit
20:23
Josh: by the year 20, or after 2029, but that's how much they're spending before 2029,
20:28
Josh: and actually earn a return on this investment?
20:30
Josh: That is where I don't know. Because at the end of the day, they're competing
20:33
Josh: for what will probably be a commodity, right?
20:36
Josh: We have Google is going for the same thing, XI is going for the same thing,
20:39
Josh: and maybe there will be hyper-specific models.
20:41
Josh: Just last week, we talked about the GPT-4B model that is good for biometrics
20:45
Josh: for these protein synthesis and extending life.
20:48
Josh: You don't need $115 billion of
20:50
Josh: processing for that. So that's the big question. Is this a CapEx bubble?
20:53
Josh: We don't know. But we are, I'm stoked to find out because it's not my problem.
20:58
Josh: They're privately held. We don't own any shares.
21:01
Josh: So hey, do your thing. And if you give us AGI for $20 a month, sign me up.
21:05
Ejaaz: I think that AI is the most crucial thing to be spending money on.
21:09
Ejaaz: So regardless of whether they know we're in a bubble or whether they don't know,
21:13
Ejaaz: I think the money is worth spending because we're going to have some kind of discovery from this.
21:18
Ejaaz: And the clearest signal for the bull case of us not being in an AI CapEx bubble
21:23
Ejaaz: came from Microsoft this week. Josh, did you see the news?
21:28
Ejaaz: So they just plowed in...
21:30
Josh: No, a multi-billion dollar agreement. Another day, another couple billion. Yep, yep.
21:34
Ejaaz: They just plowed in $17 billion. So the same amount Elon just dropped on his
21:39
Ejaaz: new Spectrum provider on this company called Nebius. And the main reason why
21:46
Ejaaz: is to acquire more compute.
21:49
Ejaaz: Nebius is a data center provider, which is primarily located in Europe,
21:54
Ejaaz: and they specialize in hardware data centers.
21:57
Ejaaz: I'm talking cold metal bare bones infrastructure out in God knows where in Norway or wherever,
22:05
Ejaaz: energy proficient, pure power to help Microsoft surface and train their models
22:12
Ejaaz: that they're doing in conjunction with OpenAI, but also to service some of the
22:15
Ejaaz: features that they have that they offer to clients, such as Microsoft Copilot.
22:21
Ejaaz: One striking take from this bit of news is, well, okay, let me ask you this, Josh.
22:28
Ejaaz: How much do you think Nebius was valued prior to this announcement?
22:32
Josh: I'm not even going to guess. I'm so far out of the ballpark in these numbers
22:35
Josh: now. I have nothing to even give.
22:38
Josh: It could be a million dollars. It could be a hundred billion dollars at this point. Got it.
22:42
Ejaaz: That total market cap, they're a public company, I believe, is $15 billion.
22:51
Josh: Whoa. So that's a very low multiple that they purchased them for.
22:54
Ejaaz: Yeah, yeah, exactly. So Microsoft plowed in. Well, this is the thing.
22:57
Ejaaz: It's not a purchase. It's a partnership.
23:01
Ejaaz: $17 billion of consumer revenue is going to Nebius.
23:06
Ejaaz: So over the next, I think it's like three to five years, Microsoft is paying
23:12
Ejaaz: them basically $17 billion.
23:14
Ejaaz: So someone did the math looking at that. I don't know if I have the tweet.
23:19
Ejaaz: Oh yeah, it's right here.
23:20
Ejaaz: Nebius is projecting $625 million of revenue by the end of 2025.
23:26
Ejaaz: Microsoft's revenue alone would increase their revenue by a growth of 3x per upon year.
23:32
Ejaaz: So this might be the best stock purchase or for people who are holding Nebius
23:36
Ejaaz: are up bigly basically in the last 24 hours.
23:41
Josh: Congrats to Nebius holders. And I guess congrats to Microsoft holders as well
23:44
Josh: because it stocked it very well as soon as this was announced, right?
23:47
Ejaaz: I don't know how much Microsoft shares went up, but what I do know is that Nebius
23:52
Ejaaz: shares jumped 53% in after hours, Josh. So we have another data AI spend.
23:59
Ejaaz: And going back to my original point, I think this is a proving point against the AI CapEx bubble.
24:04
Ejaaz: Because if Microsoft, a titan that has spent, I think it's, what is it,
24:09
Ejaaz: $35 billion that was invested in open AI when Sam Altman was ousted, right?
