how to design Highly scalable distributed message queue like Kafka?

[MUSIC] [MUSIC] [MUSIC] >> Welcome everyone to the information's TITB. My name is Akash Pasfreech. It is Friday, June 5th. We start today with some new revealing forecasts for SpaceX, which is angling for the biggest IPO ever next week. We'll break down our exclusive reporting on new numbers that Goldman Sachs has modeled out for the company with our Deputy Bureau Chief of Finance. We'll then dive deeper into the Cat and Mouse game playing out between andthropic and XAI. We've got exclusive new reporting on the relationship between the two companies. And we'll close out the show with our scoop that Data Center developer Switch is in talks to raise money at evaluation of at least $50 billion. We'll look at what that means for the broader Data Center capacity crunch. It's going to be a fun Friday show, so let's get right on into it. SpaceX is planning what will be the biggest IPO of all time next week most likely. But the information has exclusive reporting that Wallsery analysts suggest the company might need to raise even more money in the years to come. I want to bring on Deputy Bureau Chief of Finance, Cory Weinberg, to break down his reporting. Cory, welcome back to the show. It's great to have you here. So SpaceX is going to go out. It's going to try to raise $75 billion. And your reporting suggests that Wall Street thinks they're going to have to raise even more? That's exactly right. These are numbers that the investment banks are for catap things, specifically Goldman Sachs. And I've reported on these types of numbers before in the run up to sort of significant IPOs. It's pretty standard practice for the analysts at the investment banks to get their cues from the companies themselves on what their forecasts are. And then the analysts and the bankers at the investment banks essentially pass those on verbally to sort of accredited investors in the IPO. And then we get our hands on them. And usually it's an interesting exercise because who can really see the future? Are these forecasts super-regressive? Are they conservative? Regardless, the company is going to be held to this standard. And what I found was interesting about these numbers was how revealing they were in terms of just how much money Wall Street expects SpaceX to burn in the coming years. So much so that the largest IPO of all time won't get them through 2028 if their cash per number are correct. You know, it's funny, Corey, because we don't know how much goes into these models. And these are a lot of conversations, a lot of number crunching. But really what I'm imagining is some senior banker looking at the forecast and looking at some growth rate for expenses or something. And maybe a little something around the line. This looks a bit low. Let's raise it a bit. I mean, it's a modeling exercise really at the end of the day, right? It's a modeling exercise. But usually the analysts and the banks want to be within some sort of realm of consensus. They want to be around where the company themselves have told them that they're modeling. And usually they tweak it a little bit. And they of course have their own intellectual assumptions around the numbers. But yeah, no, it's the reason why I think folks should take these numbers, at least somewhat seriously. And it's not just an exercise to be like, well, they're just trying to pump up the IAPO. They're just trying to make everything look really good. Is it because after they go public, we're going to be looking, everyone's going to be looking back at these numbers. The banks are going to publish these numbers after the IAPO and the company, you know, SpaceX and then obviously any other company that goes through this are held in some ways to this standard. You know, so they are worth taking seriously. So let's talk about then the details of this forecast. So most of this burn is coming from the AI business and the capital investments that the company's going to have to make. Yeah, 100%. I think about three quarters to 80% of the capex is for AI in the coming years. Right. It's a much notable because there's a bunch of things that are spending money on their starship, their star like. And how do you figure the SpaceX would then raise this additional sum of money than after the IPO? I mean, probably the debt markets or they would tap the equity markets again. I know it in the story that in conversations with Wall Street investors, SpaceX executives, including Elon Musk has sort of played down the idea that they would raise even more equity and dilute shareholders further after going public. We'll see if that's actually true. And we'll see what they're sort of what the relative ease of raising debt versus equity would be would be for them going public. And so these are the latest forecasts that we have from the bank. What's your from from up and I should say, but the lead bank for the lead bank. Okay. So it's you know, it's what's setting the pace here. What is your what's the temperature then on? I mean, these are the numbers. What's the temperature on how analysts feel about these forecasts? What investors are feeling? Was this sort of an unwelcome surprise to them or were they expecting this? Yeah, I talked to one in a large buy side investor, you know, who said like, oh, obviously, our people are doing our own models based on the S1 based on feedback from the company. Like we're going to do our own work, not just listen to the banks. And basically, you're like, yeah, it was like to everything was 10 times larger. We thought or not. Everything was a lot larger than we thought. Revenue's cash burned, you know, they sort of turned up the dial on these numbers because let's be let's talk about some big numbers. Like let's let's sort of set the foundation here. Last year's face that's had about 8. 