원문 보기
We are seeing a huge inflection point in SpaceX's earnings estimates. This is shocking. The company, which was expected to be loss-making for years before and during the IPO, is now expected to become profitable much more quickly, according to new revenue projections. And you could almost say that it has become cheap.
This is simply an incredible turn of events. I will show you the actual statistics as well as the market rate situation and peers for these data centers. We need to sort this out because there is too much information. I want to clearly outline my own interest. Our venture fund has a stake in SpaceX, which we have held since the time when the capitalization was around 350 billion.
Do you know how much it costs now? 1.5 trillion or so , maybe even more. But the thing is , not a single stock ... Oh, now it's almost $2 trillion . No shares have been sold yet. We have n't even received... The shares haven't even fully arrived yet. The blocking process is still in the unlocking phase.
So our initial thesis about SpaceX has always been focused on space exploration, which we believe will become relevant in 2035–2040. Uh, and of course, Starlink, Starlink growth. This was our initial forecast for the company's valuation. And we watched analysts' expectations. Actually, here are some of the things we wrote.
We're like, "Hey, you know, by the end of 2030 we could be generating revenue from launch services." We could get some revenue from Starlink . Hey, we can grow to an operating profit of somewhere around $20 billion. This justifies the price of about $150. But, but, but—this is old news. And I'm not even talking about the launch table that we talked about - how exactly we scale launches with our Starbase in Louisiana with 10 sites, Cape Canaveral and Kennedy with three sites, Starbase in Texas with two, and how that could generate $80 billion to $160 billion in revenue.
This is a kind of estimate of how much these launch services will cost in the 2030s. Here's how you can hit those price targets of a few thousand dollars just through launch services. But let's put launch services aside for a moment , because Wall Street is currently adjusting SpaceX's earnings forecasts solely based on the development of AI data centers.
And news broke that made us dig deeper into the topic, in addition to previous research on Anthropic's half- trillion dollar plans for computing power. More on that in a moment. Well, let's start with what actually caused all this . The reason for all this is the following: SpaceX is reviewing plans to build data centers, which could slow down expansion.
So, when you first read this, the thought comes to mind: "Oh, that sounds like a bearish trend." This means they won't be able to increase computing power as quickly. After all, they still need to raise between four hundred and five hundred billion dollars to build the 10 gigawatts of computing power they need, which means they'll have to borrow.
And if they are going to raise debt, there is a chance that analysts are trying to slightly inflate forecasts for the company to make debt more attractive, not just to get commissions from big banks. This isn't how it's supposed to work, but let's be realistic, that's how it works. Uh, and of course, any interest rate reduction SpaceX could get from a higher investment grade rating saves them a ton of money in interest.
So when you first see this headline, you think, "Oh, this is bad." Slowing down the expansion. After all, there is now literally talk that Anthropic wants to spend up to $517 billion in commitments between the end of 2025 and the end of 2035. Well, Anthropic is a bit picky when it comes to computing resources.
They literally rent from everyone. They rent from AWS, they rent from Google, they rent from Microsoft, SpaceX , Economy, they rent from AMD and Scale Lambda. They rent from anyone they can get their hands on because they just want it all. Their deal with SpaceX actually, quote, significantly increased their computing power, according to a Bloomberg report for this, er, or that particular note.
But what is striking is that Anthropic is driving half a trillion dollars in computing power spending over the next decade. This is just one company. If you add OpenAI, you probably add another $750 billion that they want to spend by 2032, it seems. When you add that together, you start to approach two trillion dollars in computing power spending from OpenAI and Anthropic alone over the next few years.
These are these are crazy numbers. They are simply incredibly huge . But anyway, last December Anthropic told investors that it wanted to spend—I'll write that number down because it's really important to see what's changed. And this will show you why analysts' expectations for SpaceX are changing. Last December, Anthropic told analysts that it was looking to raise about $180 billion for computing purposes.
This has now been updated to an estimate of $517 billion in computing power. This is in addition to 750 or possibly more from OpenAI. This is quite impressive. As of last October, Anthropic had between 1 and 2 gigawatts of computing power. So, let's write this down here. Okay, last October they had about 1-2 gigawatts of computing power.