24:13
Ejaaz: To acquire the rights to basically all open AI models till 2030,
24:16
Ejaaz: is spending even more money on another data provider.
24:19
Ejaaz: And remember, they're like the primary compute providers via Azure to train
24:24
Ejaaz: open AI as models. if they're spending a further $17 billion to acquire more
24:28
Ejaaz: compute capacity, this is a sign that it's not a bubble and that we're actually
24:33
Ejaaz: building something useful.
24:34
Josh: There's a lot of spending. And this is not where the spending ends.
24:37
Josh: For our last story, we have one more acquisition merger too,
24:40
Josh: correct? And this one is also fairly large?
24:43
Ejaaz: Oh, yeah. Okay. So as you know...
24:46
Josh: Tell me about the browser company.
24:47
Ejaaz: Let me tell you about the browser company. Well, actually, let me tell you about
24:50
Ejaaz: our views on the browser company.
24:52
Ejaaz: So for context, Josh and I have done, I think, about three episodes so far on
24:58
Ejaaz: these things called AI browsers.
25:00
Ejaaz: Think of your internet browser, Google Chrome, Safari, Firefox for you crazy
25:03
Ejaaz: kids out there, paired with AI technology.
25:07
Ejaaz: So having a live-in chat GPT-like assistant that kind of like knows what you're
25:12
Ejaaz: browsing, knows what you're looking for, surfaces all your results,
25:15
Ejaaz: combines all these videos and summarizes it for you in one slick little interface.
25:21
Ejaaz: It kind of reads your brain before you even type any keys. Um,
25:25
Ejaaz: I think Josh and I were both kind of bearish on the AI browser.
25:29
Ejaaz: And the reason why, and Josh, please jump in here, is we kind of thought that
25:34
Ejaaz: the interface was a cop-out.
25:36
Ejaaz: We think that AI is going to enable a completely different type of OS operating system.
25:41
Ejaaz: And we don't think browsers are quite the same medium. Josh, do you agree with this?
25:45
Josh: Yeah, we spoke about this at length with Arvind, the CEO of Perplexity.
25:48
Josh: So if you have not seen that episode, go check it out.
25:50
Josh: It was interesting to hear the other side from someone who's actively building
25:54
Josh: what I would assume is one of the largest AI browsers in the world.
25:56
Josh: But the take that I had was that, well, when I'm using a browser,
26:00
Josh: I use it for two purposes, right?
26:01
Josh: I use it, one for leisure, which is watching YouTube videos or scrolling Twitter
26:05
Josh: or whatever type of engaging thing that I wanna be doing.
26:08
Josh: And for me, I want to do that. I want to personally watch the videos.
26:11
Josh: I don't want a bot doing that for me.
26:13
Josh: And then the other bucket that I use a browser for is productivity.
26:15
Josh: So I use it for working on agendas with you for the show notes or doing research
26:19
Josh: or writing papers or whatever random things it may be, or even booking a trip
26:23
Josh: like I did for this trip that I'm on right now.
26:25
Josh: But I don't really care to do that part.
26:29
Josh: Like if I can offload that to an agent to go and book my flights to give me
26:34
Josh: help on this prep and just present me with the finished thing,
26:37
Josh: I would much prefer an agent obfuscate away all of the complexities of the browser
26:41
Josh: and just present me with what I want.
26:43
Josh: I don't really care for the browser interface that is now over 25 years old, I believe.
26:49
Josh: When an agent could just go and do the engaging itself, and then return me the results.
26:53
Josh: So for that reason, we've been bearish on AI browsers, but clearly we are,
26:57
Josh: I mean, to some extent, the market's signaling we're wrong.
26:59
Josh: I don't know if we're wrong, but according to this acquisition,
27:01
Josh: the market is giving a signal that like maybe people actually do want AI browsers.
27:04
Ejaaz: Yeah, so the headline here is AI browsing company called The Browser Company
27:10
Ejaaz: just got acquired for $610 million.
27:16
Ejaaz: All cash by none other than Atlassian, which is a enterprise software product,
27:21
Ejaaz: a software as a service company famous for their product, Jira, right?
27:26
Ejaaz: I actually caught up with the guy that set up this entire deal,
27:31
Ejaaz: who is the number two at Alassian. He's my former boss, Sanchen.
27:35
Ejaaz: And he basically said that the reason why they bought it is they are betting
27:41
Ejaaz: big on the browser, but not at a consumer level, at the enterprise level.
