SpaceX's $5B Loss, OpenAI Stargate Shakeup, and Is OpenAI “Too Big to Fail?”

0:13Welcome, everyone, to the Informations TI TV. My name is Akash Basritcha. It is Friday, April 10th. Today on the show, we have got some exclusive reporting on the latest shakeup within OpenAI's Stargate Data Center build. We've also got a bull versus bear debate on OpenAI. It was a busy week of news for the company, so we will make that the subject of today's editor's cut. We've also got some more exclusive reporting around SpaceX's financials and the status of its IPO pursuits. And we'll wrap with this weekend's big read. We've got a great profile of Boris Cherney, the head of Claude Code. We'll talk about his role in the product's dominance and what questions lie ahead for him.

0:54It's going to be a great show, so let's get right on into it. OpenAI's Stargate initiative is facing more tribulations, according to exclusive reporting from the Information's Cloud and Compute reporter, Anissa Gardizi. I want to bring her on to share with us what she knows. Anissa, welcome back to the show. It's great to have you here. Thanks, Akash. What do we know? So three senior OpenAI Stargate executives have left the company, Peter Hushly, Shemez Hamani and Anuj Saharan, who were all key early members on OpenAI's first data center team. So we learned yesterday that, you know, they've either left the company or have told colleagues in Slack that they're planning to leave in the coming days.

1:37Now, this is a little bit surprising because they were early members of OpenAI's first infrastructure effort. The company is about to go public and all of the deals that these guys helped sign are now sort of underway and they're trying to execute them now. So we've also heard that these three executives are going to go to the same company. So we're working on figuring out what that company is. But this was definitely a big shakeup yesterday within OpenAI. Okay, I'm so curious what the company is. So we'll have to bring you back on when you find that out. But you've been doing some reporting on the shakeups happening at Stargate more broadly.

2:11How does this fit into that? Exactly. So last year, January 2025 was when OpenAI was at the White House announcing its$500 billion Stargate initiative. And as we've reported, that morphed from OpenAI potentially building and owning its own data centers to signing up a bunch of cloud deals with other partners. They struggled to finance and understand how exactly they were going to set up the structure, the Stargate structure with Oracle and SoftBank. And so last year, the Stargate venture really evolved. And then OpenAI late last year brought in a new compute leader, Sachin Khadi, who now reports to Greg Brockman at OpenAI.

2:52And a lot of the compute executives went from reporting to Greg to reporting to Sachin. So I think these departures that we're seeing are maybe a natural progression of, you know, leadership changes on the compute team that really started when they hired a new head of compute late last year. So Sachin is still at the company, though. He's the guy who's kind of leading the charge right now under Brockman. Exactly. I would think of Sachin as like the new compute leader. And so he joined last year. It was sort of understanding what's the team like? You know, who do we have? He did a reorg earlier this year, actually.

3:26And now we're seeing some of the early team members depart. So, look, we've talked about the talent wars with respect to AI researchers. and we've talked a little bit on the show about to what extent there is a talent war for data center talent. And it's hard to say right now if these people are being poached. We don't know where they're going. It's hard to say if they're leaving, if they're disgruntled. I mean, there's a lot of questions still, but I mean, how scarce is great data center talent right now? You talk to these people all the time. It is so scarce. Data center talent is not easy to come by.

4:01And these three executives were sort of inside OpenAI during this key period of them figuring out how they were actually going to secure all of the compute capacity that Sam Altman wants. And so I think they're sort of viewed now externally as really good hires. I was talking to people last night who, you know, wanted to get in touch with these folks to see if they were still looking for a job. So I think the market for data center talent is still really hot. And just having people with the context of OpenAI's compute needs is probably going to be really valuable to other companies that are looking to hire them or maybe other executives that are still at the company.

4:41Well, because – and I'm looking at Sachin's LinkedIn right here. I mean, you know, he was the chief technology and AI officer at Intel before. And so this was obviously a big move for him to come to OpenAI. And so I imagine this profile is pretty common for the top data center talent. So with all these departures, I want to ask you about a report that Bloomberg put out. Bloomberg reported that OpenAI has been speaking with investors and reassuring them that, hey, we're doing okay. And one of our biggest advantages is our compute. The compute obviously depends on whether or not they can build the data centers.

