Okay. I think here we go. We're going to get started with our next fireside chat. It's my pleasure to welcome SpaceXAI to the conference this year, Bret Johnsen, CFO. I am going to do a little bit of housekeeping, and then Bret and I are going to get into a conversation. With a company like this, we are going to try to cover as much ground as we possibly can, but I'm up against a shot clock. Some of the statements made today by SpaceX may be considered forward-looking.
These statements involve a number of risks and uncertainties that could cause actual results to differ materially. Any forward-looking statements made today by the company are based on assumptions as of today, and SpaceX undertakes no obligation to update them. Please refer to SpaceX's Form 10-Q for a discussion of risk factors that may impact actual results. Bret, thanks for agreeing to do the conference.
Thank you.
Great to have the opportunity to talk. Let's start with what we've even talked about in some of our work, which is the central theme I think of this company is vertical integration.
Yeah.
Talk a little bit about how that is the anchor for this company, both at the company level and even at the segment level.
Yeah. I think vertical integration is core to who we are. I totally agree with you. We really believe that we are building the infrastructure of the future, and we're taking a very different approach, and we have since inception. If you go back to the old days when we were a rocket company which wasn't too long ago, by the way, we have always taken a very vertically integrated approach.
Starting when we were really just focused on rockets in the beginning, or I should say rockets and spacecraft, a lot of it was because we couldn't find elements of the supply chain that were moving at the pace we needed or in an element of quality that we needed for sure. Quality is so important when you're flying payloads to space, especially when the payloads could be you and me.
Yeah.
Right? Certainly that was in our DNA. Then when we looked up and headed down the path of being a communications business, and started Starlink, it was the same type of thing over again. We want to own all the way from in the rockets, the metal coming in the door to building the engines, doing all the avionics and the software, and then the customer relationship. In comms, it became the same thing, where we were owning the launch, we were owning building the satellites themselves, and then going to the end customer or enterprise.
Taking that approach now to the AI side of our business, we are very vertically integrated, all the way to the point that we're building our own infrastructure. We're the GC building the building, putting up the power, and then using our own model to get directly to the end customer, whether it's consumer or enterprise. Then very soon with orbital compute as well.
Okay. Well understood. Let's come back to that launch business. I think when I think about where you're going over the next 5- 10 years, so much of it is tied to rapid reusability in the launch business tied to Starship. Talk to us a little bit about the building blocks to take Starship to where you want it to go, and what some of the key learnings have been from recent launches.
Well, I'm glad you start with Starship because Starship is so core to the rest of our business in really all the segments. It's been years and years in the making. We had a huge milestone event with flight 13. So much learning came from that. If you didn't get a chance to see it, we were able to not only deliver payloads, I would say demonstration payloads of V3 satellites, but also did a relight of the Raptor engine, but probably most importantly, had the softest splash down, kind of precision splash down over the ocean of the second stage.
Then kind of magically, we're able to tow what looked like a big whale back into shore, thanks to the efforts of pretty incredible recovery team on the SpaceX side, we're able to get that vehicle back. We were able to get our engineers climbing all over that vehicle to see exactly the environments that the heat shield went through.
Yeah.
There's been so much learning just in the last number of weeks, that we're able to take back into our flights coming forward. So very exciting related to how the heat shield performed already on flight 13, but then the learnings that we got from that will carry right into flight 14 and beyond. Then we've got another flight coming up later this month for flight 14, which for the CFO is very exciting because it's going to be a revenue-generating flight. I love the demonstration flights, and now I'm excited that we're moving into production cadence and flying production V3 Starlink satellites-
Yeah.
...to orbit on our next flight. I think that'll be a huge milestone for us. If you look beyond flight 14, starting to get a shot at bringing back both the first and second stage, will be big milestones coming up for later this year.
Okay. Let's pivot the conversation to AI infrastructure. We'll start with orbital compute. When you think about orbital compute and you try to conceptualize this for investors, how should investors think about the challenges that are being solved for to stand up and scale orbital compute and the timeline over which feasibility and scale can be built in orbital compute?
Yeah, it's interesting. I think I've heard conversations in two sides of it. One is it feels like almost every leader in the AI industry is willing to openly acknowledge that orbital compute is the future, right? It's clean energy. You don't have all the constraints. You don't affect any type of community from that perspective. I think it's clear that a large majority, if not almost everybody thinks that's the future. But most, if not everybody other than us, think it's many, many years away. The huge difference is we're actually, again, very vertically integrated, and we're in control of all the pieces that are required.