They are trying to achieve capacity based on what they have already reserved. So, this is already fixed, this is deferred capacity, because it still needs to be built, of 14.8 gigawatts. This is madness. Anthropic specifically approached Microsoft about a $30 billion deal for a gigawatt of Azure and Nvidia services.
Of course, they also have a recently announced partnership with Fluid Stack. If I haven't mentioned, Terrawolf is also involved in this compared to other companies that Anthropic has deals with. They make large-scale deals. By the way, Anthropic's computing power is 517 billion to 418. OpenAI plans to reach 30 by 2035.
So, that's almost twice as much as Anthropic says, and Anthropic's numbers are already crazy. So, OpenAI and good old Sam Altman are getting even more audacious. Some of this could explain the surge we are seeing in Dell shares, for example, due to enterprise AI, the development of cloud services or sovereign infrastructures.
Now everyone wants these data centers. And the amazing thing is that this is finally starting to be reflected in analysts' expectations for SpaceX, turning a business that seemed overvalued into a completely affordable one. This is madness. And again , we'll explain these graphs that we have here in the Meet Kevin app for course participants.
We don't have a promo code right now. I'm not trying to push a promo code on you or anything like that. I'll explain all of this in a moment. Before we explain what has changed, you need to know what the problem is . So, let's start with the problem. Last week, there was an outage due to data centers in Memphis .
SpaceX's responsibility. Grok fell, Claude fell. This was a problem. This is not the first time. Earlier this summer, around the time of the IPO, there was talk that SpaceX was leasing computing power after its own teams had problems using it. So, Ed Ludlow claims that SpaceX decided to lease the full capacity of its Colossus 1 center to Anthropic after encountering technical difficulties in developing and launching its Grok intelligent models.
The firm encountered latency issues when connecting Colossus to two other facilities located more than 10 miles apart . Training requires ultra-fast connections, and using legacy channels can create latency that slows down the entire cluster for training purposes. Anthropic may use this, we believe, for inferential purposes.
So maybe this delay is less important because they're expanding credits that can be used for Fable or something else. Of course, many people brag about the Colossus object because it was built in 122 days. It is important to remember that this facility is an older building that already existed. So they just had to come in, bring in the chips, and put up some temporary gas turbines outside.
I'll simplify this further. But that's pretty much how it all happened. Most of the time, when you hear about a data center being built, it means that there was just a vacant lot there . There is no parking, no asphalt, no walls, no existing air conditioning, no existing power grids, etc. All of this had to be actually allowed and brought to a halt.
It was like, "Hey , let's rent this and chip in." That's how they got to 122 days, but now they're facing some problems because of that speed. So Elon is doing something right now that I think is really smart. I respect that. According to this informational material, Elon meets regularly on Sundays.
He travels and meets with engineers at these facilities, somewhere it is written here about Sundays, to better control this development process. In recent months, he has visited the Memphis facility almost weekly, stopping by mostly on Sundays. In particular, he visited it very recently, last Sunday.
So, what's going on? Well, Elon decided to update the team. Apparently they're firing a bunch of old data center architects and bringing in SpaceX engineers, saying, " Look, we can't have a latency problem." We need backup systems. We need to do this right. We can't afford a bad reputation due to Claude's failures at our facilities, because then they will terminate our contract.
I don't think they'll actually break the contract, but they'll be angry and won't want to pay us that much money. If we want to earn premium revenue from computing, we will have to provide premium quality services. So, here are the current market deals we see on one sheet. And Anthropic's deal with SpaceX here comes out to about $50 billion per gigawatt per year, which is just crazy.
This is very, very expensive computing power that Anthropic needs. They pay for premium class. So they deserve premium service. They don't need disruptions. These are just some of the other rates we're seeing. Typically, deals are made for around $14–17 billion, but these are typically long-term contracts.