18.6 billion in revenue. It burned about 14 billion. So that's really free five. It's revenue is expected to more than double this year. That's in part thanks to Elon's new friends in Vropek and they're renting out colossus compute. But revenue is expected to go to about 38 billion cash burn also more than doubling to about 30 billion. And then let's fast forward to the end of the decade or to the start of the next decade in 2030. Revenue is expected to be 474 billion. So up from 18 billion this past year. So that's quite the jump. And then over that time period they're expected to burn $350 billion. That's largely from CapEx due to AI. That's from data centers in space. That's from TerraFab. So yeah, the numbers are really big. So 75 billion of what they're hoping to raise in the IPO. The 350 billion that they are slated or forecasted to burn through 2030. Did the models have any indication in there? I mean, on the 75 billion, how much they might seek to raise then in the years after? I mean, they're not going to be able to raise. They're not that level of detail in modeling. And like, let's be clear, these are coming from, there's like multiple steps removed here. Remember, like the company is forecasting what they think are going to happen. The banks then taking those cues and making their own assumptions about what's going to happen. And then the banks telling the investors verbally. These are not like, these are not numbers handed out on a page across Wall Street, leafletted around. These are like phone calls. And so, take them with sort of that sort of green of salt, I suppose. But they're not usually that level of specificity or color as they say on the street. Correct. Cory, I want to ask you about another headline this week. So the S&P 500 came out and said that they will not fast track the process for SpaceX to join the index. Was this a surprise to you? A little bit. Only because it seemed like that had been this freight train of all these index providers saying we're going to allow, you know, sort of these newly listed huge companies into our index. You know, you've seen most notably, the first one to do it was the NASDAQ 100, which is obviously part of NASDAQ, which is the exchange that SpaceX is listing their shares on. And so there's like a clear, no one sort of has like, no one's written the story or proven that like NASDAQ changed their rules to allow SpaceX in to get the listing. So let's just be clear about this. They have their own rationale, but people rate their eyebrow at that decision and lo and behold, they did change the rules and SpaceX is listing on NASDAQ. The S&P has always been seen as a little bit more, I don't know, by the book conservative, they have a little bit less of a commercial incentive to change the rules. And from what I've been noticing online and also from talking to folks on Wall Street, like there's been a bit more pushback on this idea of sort of getting rid of sort of various rules around season and periods after an IPO to, you know, sort of and also not requiring them to hit any kind of level of profitability. And so there's been a bit of a pushback. And so we don't have visibility into a lot, you know, exactly why S&P sort of did this, but what about the implications of this though? So I mean, the reason that happened is unclear, but now looking ahead, it's not saying, yes, it's not saying SpaceX will never make it in, but in terms of investor uptake, you know, even just marketing just at the outset, what are the implications for it? It means they won't, you know, in the months ahead after the IPO, when there is going to be a significant unlock of insider shares available for sale, when employees, former employees, investors are finally going to be able to sell, there won't be the sort of wall of automatic or passive demand to scoop up those shares at whatever price. There's going to have to be more active managers or active retail investors making the decision themselves to buy the shares and saying this is a good deal. It's not just going to be an automatic decision that sort of an ETF for, and it, you know, does because the index that owns it. And so that's pretty big. And S&P is obviously the big corner there. - Right, and they'd have to look at those projections, $350 billion in burn through 2030. They'd have to make their own decision saying, "You know what, I'm willing to invest in that." It is certainly an interesting dynamic to watch. Cory, I want to thank you for coming on. That is Cory Weinberg, our Deputy Bergeef of Finance here at the information. For more coverage on the SpaceX IPO, I want to go deeper on XAI. The information's Elon Musk reporter Grace Kaye has a story out this morning unpacking the complicated dynamics between XAI and Anthropic. She has some great inside details on their relationship. And to put those details in broader context, I want to bring on co-executive editor Martin Pierce for this week's edition of the editor's cut, Martin, welcome back to the show. - Hey, I caught you. - I'm doing well. So we know that Anthropic is renting compute from XAI. What else did we find out from Grace's reporting about the relationship between these two companies? - Well, her story kind of revealed how over the last year or so, XAI has played this kind of cat and mouse game where they were tapping into Anthropics AI to help develop their own models and how Anthropic tried various things to cut off people who were doing that. And XAI tried to various ways to get around that, including by having their staff access to Anthropic through personal accounts. So I mean, I think