We're actually targeting to fly our first orbital compute satellites next year. I think that will be an enormous moment when they start, and when everybody starts to see that we're putting up huge amounts of compute into space going into 2028, because that's the right way to do this anyway, and then everyone will see that the timeline is far shorter than people had expected. Some of that circles right back to Starship.
What these are, is these are largely the same V3 bus that we were just talking about for Starlink that we're flying literally later this month, coming days and weeks. Then changing out the payload, giving it more solar, so certainly the solar arrays will be larger. I think that is incredibly important, and people will see that and then realize how close this is to becoming a reality. It doesn't mean we're not continuing to expand our terrestrial footprint on compute as well, but I do think orbital compute is definitely the future.
Okay, maybe just one last quick one on orbital before we turn to terrestrial. You touched upon it a little bit, but just help us better understand how to think about the cost advantages of orbital compute relative to terrestrial compute, because that could be a real-
Yeah.
...competitive differentiation factor for the company over the long term.
Yeah. I think that the moment when it becomes better than terrestrial is all about reusability.
Yeah.
Really, that's been a core capability of our company for over a decade now. We brought back first stage boosters for Falcon 9 in December of 2015 and started reflying them the year after that. For now a decade, we've been reflying, I think, over 500 rocket boosters now to space, reflown. It is knowledge that we're carrying into that Starship platform.
When you get to the point, hopefully very soon. Because we've already recovered a first stage multiple times and reflown a first stage of Starship. You just heard me talk about the recovery progress on the second stage. When you get into, hopefully as soon as next year, where you're reflying both first and second stage of Starship, the cost curve goes down dramatically. What's really great about it is you look terrestrial versus orbital. In terrestrial, everything's getting actually more expensive.
Yeah.
Just inflation plays out and power gets more expensive, cooling, everything else, buildings, construction, real estate. You're actually on the wrong curve for all of us in tech. In tech, where you used to, I was a semiconductor guy before a space guy. You're used to going down that cost curve.
Yeah.
The great part is when you're flying orbital compute, you're going down the cost curve. The rockets are going to get more and more efficient, more and more capable. The cost structure will go down related to the deployment cost, and the satellite will get more and more cost-effective. You'll benefit from Moore's law or whatever the equivalent there is and following is. You'll be able to get to scale and get the economies to scale as well. I think you're going to end up going down a cost curve even when you hit parity, maybe as soon as next year.
Okay. Super interesting. Let's come back to terrestrial compute. On the earnings call, you guys talked about a goal and a range of getting to terrestrial compute that could be stood up over the next 18 months. Talk to us a little bit about the visibility you have into those potential outcomes, how to think about the cost of standing up that terrestrial compute relative to what you did with Colossus 1 and Colossus 2.
Yeah.
Would love to understand the framing of what was put out there.
Yeah, absolutely. Well, one of the other things that we did put out, certainly, and it's related to the terrestrial compute and the results thereof, is we talked about the fact that by the end of this year, with annualizing our December number, we're on track, or we believe we're on track to hit $100 billion ARR. What I would tell you, an update to that is that just earlier this month, we closed another hosting deal, and that translates into about $1.11 billion a month starting December 1st of this year, which is another roughly $13 billion of ARR that starts December 1st of this year. I would say we have even more conviction now-
Got it.
...about that $100 billion ARR target by the end of this year. I think that's incredibly compelling. If you look at where our company was even a couple quarters ago to where we'll be by Q4 of this year, a lot of that's driven by certainly the growth in terrestrial compute.
Yeah.
We talked about a range next year of 5 GW-10 GW deployed.
Yeah.
Ending this year with a little over 2 GW.
Yeah.
That's a lot of additional growth in gigawatts in one year. The reason that we have the confidence is again, that vertical integration. Circling back, Eric, to that same comment, the fact that we have line of sight to the power specifically-
Yeah.
...because some of it we're going to stand up ourselves-
Yeah.
