Inside Starcloud’s $2B Space Factory
Molly O'SheaPhilip JohnstonEzra FeildenAdi OlteanBailey MontañoMiguel RiveraCamilo Lopez
- Starcloud’s claim to cheaper orbital compute depends on launch prices below around $500/kg. Philip Johnston compares that threshold with about $5,000/kg on Falcon 9 and Starship’s designed internal costs of $20–$40/kg. Continuous sunlight removes battery and backup-power requirements; he also cites an eightfold per-square-meter energy advantage for space solar panels. “The whole thing is actually contingent on the launch cost.”
- The $2 billion company established its operating model with a roughly $2.5 million satellite built and launched within 21 months of founding. Johnston says a prime contractor quoted $75 million–$100 million for Starcloud 1; the company now has 30 employees, only two non-engineers. A broken $200,000 vibration table repaired over a weekend captures the approach: “not take no for an answer.”
- Chip reliability could break the economics even if cooling works. Johnston says a 10% or 20% higher chip failure rate in space would make “the whole economics of the model fall apart,” because chips dominate costs. Starcloud 2 underwent three months of chip burn-in; its planned January launch is the next proof point for radiators Johnston claims deliver at least tenfold lower mass per watt than the ISS.
- Government satellite inference is an initial use case before orbital compute reaches terrestrial cost parity. Johnston describes test-phase workloads processing radar and hyperspectral imagery, replacing large image downlinks with actionable coordinates in “minutes rather than hours and days.” Longer term, he expects coding agents, business-processing agents and ChatGPT queries, while tightly networked frontier-model training remains harder to accommodate.
- The scaling plan combines small independent compute platforms with mass manufacturing. Starcloud 3 is designed around roughly two tons and 235 kW, with up to 50 vehicles per Starship launch; Johnston describes an initial 88,000-satellite constellation supporting about 20 GW. His larger estimate is about 10 TW in dawn-dusk sun-synchronous orbit, approximately 20 times the entire US power grid.
- The $450 million raised is funding launch reservations and a factory before the launch market fully opens. Johnston hopes commercial Starship capacity becomes available from 2028 onward, while Starcloud has also booked alternatives including Stoke and Firefly. Approximately 100,000 square feet could support production of 100 satellites weekly, with an eventual line “like an auto manufacturing line.”
- Johnston’s largest investment prediction is that SpaceX might exceed $100 trillion in value within ten years. Molly O’Shea frames that outcome as contingent on execution and no world war; Johnston’s mechanism is physical recursive manufacturing once “Optimus is producing Optimuses.” He also categorically predicts SpaceX will buy Tesla, calling it “just a fact”; the interview supplies his assertion and timing theory, rather than evidence of a transaction.
1. A $2.5 million satellite established the operating model
- Johnston says Starcloud went from founding to orbit in 21 months, a record against a previous four-year benchmark. He says Adi and Ezra built the earlier fastest satellite by hand. Starcloud 1 launched in November “last year” carrying NVIDIA H100 hardware. A prime contractor quoted $75 million–$100 million for the mission; Starcloud did it for about $2.5 million. His lesson: “you can do things very cheaply.”
- Seattle supplies the relevant combination of skills. Johnston says 90% of satellite mass launched in the preceding few years was designed, built and tested in Redmond. About half the team previously worked at SpaceX; the other half comes from hyperscalers, with AWS and Azure nearby. The company has 30 employees, only two non-engineers, and deliberately keeps staffing lean.
- Procurement illustrates the approach: a new vibration table cost about $1.5 million with a roughly two-year lead time, so Starcloud bought a broken one for $200,000. The 16-ton machine arrived waterlogged, with a major internal leak. Despite an expert declaring repair impossible, Feilden and the team rebuilt it over a weekend sufficiently to complete spacecraft testing: “not take no for an answer.”
2. Cooling is a manufacturing problem with an orbital test ahead
- Starcloud representatives identify cooling and radiators as the largest technical challenge, while framing the remaining problem as engineering cost rather than feasibility: “We know it’s possible.” Johnston supplies the commercial question: “how do you make it super cheap and how you make it manufacturable at scale.” The team adapts liquid-cooling technology used in terrestrial data centers.
- Starcloud 2 has a roughly 120-foot deployed wingspan. Johnston says its January launch will carry “by far the largest commercial deployable radiator in space.” Against the International Space Station, he claims at least tenfold lower radiator mass per watt dissipated and about 500-fold lower cost. He concedes the ISS is an easy cost benchmark; the mass improvement is the achievement.