what the story is a really good reminder of is that XAI's AI development is a mess. I mean, we all know that Elon has fired most people who are originally hired at XAI. He has a tendency to set unrealistic deadlines and when they not met, he just fired his people. And I, you know, why this is hugely important is that SpaceX, in what they are telling investors, they are projecting that AI will make up, I think in Corey's story last, not two thirds of the projected revenue that they will earn by 2030. And yet their AI business is a mess. And, you know, certainly right now, the only revenue they really have in the AI unit is actually coming from X, which is a non-AI service. It's mostly advertising. So if you're investing in SpaceX for its AI angle, think again. - So let's go back to the great story for a second. So this idea that Anthropic is cutting off access to XAI from using its tools. I mean, they are rivals, they're two competing AI labs. Is that common? I mean, I'm thinking about like cloud service providers. - Not only XAI, they also cut off open AI. I mean, this is, I think it's fairly common for AI labs to rely on other people's AI to help within the development of models. It just seems that Anthropic tried to cut off people doing it. And Elon Musk has admitted under oath that XAI has done this. I think he was, he talked about how they used open AI's tech, but our, I mean, grazes reporting details how they were also using Anthropic. And all that was, I think before XAI decided that they had all this spare computing capacity, they may have surrendered out to Anthropic. So that is the irony of that rental deal. - Right. But I guess, I mean, if you just think about all the other waves of technology innovation, like have we seen a version of this story before where rivals use each other's technology without them knowing it, cut off access, you know, and I'm maybe thinking about the corporate data wars too, you know, there's that rivalry playing out. Is this a common thing that happens with rivals cutting off access to each other's tools? - Oh, I mean, the data wars, as you mentioned, is the best recent example of that, where software firms tried to stop the customers from getting access to data stored in their apps, because the customers wanted to use it in other apps that competed with them. But I think there's a long history of tech firms, I mean, tech firms using other companies tech to develop their own. I think the most famous example might be Apple using Xeroxes, technology to develop, it's a original user interface. So it's not, I mean, I think people build on each other's technology. The real point here is that XAI isn't really in a position to be competitive in developing its own models, largely because Elon doesn't have the patience to allow researchers to do the work. He sets these deadlines, they don't meet the deadlines, and he vies them. And then this came up during the OpenAI Elon trial, where I think Greg Brockman made the point that when Elon was at OpenAI, he didn't have the patience and he didn't have the temperament for the long times that it takes to develop AI technology. And this is hugely important for people who are looking at the SpaceX IPO, because they are pitching themselves as an AI firm. And you know, if you think Elon Musk is gonna develop new AI, then maybe you should think again. - Now, the other interesting detail in Grace's story was the cursor integration here with XAI. It sounds like they are increasingly looking to cursor for maybe some of the answers here. And there was even the detail that cursor staff are spending more time at the office, even. - Right. And that's something I think we've actually reported before, that Elon brought in the cursor people to talk to XAI engineers and to that they fired all these XAI engineers. So I think it's just more of the same as what I'm talking about is that he, he, you know, his frustration with the rate of development manifested itself by him just firing people, which, you know, maybe that works for Tesla. It certainly did not work at Twitter, where he basically destroyed that business and he's hidden that by merging it into XAI and now merging XAI into SpaceX. But, you know, that is it. - Martin, I wanna ask you about another topic that was in the news this week. So we had a bunch of cybersecurity companies reporting earnings. you wrote about it in your briefing we had Netscope, we of course had the CEO on the show, we had CrowdStrike, I believe Palo Alto, networks may have also reported. ZScaler was last week, I believe. What was your big takeaway here from seeing all these earnings play out? Well, I know you're aware of this, Akash, because you talked to the Netscope CEO about it yesterday. So, you know, the takeaway is that these type of security executives love the fact that we're moving into AI, because AI is going to allow hackers to much more easily infiltrate corporate data systems and threaten everyone's data and maybe steal the money out of their bank account. No one likes me talking about that because they don't really want to believe that could happen, but believe me, it will. And so, the security people think this is a great opportunity and that's what they've all been talking about. Point I was making in a column earlier in the week was, well, if that's happening, why is there revenue growth not speeding up? In fact, for many companies, it's actually slowing down. The answer might be that it's still to come. They're all meeting with companies now to talk about myth, us and AI. So, maybe we'll see more growth in the next year. Maybe what will happen is that businesses will cut back on spending on some kinds of cyber tech and replace it with AI defense. We