...in existing locations, I think is hugely important. Elon talked about even, I don't know, beginning of this year, maybe even last year, about the fact that he thought power would become the constraint and that people would work through the allocation of some of the rest of the supply chain, but you're seeing that already play out. In fact, even when we went through the IPO process a couple months back, when we would say, "I think power is actually going to be the issue, and that's why orbital compute will become so compelling," people were still a little, "Well, I'm not sure." I think everybody's seeing it.
Yeah.
You're seeing it across the industry now. It's pretty magical that orbital compute is clearly the solution here on the power and cooling and permitting side. We wouldn't give that range for next year unless we had line of sight to power specifically, but power and facilities and permitting needed. Certainly, we have a strong relationship with NVIDIA. In fact, we came out and talked about the fact that we're NVIDIA exclusive.
That strong relationship, we hope certainly will translate into good allocation from a compute perspective because that's the piece that we don't control. Having a great partner like that is, I think, hugely important. I think when you talk about that, the other thing that we do that certainly I think is helpful for them is when we get this allocation, we stand it up.
Yeah.
I think having certainty that you have a partner who's actually going to stand up that compute, not maybe not just for themselves, but also for the industry-
Yeah.
...it's a different dynamic, right? I think that relationship has been very strong and will continue to be very strong for us.
Yeah, that certainly was an interesting part of the earnings, the partnership with NVIDIA and the-
Yeah.
...framing of that. Talk to us a little bit broadly about your own silicon solutions over the longer term. You obviously have a plan around Terafab, but you're this very important partner to NVIDIA. How do you think about the landscape evolving between first party and third party solutions around silicon for, again, against your ambitions of how much compute-
Yeah.
...you want to try to stand up?
I think it is an interesting world for us right now. We have done these compute deals, right? Almost all of them, if not all of them, have been, in essence, 90-day with a 90-day out, so roughly six-month commits.
Yeah.
In fact, the one I just talked about that we just did earlier this month is also the same structure and timeline. I think largely that is because we have so much internal conviction about our own products. In fact, if you look at our own product cycles, it is pretty great that we closed the deal with Cursor just a few weeks ago. That brings a huge, very capable team to us officially now. You are already seeing improved product cycles. Announcing Grok Bot and getting that out there, that is a viral product. It is-
Yeah.
...really amazing. If you have not tried it, you should. It is just getting started down that path, right? Certainly it takes a little longer on the enterprise side than it does the consumer side related to security and other things you need to work through. But as people start using it and seeing all of the benefits of it and starting to talk to their friends, anecdotally, that is what I am seeing constantly around this. We are starting to see that. We saw Grok 4.6 come out and be a notable-
Yeah.
...improvement from Grok 4.5. We are talking about Grok 4.7 here shortly. The product cycle is improving, the size of the models, the capability. We are in this interesting dynamic where we have so much conviction about these products and more coming that you do not want to get in a situation where you constrain your own products-
Yeah.
...in the years to come because you were giving away the compute forever. For me and for my finance team, it is an interesting one because you are actually basically building this model of I am going to have this much compute from a wattage perspective, and this is my assumption of how I am going to monetize it. We know for a certain period of time what the range is. I think on the earnings call, we talked about $30-$50 is where we expect next year to be from a range of monetization, and that we were at the high end of that range-
Yeah.
...is probably, I think, the wording we used. If not, we are on the high end of the range. Really now it sets the bar for our internal team.
Yeah.
Right? Because you do not want to be pivoting over to your internal solutions and have any degradation related to the monetization.
Yeah.
You want to have improvement. I do not have the financials of some of these big guys like Anthropic or OpenAI related to visibility of their financials, I should say. I think our financials might be better, but whatever. But visibility to their financials to know exactly what they are getting from a monetization of dollars per watt. Based upon what we are seeing, it is higher than the range.
Yeah.
Right? That is why it makes sense for them to be coming to someone like us to get the compute. We have a very strong relationship with Anthropic and Google, for example.
Yeah.
You see that and you say to your team, "Well, look, if you just forecast that range for monetization, when our tools start ramping, it will actually be an improvement to that range.
Yeah.
Right? When we are using our compute and our software models.
Yeah.
It is pretty great because you could take the situation where you guys want to model something in that range and say, "Okay, well, I see certainty to that. I actually do not even have to have any conviction on their products and see that." All of the upside of the product starting to kick in the Grok world will all be upside to what I have in my model.
Right.