- A Starcloud representative explains the accumulator being vibration-tested: heated coolant expands, and the component accommodates that expansion without bursting the cooling system. Launch qualification also checks resonance. Johnston thinks SpaceX’s minimum first resonant frequency is around 40 Hz; a structure oscillating slowly could move neighboring payloads and destabilize the rideshare stack.
- Engineering manager Bailey Montaño designed and fabricated the radiators and recalls the first fully integrated cooling-system test in Colorado. Final satellite assembly then had only three weeks available, with Montaño building the bus at the provider’s office. Testing is Montaño’s favorite stage: “That’s where the rubber really hits the road.”
3. Launch below $500/kg is the terrestrial crossover
- Johnston’s comparison starts with a new solar project powering a terrestrial data center. Its largest cost is permitted land, depending on location; next come batteries and backup power, then solar cells. In the intended orbit, he says Starcloud needs neither permitted land nor batteries and backup, because it remains in sunlight continuously.
- Johnston says Starcloud needs eight times fewer solar cells, citing an eightfold per-square-meter energy advantage for space solar panels. Launch is the principal added expense. Combining these savings and costs produces his crossover of “around $500 a kilo,” versus about $5,000/kg today on Falcon 9.
- Starship’s designed internal launch cost, in Johnston’s account, is $20–$40/kg. The model therefore works if SpaceX sells capacity at anything below roughly ten times marginal cost, or potentially more. He considers that plausible because Falcon 9 currently charges two or three times marginal launch cost. Those are his expectations for future pricing, rather than an available commercial Starship quote.
4. Radiation and chip failures can erase the savings
- Johnston considers chip survival more pertinent than cooling, which he says the team has “basically solved.” Radiation shielding and particle-accelerator testing matter because chips are the most expensive component. If their failure rate is 10% or 20% higher in space, “the whole economics of the model fall apart.” Lower infrastructure costs cannot absorb that loss.
- The satellites will be inaccessible after launch, so the operating model resembles Starlink: duplicate critical systems such as pumps and over-provision components that degrade. Solar panels, for example, lose a few percent annually. Shielding and testing address radiation exposure; redundancy and excess capacity address failures the company still expects during service.
- Starcloud 2’s chips underwent three months of hot, full-power burn-in to expose early failures before launch. Johnston calls this chip “infant mortality” and says subsequent failure probability diminishes dramatically after surviving that period. Asked why early failures happen, his answer is simply “I don’t know”; the mitigation rests on the observed failure pattern.
5. Government inference supplies the first workloads
- Starcloud 1’s payload includes three NVIDIA chips, two Arm chips and AMD CPUs. Johnston says it trained the first model in space, nanoGPT from Andrej Karpathy, and “ran the first version of Gemini.” The satellite has since moved into useful, mainly test-phase workloads for government customers, including inference on synthetic aperture radar and hyperspectral data.
- His concrete example is monitoring whether a vessel crosses a boundary. Conventional satellites collect enormous imagery volumes, may wait at least an hour and a half for a ground station, and cannot necessarily downlink much. Optical links can instead send imagery to Starcloud for inference; it then transmits vessel coordinates in “minutes rather than hours and days.”
- On security, Johnston argues that striking a satellite moving about 17,000 miles an hour is harder than attacking a Virginia data center. He points to Starlink and Starshield usage in Ukraine as existing cases of adversarial pressure. Starcloud is using the same kind of cryptography, he says, to protect its satellites.
- Johnston also argues that approaching objects would be visible, claiming objects larger than a one-inch cube are trackable. He describes interference with the satellites as an act of war against the US and invokes the Outer Space Treaty and Space Force’s deterrence role. These are his security and legal characterizations.
6. Launch availability constrains the transition
- Feilden says booking Starcloud 1’s Falcon 9 rideshare roughly 21 months ahead was essentially the company’s first action; the initial reservation could be made by credit card. That simplicity is changing. He describes Falcon 9 rideshare as going away while Starship has yet to ramp: “we’re definitely in a crunch time at the moment.”
- Launch insurance was relatively inexpensive given Falcon 9’s record, which Feilden describes as 300 consecutive launches without a failure. Insuring a first-of-its-kind spacecraft’s subsequent operation is harder. He remains “slightly on the fence” about payload insurance: financial compensation cannot recover lost time, the startup’s most valuable resource. Future coverage will be decided case by case.