just don't know, but investors certainly are very hopeful. They've been really, they've lifted the stocks of CrowdStrike and Palo Alto by 50%. So far, this year and yet, so far, the impact on revenue was not that great. So, yeah. Well, and the point that the Netsuko CEO made to us is he said, well, we're just training our teams. It's coming. We're training. The point that I made back to him was, but you're still dissellering. Listen to the interview and he definitely avoided answering the question, which is what old CEOs do when they are confronted by reality. Right. And so, we'll have to see how the forecast come, but it's certainly something to watch. Martin, I want to thank you for coming on. That is Martin Pierce, our co-executive editor here at the information. The information has exclusive reporting that data center developer switch is in talks to raise billions of dollars at evaluation of at least 50 billion dollars. My colleagues, Anisa Gardese and Valida Poe reported that story. I want to bring on our San Francisco bureau chief Jason Dean to break it all down. Jason, welcome back to the show. It's great to have you back. What is switch the data center company? They are just that. They're a data center company that get around for quite a while. And obviously, these days, all about AI. Are they like, are they one of the biggest, like, are they helping the big hyperscalers develop data centers or how big are they? They're sizeable, but no, they're not, you know, in the hyperscaler realm, although they have five locations. They're big in Nevada where they were founded. They're also in Michigan, they're in Atlanta, they're in Texas. And of course, they are trying to grow, which is why they want to raise more money. So, do we know how much money, exactly, they're looking to raise? Looking at billions of dollars, we don't know exactly how much, and, you know, as Anisa and Valida reported, the details of the plans are not finalized. They're still in flux, so things could change, but billions of dollars at evaluation above 50 billion. But there are some big names who they reported are considering investing, right? That's right. They're talking to a range of investors, but including PE firms with big names among them, KKR, and Brookfield. So, you know, those are obviously serious players in the investment worlds. And yeah, it's not entirely clear what the specific plans are, but presumably that money will be funneled into building more data centers and expanding the ones they have. Right. So, look, I think it's easy for these data center companies to sort of think of them as just another data center, but I mean, if we just open the aperture a little bit, I want to ask you about this risk of overcapacity, which is something that all these data center companies face, Hey, Mithin Asia, the CEO of General Catalyst was on the show last year, and he said, there might be some pain, you know, some companies will suffer. When you think of the data center companies, is it the case that they lock in these long-term contracts? And if you lock them in now, you're sitting pretty for the next three years, or is it the type of business model where you start the project, and then it could be interrupted in a year if the customer says, hey, we don't actually want to build it down the line. I mean, it's a great question. I think, you know, we don't know the details of their financial agreements with their customers, but I would venture to say that if we switch into starting up unintended a period of overcapacity in AI data center worlds, I don't think anybody is going to be fully insulated from that. You know, we're just the whole thing we're seeing right now, and this is a part of it, is an incredible go-go-go mentality that stretches from the biggest companies in the world to smaller niche players in the AI data center worlds that are able to build capacity and sell it, including to the Iberscales, because there's just so much demand right now. If that changes, yeah, I think people re-examine contracts and companies that are sort of not front-center in the sense of the Iberscales, but a little off further toward the margins will be at some risk. Now, Switch was once a public company itself, right? That's right. Yeah, they've been around for a long time. This is a very established company. They were founded in 2000. They went public, I think in 2017, and then they were taking private in 2022. They were taking private at a valuation of $11 billion, including debt. So they're looking at quite an increase in the valuation. They've also raised a lot of money from debt issuance in the past several years. The last summer, they said they'd raised $20 billion just since 2024, and they've announced a bunch of additional debt issuance since then. So this equity deal that they're talking about is far from the only way that they've been raising cash to invest. And maybe they might go public in the future as well if they need to. That's right. Yeah, I need some believe it reported that this could set them up. Obviously, there are lots of caveats and ifs, but for an IPO as soon as next year. Great. Well, Jason, I want to thank you for coming on. That is Jason Dean, our San Francisco Bureau chief here at the information. That does it for today's show. A reminder, we're on this stream Monday through Friday at 10 a.m. Pacific 1 p.m. Eastern. If you can't make it then episodes are available on the information.com on our YouTube channel or wherever you get your podcasts. Make sure to follow us on social media on X Instagram, TikTok and LinkedIn. I'm looking forward to our next show on Monday. Have a great rest of your Friday and have a great weekend. Bye bye for now. [BLANK_AUDIO] [BLANK_AUDIO]