Certainly we are at the very early days of something incredibly compelling on the AI evolution, revolution, and really, we are still only in the digital AI side.
Yeah.
When you start talking about physical AI in the years to come, that even expands the market another fold.
Yeah. Understood. There are maybe two themes in there. I just want to pull on the string a little bit.
Yeah, sure.
Obviously you're talking about this third deal for compute. How do you think about the array of potential partners for those types of deals? Because you're standing up compute in a very constrained environment and delivering a solution to the industry that there's a high degree of demand for. How long do we think that narrative could sort of play out? Or how do you even think about who the people are that will call and ask you about that-
Yeah.
...sort of availability?
It feels like almost everybody.
Okay.
There might be one company that hasn't called us, and we haven't called them. But other than that, almost everybody else in the industry, we have, first of all, very strong relationships with in many ways and are talking to. I think that just It speaks to the environment that we're all seeing, right? This hyper growth dynamic that from everything we're seeing is not slowing down by any stretch. We don't predict anything other than up and to the right in a meaningful way into next year as well from everything we're seeing.
When you see that, first of all, you want to continue to invest with great software solutions and do deals like Cursor to make sure anything you can do to get your product right up there is important. But you also then have a lot of conviction about making investments for the future. The other side of this is, the risk related to these investments for us is dramatically less when we are in less than a one-year payback mode.
Yeah.
Right? Because you can see a situation, even when you are doing six months at a time, that likely continues beyond that. Now you are sitting with compute that will likely last for many years. In fact, you are seeing the monetization of many generations old now compute from NVIDIA still holding its own from a pricing perspective. And so that tells you, related to the residual value on these GPUs, that it is going to be a great product cycle for a long time. It also opens up markets from a financing perspective, right?
Because now you are taking, in essence, creating a whole asset class related to GPU financing because everyone is seeing the value of this. And in fact, in the earnings call, I talked about the fact, and one of my buddies made fun of me, said, "I am going to use that line," that not all capital is the same, right? CapEx is not the same.
When you are talking about assets that you can buy that others could use as well that have less than a one-year payback, that is very different from what the rocket guy would say is like a hangar or a launch tower that let me talk about the next couple of decades for that asset. So it is an interesting time, and it has kind of a ripple effect to a lot of other pieces of the business.
Okay, so one more piece I just want to click on that you mentioned, just to give you the opportunity. Because if I think about what has changed over the last couple of months, the closing of Cursor, the success you have had on the model side, that was not where the conversation was four months ago, and now it is where it is now, at least based on my experience. So talk a little bit about how that is evolving the strategy. You talked a little bit before about striking a balance in the business. There is compute deals, but you do not want to starve your internal operations that want to go-
Yep.
...and external for the long term. Talk about what you are learning about the opportunity set as a foundational model company, as a solution like Cursor into the enterprise space, and how that could build in momentum.
Yeah, we have certainly, up till this year, and I have been the CFO for 15 years, so we have certainly been all about organic growth, right? I think, really proud of our track record related to capital allocation and value creation. This year we certainly did some M&A between AI and Cursor so far this year. But it is because when you are faced with a product cycle as amazing as the AI product cycle, time to market becomes so critical, you then look at how do I get there, and how do I get to the very, not just the frontier, but how am I winning at the frontier?
We feel like we have a very logical strategic advantage because we are an infrastructure company. Because of the fact that we have reusable rocketry that nobody really else has. I think the Chinese are trying to get there, right? To their credit, they have got an amazing space program, and I do think in the years to come, we will see things. But not on a Starship level anytime soon, let us put it that way.
When you have that asset and you believe that that should be the solution for infrastructure for AI, and it will drive our cost of compute dramatically lower than others in the industry, then you do want to play it forward, even maybe as soon as next year, and say, well, if there is multiple solutions out there and our model is even close to at the tier one, if not better, hopefully, and our token cost is less, I think that puts us in the best position related to our customers.
Yeah.
I think in the end, that is the key, is you want to be able to say that you are delivering a better service. We can absolutely say that with rockets, most reliable rocket ever flown, let alone lower cost per kilogram. We can absolutely say that with Starlink. If you look at our Starlink product offering, no one has ever had that before. If you go on an airplane now and it has Starlink, you will know what I mean, or on a cruise ship or at your house for that matter. Now we want to be able to do the same thing with AI.