5:18What did you make of that report? I thought it made a lot of sense that OpenAI sent that note to investors because, you know, everyone is talking about how Anthropic is, you know, they just announced their$30 billion revenue run rate. Now everyone's wondering, do they have enough compute? And OpenAI has taken the stance of being so aggressive when it comes to publicly talking about how much compute they have, whereas Anthropic is a lot more quiet about it. And so it made a lot of sense that OpenAI went to investors and said, hey, our advantage is actually how much money we're spending or how many deals we're assigning.

5:53And so they gave some number comparisons of how OpenAI believes it's ramping its compute capacity faster than anthropic. But these numbers were obviously estimated by OpenAI. And so, you know, it's a little unclear how accurate their estimate of their rival's compute capacity will be. But I think it speaks to the moment we're in right now. Compute is the biggest constraint and startups are going to say, hey, our willingness to spend money on compute is our advantage. I mean, why wouldn't you say that if you're OpenAI? Right. And I mean, this is the whole seesaw that Dario talks about is, on one hand, you know, if you build too much, then you overspend.

6:32If you underbuild, then you don't have the capacity. We saw today that Anthropic signed a deal with CoreWeave. And so it's obviously thinking about its compute capacity as well. I want to ask you about Anthropic, though. So Reuters had a report that maybe they're looking to develop their own chips, which is another interesting development here in the store. We know OpenAI has been looking at perhaps making their own chips. I think they have a deal with Broadcom, if I'm not mistaken. Exactly, exactly. So what did you make of the Reuters story about Anthropix chips? The Reuters story on Anthropix, potentially thinking about designing its own chip, was not too surprising because every company under the sun that relies on NVIDIA chips or someone else's hardware has thought at one point, should we make our own chip?

7:23I think one good data point to look at is that we've been reporting on OpenAI building its own chip for many years. And today that chip is not in OpenAI's data centers and OpenAI's signing deals with all other kinds of chip companies. So if Anthropic is just starting to think about it right now, we have at least two to three years before it becomes important. But I think, you know, every company that relies on NVIDIA or even in Anthropics case that relies on Tranium from Amazon and also TPUs from Google, it would be silly if they weren't thinking about building their own chip. But this is something that, you know, won't really make a difference for years to come.

8:02Let me ask you a bit of a broader question to close up here. your conversations with folks in the data center space. I mean, we are waiting to see how many of these data centers actually get built, how many of them are delayed. We were talking to the CTO of CoreWeave a little bit earlier in the week, and we were asking him about delays. He seems pretty confident with their builds. But what's the vibe generally right now in data center land compared to when you were in Hawaii at that big data center conference, has anything changed really? I would say that the vibe on the ground within data center circles is that projects are delayed and the costs are increasing to the point where some of these projects are just going to be really hard to build in the near term or people will have to push them out into the long term simply because it would be way too expensive to build right now.

8:58Now, everyone is saying this about their competitors' projects, so that is always something to keep in mind. But I think cost overruns and delays, labor shortages, that's really on everyone's minds. So this is something that we will be following this year for sure. Great. Well, Anissa, I want to thank you for coming on. That is Anissa Gardizi, our cloud and compute reporter here at The Information. Speaking of OpenAI, this week on The Editor's Cut, we are going to be hosting a bull versus bear debate on the company. OpenAI has made a ton of news this week, and it seems we could be at an inflection point.

9:32The question, of course, being whether or not it is a good or not so good inflection. I want to bring on our co-executive editor, Martin Pierce, and our San Francisco Bureau chief, Jason Dean, for help us with the conversation. Welcome to you both. It's great to have you here. Hey, gosh. Great to be here. Thanks. So, Jason, it's your first time on the show. Welcome to TI-TV. We're excited to have you here. You are taking the bear perspective, is what we decided, for UnOpenAI. So let's walk through the events of the week. How are you thinking about it? What does it tell us about OpenAI's trajectory?