- Johnston expects Starship reusability probably within a year, or at least upper-stage capture; reflight might take longer. From 2028 onward, Starcloud hopes SpaceX will sell commercial capacity. It has also booked other vehicles, with Stoke and Firefly already announced and further announcements expected, to provide options if that timing slips.
- Asked about Rocket Lab’s delayed Neutron, Feilden says the company announced a 2027 first flight the previous day. “Composite manufacturing is really hard,” especially large structures and first-time tooling. Nevertheless, he has substantial confidence in Rocket Lab’s ability to operate regularly, citing its 100 Electron launches: “we’re definitely rooting for them as well.”
7. Independent modules are the route to gigawatts
- The Starcloud 3 manufacturing demonstrator represents a roughly two-ton fully loaded vehicle rated at 235 kW. It is designed to pass through Starship’s “PEZ dispenser door,” with up to 50 per launch: two across and 25 layers tall. Upcoming individual missions still use Falcon; “the future is definitely Starship,” Feilden says.
- Johnston distinguishes this architecture from terrestrial frontier-model training, which requires a highly interconnected cluster. For a long time, he expects Starcloud will probably avoid docking spacecraft together. Most orbital workloads would therefore be inference: coding agents, back-office business-processing agents and ChatGPT queries, distributed across independent platforms.
- Using 200 kW units in his deployment arithmetic, Johnston estimates close to 10 MW of additional compute per Starship launch. He describes a filing for 88,000 satellites supporting the first 20 GW. “That’s just the first 88,000,” he adds; the planned constellation is an initial deployment rather than his eventual limit.
- Johnston thinks dawn-dusk sun-synchronous orbit can accommodate about 10 TW, roughly 20 times the entire US power grid. Beyond that, he names cislunar space, MEO, beyond GEO and Lagrange points. Answering Molly’s concern about blocking sunlight, he says the polar geometry keeps satellites out of Earth’s shadow and prevents them casting shadows on Earth.
8. The capital funds launch slots and a production line
- Starcloud is expanding into adjacent buildings, with one section coming online in November and another available in February next year. About 100,000 square feet could support 100 satellites weekly. Space is needed both for manufacturing and for testing Starcloud 3’s 400-foot deployables. The intended result looks “like an auto manufacturing line,” with chassis receiving components as they move.
- Molly puts total capital raised at $450 million. Johnston says part is reserving constrained launch capacity, particularly from 2028–2029 onward; the remainder supports factory fit-out, hiring and materials. He says the company is making about a million dollars a month in interest. The cash will increasingly be spent as the build-out progresses.
- Staffing could reach 50 by year-end and maybe 100 by the following year. Engineers alternate between CAD work upstairs and hands-on wiring and testing below. A technician group—four, or perhaps five, Johnston says—handles fabrication, including the aluminum compute-module enclosure. The production model keeps design and physical assembly closely connected.
9. Johnston’s SpaceX prediction depends on recursive manufacturing
- Johnston thinks SpaceX “might be worth more than $100 trillion in 10 years.” Molly explicitly conditions that prospect on plans working and no world war. His mechanism is Optimus building Optimus factories: once “Optimus is producing Optimuses,” manufacturing growth accelerates. He compares the process with recursive self-improvement transferred into the physical world.
- His Tesla prediction is categorical: “This isn’t a hot take. This is just a fact. SpaceX is going to buy Tesla.” Johnston ties timing to stock unlocks and concerns that participants would dispute the price. He predicts few Optimus announcements for the next year, with more following a merger. The conversation provides no transaction evidence beyond his assertions.
- On the Roadster announcement expected that evening, Johnston says he has heard about cold-gas thrusters from the spacecraft and a compressor compressing air to fire out the back, enabling extraordinary acceleration and cornering. He also thinks lift is possible. Molly suggests it might be an eVTOL. He disagrees: “I would be extremely surprised if it’s an eVTOL,” arguing that such a development would already be known.
10. Fermi Explorer extends the cheap-spacecraft thesis to another star
- Johnston describes a separate nonprofit effort, Fermi Explorer, to send the first spacecraft toward Alpha Centauri, about four light-years away. Starcloud 1’s cost experience encouraged the idea. An MIT physical AI lab analyzed a trajectory he says could put a spacecraft on that journey for about $10 million–$15 million.