Honestly, this is a company that is only used to being in number one and winning. What I would tell you is since the IPO, I have a couple of friends that are like, "Oh, at least you got through that. Now you can relax a bit." I'm like, "Are you kidding me?" It's probably more intense now than it was. Because it's such an exciting time, and all we want to do is get out there and deliver the best product.
Yeah.
I think there's an opportunity for us, and you're seeing Grok Bot start to be part of that, and Grok continuing to evolve to that. We're not going to settle.
Yeah.
Not going to settle for anything other than going after being the best product for our customers. I think that's really compelling, and I think people are starting to see that, and that's probably the change, to your point-
Yeah.
...of maybe the talking points from a couple of months ago is people said, "Yes, I get it. I understand why you guys could be the best infrastructure solution for AI." But they were a little hesitant related to the products, and already, like 90 days later, not even-
Yeah.
...I think people are seeing, oh, they were very serious about this side of the business as well.
Yeah.
It's exciting.
Yeah, definitely. I do want to touch upon connectivity. Let's go with Starlink broadband first. You brought up the example of if you experience the product and then don't experience the product, I can personally attest there is a difference for what it's worth. But talk a little bit about what you've learned about the addressable market for Starlink broadband, especially some of the key verticals that you're attacking globally, and how you think about building or scaling competitive advantages around that business.
Yeah. What's interesting is we started out, and actually, the first year we called it Better Than Nothing Service in 2020 or 2021. We were learning, and we were getting satellites up there, and then a couple of years later, it was clear that we had a very strong service. Consumer was still a big focus because we hadn't gotten to the point of being able to go to a CTO, CIO, and show that we were three or four nines uptime and that we could give commitment levels and SLAs. But in the last couple of years, we're there, and enterprise has been a big push, and everybody wants to talk to us.
Because when you're now in a world where resiliency is so important, not having Starlink be part of your networking solution just doesn't make any sense. Right? There isn't really another solution that's anything close to Starlink at all. You're seeing that, and we're seeing that. In almost every industry, this makes sense. Whether you want to use Starlink as your backup or your primary, why wouldn't you do that? Because certainly, any Board member could look up in any of those companies and go, "Why were you down? Why wasn't there Starlink in the solution?"
I think we're seeing that on the fixed wireless side, and then we have a huge opportunity in anything related to mobility. We were talking about aircraft. We have a huge backlog, actually, on the aircraft side that we're actually ramping up our production for right now and addressing. Pretty exciting on that front. Almost every major airline, three of the four largest, we can't even talk about the one that's not in the fold because it's just kind of embarrassing. We'll make sure that we embarrass. But cruise ships, yachts, trains, any time where the connectivity wasn't there, it now can be there.
Yeah.
What's great if you just say cruises or airlines for a minute, is now it's exposure to millions and millions of folks that are getting great connectivity from Starlink that otherwise wouldn't. Now thinking, "Geez, I could have this at my home or office.
Yeah.
What we were facing in the past is actually most of our issue, if we didn't have an issue because we were selling every terminal we were making, but issue overall, on the demand side was just awareness. People weren't aware that we were there. We still face a lot of that outside the U.S., especially, where you go places and you say, "Well, you could have Starlink." They ask you what Starlink is, and you're like, it just means we have some work to do.
Opportunity set.
But it certainly is a huge opportunity, and I am very excited actually about what the next step is. What I mean by that is even on the broadband, let alone what we can talk about mobile in a minute, but the fact that AI and physical AI will probably be a large conversation when we start into next year and beyond of devices everywhere. These devices are going to want to have AI embed from a knowledge perspective, and capabilities will be there, but they also need to be connected basically at all times.
Yeah.
They can't be down. Connectivity to a humanoid robot or autonomous car or other any type of flying vehicle, that is not something that you're going to do with threshold solutions.
Yeah.
I think it's going to open up the connectivity market dramatically when you see these solutions coming out because they need so much more data for that connectivity. It's really exciting on that front.
Understood. Let's turn to the medium to longer term mobile ambitions of the company.
Yeah.
You have been very upfront about these ambitions. You have invested in assets that can build out a portfolio of solutions to deliver globally and to consumers. How are you thinking about investing, building, and scaling against the broader mobile telephony opportunity that could present in front of the company in the years ahead?