10:08Yeah, I mean, it certainly wasn't the best week for OpenAI. It started with our scoop by our colleagues Anissa and Amir that CFO Sarah Fryer and CEO Sam Altman are sort of at odds over the IPO timing and some of the spending plans. Monday, you had a huge New Yorker piece that revisited the whole narrative that Sam Altman can't be trusted. And then Anthropic has been racing ahead. And we learned this week that their annualized revenues hit$30 billion, which means that for the first time, they're now actually bigger than OpenAI. Plus, we've been hearing about this new model from Anthropic that is so powerful, supposedly.

10:56that it risks being able to hack every piece of software on the planet, and therefore they're not fully releasing it. So there's been a lot of tough news for OpenAI and good news for its rival. And then, of course, to cap it off, the story that Anissa was just talking about, that several of the top executives at OpenAI in charge of their compute have left. So tough times, of course, the news ebbs and flows, and they also have had some good news recently about fundraising. But at a minimum, the competition's gotten a lot closer with Anthropic and we're seeing, again, signs of fissures in the leadership at OpenAI.

11:38Okay. Martin, he's making some good points. You are the bull, which is something I got to see. So let's see. The only inflection point here is in the media's coverage of the company. OpenAI was the darling of the media, and now it's not. That always happens. Really, ChatGPT is still the standard for the vast majority of people who use AI. that is not likely to change very easily because people just, you know, get these, you know, that they get very set in their ways. So I really, firstly, on that point. Secondly, on the ARR point, the two companies, Anthropic and OpenAI, they do not calculate revenue in the same way.

12:30And if they did, I think you would find that OpenAI was probably bigger than it appears. And in terms of the management issues, look, I agree that there is some management concerns here, but that could be easily resolved. They just have to replace the management, which they can do and probably will do if they have to. This company, I think, is still very well positioned. And, you know, we should never in an industry like this where it will take years for AI to really develop. we cannot over-index on one week of bad news. Okay. So, Jason, what do you think? I mean, is Sam Altman the right person to be running the company?

13:14You know, is it just media coverage? Well, I would differ with Martin on that. I mean, I think there's some substance here. I'm not going to opine on whether Sam is the right person to lead it or not, but the whole idea that there's a discussion at this point in a year, in April of a year where Altman has said he wants to go public, that whether he is even the guy to lead this company or whether they might need to change CFOs or other top executives suggests that things are not in the situation that you would want if you're talking about a company that's supposed to have a trillion dollar IPO.

13:53So I think the leadership issues are significant. they've been significant and they've weathered them in the past but you know as an investor as a potential shareholder you do not want to see that in a company with a valuation like that that is you know still burning enormous amounts of cash every single day of the week um can we just pull back the camera a bit I mean this obsession with when they go public I mean I know the media loves to obsess when this company will go public. Going public is a minor development in the evolution of any company. Over the long term, whether they go public this year or next year, it's not going to actually matter at all.

14:38So I think that is one of the problems. Yes, it's not ideal for them to be, for this reported debate, among these top people about whether this is the right year. But in the long term, this is not going to matter at all. I would say if you were an investor, this is a great buying opportunity because everyone is now down on open AI. And I think you could argue that this is the time to be buying. Well, I mean, you know, they're at parity in size, even if... We don't know that, Jason. Even if your accounting point is right, Martin, And Anthropic is a lot closer to the same size to OpenAI than they were a while ago.

15:24And their valuation is less than half. So I think if you're looking around for AI buying opportunities, maybe you're looking at Anthropic and not OpenAI. I agree with you that the month of the IPO timing or maybe even the quarter is not that important. But let's remember that I'm not the one who came up with that idea. It's Sam Altman himself, who's the leader. He obviously thinks it's very important. and he thinks it's important not to let Anthropics steal the march with investors. And they still need capital. They just raised$122 billion, but they need more. And the private markets are reaching the limits of their ability to supply that capital.