27:47
Ejaaz: They think that a number of their clients are going to want to have their own
27:51
Ejaaz: in-house browser that they can use for their own data set, for their own operating
27:57
Ejaaz: system, for their own products and systems.
28:00
Ejaaz: And what he wants to do with part of the browser strategies company is build
28:06
Ejaaz: out this kind of like enterprise grade browsing solution.
28:09
Ejaaz: How much revenue do you think the AI browser company, sorry,
28:12
Ejaaz: the browser company is making?
28:14
Josh: Zero dollars and zero cents.
28:17
Ejaaz: Um, you are accurate.
28:19
Josh: Yeah, that's, that's insane. Uh, $0 of revenue and $610 million in acquisition
28:25
Josh: cost is crazy. Congratulations to the team.
28:28
Josh: Uh, that is an unbelievable accomplishment for the Atlassian company.
28:33
Josh: Oh my gosh, that feels scary.
28:35
Josh: Um, you are selling B2B products in a world where
28:38
Josh: software will be hyper-customized and
28:42
Josh: to productize this as a
28:45
Josh: generic product that you can customize the data for and then
28:47
Josh: serve to an employee base a company
28:51
Josh: seems scary and risky when there
28:53
Josh: will be a world in which you can create a prompt that generates
28:56
Josh: an entire software suite custom to
29:00
Josh: each company so this is an interesting bet i mean jira is kind of what a project
29:06
Josh: management software for like yep corporate b2b services it's kind of slow it's
29:10
Josh: kind of clunky yep yeah we use asana also kind of slow kind of clunky it's just
29:14
Josh: it feels like this this outdated,
29:18
Josh: type of form factor and business plan when the new and exciting thing feels like,
29:24
Josh: hey, we actually just have an incredible AI service that can write unbelievable
29:28
Josh: code and create a custom software stack with all your data for exactly how you
29:31
Josh: need it without needing to use a browser.
29:34
Josh: It's interesting. I mean, they are clearly well-researched if they have that
29:38
Josh: much money to spend on a company that is making zero revenue.
29:42
Josh: So it's an interesting purchase.
29:44
Ejaaz: And that's it for this episode, folks. It has been a crazy week so far,
29:49
Ejaaz: and I have a feeling this is just the start.
29:52
Ejaaz: Remember, this is Tech-temper, and in under an hour from now,
29:57
Ejaaz: Apple is streaming live about their latest hardware and software,
30:01
Ejaaz: and we have our biggest Apple fanboy on this show, so you know we're going to
30:05
Ejaaz: get you a hot episode out very soon. Josh, any initial takes?
30:09
Josh: Guys, this is my Super Bowl. This is a big deal. If you don't tune into the
30:12
Josh: episode this week, you are going to miss out on some pretty spicy takes.
30:15
Josh: There are not many companies that I've followed for this long this closely
30:18
Josh: and care this passionately about than Apple and this is very
30:21
Josh: much the Super Bowl event of the year it's the new iPhone it's the
30:24
Josh: new watch it's new AirPods and if we are lucky God willing we will get some
30:28
Josh: AI news as well so there's a lot riding on this the future of Apple depends
30:33
Josh: on it we are going to go tune out and tune into the live stream watch that and
30:37
Josh: then report back to you shortly so if you are listening to this tune in it's
30:40
Josh: coming soon it's going to be a great episode but in the meantime I hope you enjoyed this one.
30:43
Josh: Thank you so much for listening. Don't forget to like, subscribe,
30:45
Josh: share with your friends, all the other great things that help our show grow.
30:49
Josh: We really appreciate you sticking to the 33 minute mark. This was a little bit
30:52
Josh: longer, nice little roundup. Hope everyone enjoyed.
30:54
Josh: And I guess for the fun comments, I've been loving the comment section.
30:57
Josh: I've been trying to reply to a lot of the comments, and people are really engaged. They're great.
31:00
Josh: So if you have anything to say, please don't forget to leave a comment.
31:04
Josh: If anything, just let us know if you think it's a bubble or not.
31:06
Josh: Are we in the CapEx bubble or are we not?
31:08
Josh: I think this is kind of a point of contention we go back and forth on.
31:10
Josh: So I'd love to hear your takes. But anyway, thank you so much for listening.
31:14
Josh: We will be back tomorrow with our lovely and exciting, hopefully exciting Apple episode.
31:18
Music: Music