These are 6-year contracts. The $ 50 billion Anthropic and SpaceX deal —it's, you know , a 90-day cancellation policy on a 3- year contract. And that's not even as much computing power as Anthropic needs. Here are some more comparisons. You have Nebius with very…it's just different sizes of deals. We don't know how much they cost per year because, unfortunately, some of these deals don't disclose the terms of the contract or capacity.
So we just hear, "Oh, a 40 billion deal with Nebius" or "a 14 billion deal with Microsoft and Enscale" or "Microsoft for 9.7 billion with Iren," but you still have to calculate what that means in the end. In fact, the deal with Iren and Nebius comes out to about $10 billion per gigawatt. So it makes sense that Anthropic is paying a lot of money for SpaceX's computing power to significantly expand its cloud services.
But now you have a problem. If your first Colossus is a bit of a failure and you want to scale from 1.5 gigawatts to 10 gigawatts , you're going to start borrowing at scale. And the last thing you want when you're going to borrow a lot of money is to spend it inappropriately. You don't want to do this.
So I think Elon is doing what befits an entrepreneur. It's really cool to come out and say, "Hey, you know what? We're going to take control of this. We're going to move away from the rush and take it a little bit slower. But because we're going to borrow, you know, over $400 billion, this is a fundraising up front.
Because we're going to borrow over $400 billion , we're going to bring in a new team of SpaceX engineers to do it right and a little bit slower. We're going to add more redundancy from the beginning. We're going to rework parts of the existing data centers. We're going to have more redundancy. We're going to have a new team .
We're going to increase reliability . And we're going to make sure that we're not creating problems by interfering with each other during the construction process. " So, with that in mind, I decided to look at what the earnings expectations are right now and what analysts are saying about Tesla or SpaceX.
And I was just amazed. So if I go into the Meet Kevin app, in the stocks section for course participants, I see that the valuations revenue for 2026 was somewhere around $35 billion in total. That's revenue. Those 2026 revenue estimates are now up about 50%. For 2027, the consensus revenue estimate is up from about 65 to 102.
That's a 57% increase. Part of that happened, as you can see, right here in the second quarter report. What happened in the second quarter report? In the second quarter report, they received an upfront payment from Anthropic of $1.6 billion . Yes, there's a 90-day cancellation policy. But it's not going to be canceled.
There it is . $1.6 billion in AI infrastructure revenue . And then we wrote down some catalysts here, which we'll talk about in detail in a moment. But the crazy thing is, it's not just those estimates that have changed. It gets even crazier. Ready for it? Check it out. Check out the FY28 estimates, which are in blue, FY29 and FY30.
That's operating income . So I just went from revenue to operating income. I want to make that clear. We're talking about operating income projections right now. These numbers are amazing, and they actually make SpaceX cheap right now. What? This is the point where people will say, " But Kevin, you said you own SpaceX stock in a venture fund, which we do, you know, millions of dollars, and you don't want to buy it until the price drops to 80, 70 on average." This is true.
Based on that forecast, I say, "This would be a great place to make an incredible rate of return." "I didn't say I wanted it to go down to that level . I said that's where I think you can get a really great rate of return. It would have already made a great rate of return, but obviously the price never went down to that level.
We were, however, bearish after the IPO because you're taking a lot of money out of the markets. You're taking over a hundred billion dollars out of the markets. 80 something, I think it was like 85 plus another 25 billion in financing. And there's going to be another $400 billion in financing . But anyway, we're bearish after the IPO.
Okay, so the price went down from a high of 220. So what? It's 150 now. Well, whatever. What, what do we say? It's down 30 percent or something. 220, that's , wow, 31.9%. Cool. But that doesn't mean that we can't look at the data and say, 'Oh my God, if this is true, this is old news.'" This is what I was thinking on the Disney cruise: "I can't figure out these finances." Now I'm like, "Dude, either these numbers are lying to me and the analysts are lying to me, or the OpenAI and Anthropic spending is probably a bubble, maybe unsustainable.