15:58So I think the idea of IPOs for these companies become much more important than they are for typical companies that are thinking about going public. They need access to additional pools of capital if they're going to continue to burn tens of billions of dollars a year. Life is long, though. Let's remember that Anthropik has some issues with capacity. We've reported that users have been having increasing problems using their products. So I think the fact they don't have enough capacity will become more of an issue. And we could easily see that the story about these two reverses fairly soon. That is a fair point.

16:48I would just say that the reason that Anthropics' capacity crunch is so severe right now is because their product has been so successful and it's a product that monetizes. You were talking earlier about ChatGPT being a far better known product, and that is certainly true. But right now, we have OpenAI projecting enormous amounts of advertising revenue that can leverage that. But their main move recently, strategically, has been to try to follow Anthropik's lead, and it's a big lead, in terms of the enterprise business, where Cloud Code has been so successful, and Cloud Cowork is also ahead of what OpenAI has been doing in that area.

17:30Can I just remind you guys of some, if you go back in time, AT &T had the iPhone before anybody else. The usage was so enormous, the AT &T network collapsed. And that when the iPhone was made available to the other carriers, people moved. It's not always good to be that successful. What do you mean by that? What do you mean by that? What I'm saying is that if you have that much usage and the system cannot handle it, then it might end up rebounding on you, and that could be a bad thing, which I think is beginning to happen. So I'm just arguing that the bull case on OpenAI is underappreciated right now, and we'll see what happens next week.

18:26Can I ask you, look, I want to ask you about the question, both of you here, the idea that so many investors, hyperscalers, venture capitalists, I mean, there are so many players that have a stake in OpenAI, whether it's a literal stake or it's just a vested interest in having AI companies thrive. I mean, you know, Jason... I catch that is a very good point. Are all these companies going to let this company die? No. So you're saying it's too big to fail? I think I've heard that somewhere before. I've asked you. Too big to fail. Therefore, it's a bull case. Jason? I feel like maybe history has not always served that argument terribly well.

19:13I think we've had some pretty epic companies that have struggled. But I'm not predicting the collapse of open AI, to be clear. I think, you know, there's plenty of room for multiple competitors in this race in the future. I do think that the idea that we have even, you know, some level of parity in revenue right now, that is not a short-term blip. You know, the model capability horse race has changed leads more times than I can count over the last several years. But consistently, OpenAI has been ahead of Anthropik in size and ahead by quite a bit since this whole race began. Now they're not. Anthropik has closed that distance and they've closed that distance with continued momentum.

19:57Now, you know, it's a long race. As Martin has said, I agree with that. This is, we are nowhere near the end of it. But that is a pretty fundamental change in the dynamic of the race. ARR is not a very good metric. So I don't think we should be resting the argument on that. Let's just say that. I think that what is underappreciated about OpenAI is that if they brought in some seriously good management, the company would take off. So that's the biggest reason to buy the stock. It can only go up from here. Okay, so Martin, who should they bring in? That's a good question. I don't know. That is a very good question, though.

20:41Jason? Well, you're fine with Sam. Yeah, I mean, I just would say that when your argument for a trillion-dollar-valued company is like, well, if they had good management, they would be great, it's maybe not the greatest argument. I was told to argue this guy.

21:03Okay. Well, all right. Thanks for that, guys. And thank you, Martin, for breaking the fourth wall. We appreciate you both being good sports about this. Next up, we're going to have to have the SpaceX debate, and I'm very excited for that one. Stay tuned. Sorry. We'll bring you back on for SpaceX later on. Okay. All right. That is Martin Pierce, our co-executive editor, and Jason Dean, our San Francisco bureau chief, here at The Information. Thank you. Speaking of SpaceX, SpaceX's revenue might be more than we thought, but it is still burning a ton of cash. That is according to exclusive reporting from our deputy bureau chief of finance, Corey Weinberg.