Yeah. Someone asked me just recently, "Well, is it a distraction at this point?" I said, "Not at all." What is beautiful about this model is we are going to take largely the same V3 bus that you are about to see fly in production for our V3 satellites for broadband, and we are going to change the payload out and be able to deliver starting next year, flying satellites for building out our network for direct to device. Right? For the mobile.
That is probably all through next year. Then the first half of 2028, we will be able to turn on a service that is completely differentiated from anything out there. I think that is pretty magical because then you can put that team onto the AI satellite as well next year. Right? You basically have all because of Starship, which is why I want to start the conversation always when we talk about SpaceX with Starship, this capability that nobody else has-
Yeah.
...and won't for years and years, delivers to all three of those businesses. That broadband business, that mobile direct to device business, and the AI satellite business. Flying those satellites in those constellations will be really, really terrific related to getting our cadence up on our own rocket business as well. But circling back to direct to device, we're planning to fly the satellites next year and have the capability to turn on service first half of 2028. Now what that means is we've got some time to figure out exactly if there are other assets we need related to it. We can look at that.
But what we went out and did is we saw that this is where the business was going, and it just looked like a missed opportunity to not be also in mobile when you have this capability to deliver from space that no one's ever had before. We turned on our version one of this and flew 600 satellites very quickly within a year. Right?
And turned on the service that, for example, in the U.S., you look at T-Mobile, and they call it T-Satellite, I think, and that's the Starlink version one. Right? But that's more like a text-only service and maybe light voice, and it's awesome for disaster recovery. It's awesome from an emergency perspective. If you get into a dead zone somewhere, you can still connect, but it's not what you and I are used to with full 5G.
Yeah.
The next version of this, what we're going to start flying next year with the satellites, is full 5G quality service. Now it's a matter of when we go to market, we bought the spectrum, the mid-band spectrum from EchoStar for U.S.
Yeah.
We feel like we've got the spectrum where we can go turn on service. We've got that, and it has direct from space and terrestrial capability already, FCC approved. So if we need to, we can stand up terrestrial solutions to put the terrestrial piece in place too, or we could work with existing carriers. We're open to, I think, a lot of different ways to do it, but we're marching down that path in the U.S. Overseas, we're probably going to have to go through the same process that we did with broadband.
We started out with a service that was approved in the U.S., and then over the next couple of years went, and now we're in over 170 countries around the world with Starlink on the broadband side. We worked with each regulator in each country really just to demonstrate that we are bringing a product offering that is better for the consumer. Again, I think it circles back to, is this a product that is better than what is out there? For sure it is on the broadband side, and for sure it will be in an industry that just has not innovated for decades, if you think about the wireless industry, right? In essence, the service you have is largely the same.
Yeah.
I do not even remember when 5G was talked about. It was a long time ago.
Yeah.
I was at Qualcomm way back in the day myself. Now you are talking about being able to come to market with a product that should not have dead zones, can have global roaming. You can go anywhere you need to with it, and that is pretty magical, especially if you are charging same or slightly less than a service that does not have any of those things.
I think that is why the big three operators were a little concerned about us and wanted to create a cartel maybe. I do not know what this term is. I am not a lawyer, and I will let them figure that out. The point being, I think that the whole industry is seeing that we have the ability to deliver a service that will be really compelling for people.
Yeah. Let me try to tie it together with one final question. You face all these decisions about what to invest in. All these different durations and return profiles. Talk a little bit about allocating capital to all of these opportunities, and then bring it home with the theme of what are you the most excited about to continue to demonstrate to the market in the years ahead?
Yeah. I think our track record on capital allocation kind of speaks for itself. What it looks like, there's a lot of different things we're investing in. I think really right now, you should think we're investing in Starship specifically because it enables our other businesses, and we're investing in terrestrial compute because the better we get at that, I think will help us on the orbital compute side and will also be an incredible funding mechanism for some of the other pieces of the business.
I think in the longer term, I think right now the piece that's still a little misunderstood is orbital compute, simply because I think the timeframe that others have related to when is this going to be bringing gigawatts or tens of gigawatts to the table is different than what we're actually going to be able to deliver.
Okay.
I really appreciate the opportunity. Thank you.
Bret, thanks for coming. We really appreciate you being part of the conference. Please join me in thanking SpaceX for being a part of-