We'll see when we get their first reports, right? Anthropic's S-1 should be out soon. And then, closer to next year, we'll get a lot more clarity. But look at this. This is crazy. This is operating income . Okay, just to be clear, my operating income estimate was about $22 billion for 2030. $22 billion for 2030.
for 2030. I'm just going to check that. And keep in mind, I didn't include the startup numbers in there . We talked about that before. Startup numbers separately. But if I go here, here. The operating income from AI infrastructure, my overall operating income estimate was about $21.5 billion . Got it?
The operating profit estimate is $21.5 billion . And that's before interest. They're going to have to pay a lot of interest on all that financing, obviously. You know, if you raise $400 billion in debt and you pay 7% interest, you're paying $28 billion in interest every year. So , to be fair, you probably have to subtract $28 billion in interest from all of those numbers, because interest comes before net income and after operating income.
So you're going to have to subtract interest. But even if you subtract interest, man, look, look. If I subtract interest from the 2028 estimate now, it's 40, that estimate is now twice my estimate right here. And if you look over here, you'll see the changes over time. My spreadsheet has been rewritten in this area.
You know, there was a 2028 estimate where they would have $15 in operating income. Okay, if I take that bottom point of 15 and go up to the current 68, SpaceX's potential revenue has grown literally 4.5 times. That's right. I mean, look at this. In 2029, they literally raised their operating profit potential from $20 to $120.
That's 6 times. And here they went from $40 to $200. That's 5 times. So we have 4, 6, and 5 times growth in operating profit. What the hell, man? That's so weird. Look at the consensus revenue forecast. Same here. This is the revenue graph. From $ 100 billion in 2028 to over $400 billion in 2030. And this, you know, growth...
it's not that big of a multiple here. Operating profit shows a bigger jump. It's a little less than double. It's a little less than double. A little less than double the total revenue growth. So in other words, they think the payback period is going to be very short. And that the computer lease is going to have very high margins.
You're essentially betting all-in on Anthropic and OpenAI. That's what it is . It's a bet on Anthropic and OpenAI. Okay. But look at this. These were estimates. Look at this. These are estimates of peak performance after reporting. I think it's operating... wait, wait. 197. I'd like to see the signature to that.
Okay, let me find what that signature is. But you can see on the right the percentage change since the IPO. Look at this. This is operating income . Okay, I thought so , but I wanted to make sure. My operating income for 2030 was somewhere in the $20 billion range, and I was very cautious in my projections.
Okay. Now it's up. I was cautious and slow in time. It's more than doubled. And here it's more than tripled. It's crazy. And some of these actual levels. Let's look at if I look at... let's look. Let's look. Let's look . Margin. Here are the margin projections right now. Analysts are projecting operating margin to be 39% to 47%.
What? That's just crazy. So I don't know. But honestly, if these analyst numbers are even remotely accurate, as strange as it sounds, it's really cheap right now. How can you justify that? You take their projections for 2027 because it squeezes the growth rate, because they're unlikely to be profitable this year.
4 cents per share this year. Next year, they're expecting $1.72. The percentage growth between them is not a comparable number, so we skip that. We move forward one year and then look at the growth after that. 1.66; 4.47; 8.51. That's an average of 88% EPS growth per year. So their current price-to- earnings ratio, based on next year's EPS forecast of $1.72, is about 88.
So a P/E ratio of 88, an 88% growth rate, what does that mean? SpaceX now essentially has a PEG ratio of one. So if you're brave enough to bet that Anthropic and OpenAI are going to keep spending like drunken sailors, you can get SpaceX at a PEG ratio of 1 based on current analyst expectations . So catalyst number one: don't lose the Anthropic contract.
I don't think you will, because it's a very high premium, but Anthropic needs it. Catalyst number two: become the most reliable provider of computing power. SpaceX can significantly outperform cloud services. Catalyst number three, it's a small catalyst, but stopping X monetization will actually increase their profitability.
For most content creators, X monetization has been essentially destroyed. And the $920 million Google deal that was signed in June is going to be at full capacity by October. So we're in the build-up phase now. This is crazy. So SpaceX is pretty amazing and very, very exciting . And I have to say something else. . Why not advertise the stuff .
Comments 0