21:42He got SpaceX's latest financials. I want to bring him on, and I also want to bring on Theo Waite, our Elon Musk reporter, for his latest reporting on the SpaceX IPO. Welcome to you both. It's great to have you here. Hey, Kyle. Corey, what do we know? What's the latest on SpaceX's top and bottom line? This is the first time that anyone has reported the consolidated financials of SpaceX and XAI. And it is insanely interesting. What we have here is a company that is spending more on capital expenditures than it brought in in revenue last year. So it brought in more than$18.5 billion in revenue. its capital expenditures neared 21 billion.

22:27And what we see very plainly is with SpaceX buying XAI, it absorbed all of that spending and went from a company that looked pretty clean on a profitability front because of its Starlink connectivity segment. And now it's burning cash, it's losing money. It's net loss using gap accounting principles was nearly$5 billion last year. So you talked about the XAI factor in here. What do we know about the revenue breakdown in terms of how much is XAI and what the different business lines within the space business are? Do we have any clarity there? We know how it breaks down in terms of adjusted EBITDA.

23:14So, you know, sort of that very, you know, somewhat crude metric of profitability that investors do pay attention to that excludes major costs like depreciation and stock-based compensation. But the Starlink business is the huge driver of that profitability. It generated more than$7 billion in adjusted EBITDA last year. And even the launch business that SpaceX has was slightly profitable, nearly nearly a billion in adjusted EBITDA. The AI division, not as much. It lost between a billion and a billion five on that measure. And so very quickly, before I get to you, Theo, Corey, general reaction to this?

24:02I mean, was this better or worse than you thought? it shows you know sort of i think in in sort of pretty living color what everyone had a somewhat assumed after a space exit bought xai which was uh you know this this is absorbing a part of elon musk's empire that is just unproven um whereas spacex is has a pretty good business on its hand. And I think, you know, where my mind really goes from here is they're going to have to show some pretty rosy forward looking projections to justify the, you know, 1.5 to$2 trillion valuation they may seek. Okay, so Theo, what's the latest on how they are preparing for the IPO then?

24:57Theo Wayt:I mean, I, you know, I think the numbers Corey reported show the necessity of cutting costs at XAI and reducing the amount of money they're burning there. And I think they're in the process of doing that. I mean, they've put this guy named Michael Nichols, who is the head of Starlink currently. He's now also a big executive at XAI and is kind of moving in, bringing a bunch of SpaceX people in to clean things up and try to, you know, reduce the burn there and make as much staunch the bleeding, essentially, basically. And I think the other thing that's important to remember when we think about XAI's financials is that the majority of the revenue XAI is making is almost certainly coming from X and not actually from selling its AI models.

25:50Theo Wayt:So the other weird thing about this IPO is that we have this Elon Twitter takeover tucked into xai and and now tucked into spacex and making that all into a coherent package is like you know pretty pretty challenging i would think cory do you have any sense for how investors are reacting to this pitch you're gonna have investors that say um we believe in elon no matter what You know, that's going to be your cohort of retail investors. It's going to be certainly the existing investors that have made a ton of money off of his companies on Wall Street and beyond. You're going to have a group of investors that say, okay, I see the vision.

26:42You have AI plus space, maybe data centers in space, maybe Starlink direct-to-sell business. like, okay, I see it. Great company. I want in. I'll pay a premium. Maybe this thing gets bought by Tesla eventually and I'll make that bet. And you're going to have investors that are just like, yeah, this is way too rich for me. I'll pass. I have other things I can invest in. I think how big that middle cohort is going to be, you know, really crucial, not just for the IPO where they're trying to raise, you know, 50, 75, 100 billion, largest IPO of all time. But how do investors receive this as existing SpaceX shareholders need to sell hundreds of billions of dollars worth of shares into the markets?

27:32What is, you know, sort of as we can see, SpaceX not making money right now when you include the XAI division. And so it will really need this capital. It needs this IPO to go well. So, Theo, take us a little bit inside this roadshow that's happening right now. I mean, you had some great color in your story about how the company is sort of wooing investors. Just give us a little bit of the on the ground, you know, anecdotes.

28:01Theo Wayt:I will say a lot of this is Corey's reporting. Okay, fine. Corey, go.