- The mission would use existing technology: solar-powered Hall-effect ion thrusters, five years spiraling toward the Sun, then another five years firing around perihelion into expanding ellipses before departure. It targets where the moving star system will intersect the spacecraft’s path in about 80,000 years, rather than its present location.
- Johnston hopes it becomes both “the first to leave” and “the last to arrive,” overtaken by later propulsion technologies. Even propulsion only 20% faster, he imagines, might allow a colony to have existed for 15,000 years before arrival. Its inhabitants could applaud the spacecraft across the finish line and put it in a museum.
- The underlying motivation is the Fermi paradox. Johnston estimates even today’s technology could settle the galaxy in roughly five to ten million years, making the lack of visible evidence troubling. He sees two possibilities: humanity is exceptionally rare, which he welcomes, or “intelligent life doesn’t last very long.” The mission is intended to make people consider that choice.
Full transcript
We’re with Philip Johnston, the CEO, and we’re going to go through the $2 billion company and what they’re building behind us.
Well, that’s the satellite we launched in November last year with the first NVIDIA H100s on board.
Starcloud 1 separation confirmed.
They told us it would cost $75 million to $100 million to do Starcloud 1. We did it for about $2.5 million. If the launch cost gets below around $500 a kilo, then it will be much cheaper to do this in space than terrestrially. We think we can fit about 10 terawatts in this dawn-dusk sun-synchronous orbit, so that’s about 20 times the entire U.S. power grid.
It’s around 2 tons, fully loaded, 235 kilowatts, and we can deploy up to 50 of them per Starship launch.
Do you think Tesla will merge with SpaceX?
This isn’t a hot take. This is just a fact. SpaceX is going to buy Tesla.
We’re here at Starcloud in the Seattle region. We’re not going to disclose the exact location because it’s a secret, right, Philip?
Top secret.
Top secret. Okay. We’re with Philip Johnston, the CEO, and we’re going to go through the $2 billion company and what they’re building behind us.
Welcome to Starcloud. On the screen, we actually have the location of Starcloud 1. That’s the satellite we launched in November last year with the first NVIDIA H100s on board. This little green area shows everywhere on Earth where you have a line of sight to the satellite. That means if we have ground stations anywhere in this green area, we can see the satellite. It’s moving about 17,000 miles an hour, so it takes about an hour and a half to circumnavigate the globe. If we come back in half an hour, you’ll probably find it over Africa somewhere.
Oh my God.
Yeah, it’s moving fast.
Before we walked over here, it was down here. Did something special happen over here?
Oh, yeah.
Why?
It actually just came over Cabo, which is where I proposed to my fiancée. It only comes over there once every 3 days, so I waited until 6:00 a.m. and then showed her on the phone, because I’d written the proposal on Starcloud 1.
Yeah.
So, yeah.
It’s cute. Check out the post if you haven’t yet. Okay. All right, we’re going to walk through the facility. Just give us a sense of what we’ll be looking at and what stages of development. Are we going to see Starcloud 1, Starcloud 2, maybe—
Yeah.
—Starcloud 3 anywhere close?
Yeah. Starcloud 1 is launched. Starcloud 2 is in the assembly facility. As I say, we’re midway through move-in, so please take it with a pinch of salt. There are boxes everywhere. Then we’ll go through to the fab shop. We’ve welded up a mock-up of the chassis of Starcloud 3, so you can get a sense of how big it is. Then we’ll go through and finish the tour in some of the newer facilities that we’re bringing online next year.
Awesome. Cool.
Cool. This is the main assembly area for Starcloud 2, and this is the satellite in the background. You can see Starcloud 2. On top, you can see the 2 lasers, and then we have some deployables. It has about a 120-foot wingspan when deployed. We just put this through vibration testing, so this is our vibe table here.
A fun story about this vibe table: we couldn’t find a vibe table in the U.S. that would vibe with a pressure vessel on it in the time we needed. The lead time on new vibe tables is almost 2 years now.
Oh no.
Yeah, and it’s about $1.5 million for a new vibe table, so we bought this secondhand broken vibe table for $200,000. Then the team worked all weekend to fix it up. I don’t know if you can see—maybe we can go around so you can see how janky the amp we bought for it is.
All right, we have a new entrant on the video. We’re here with Ezra. Ezra is also a co-founder. Do you want to give a couple of sentences on your background? Then we’ll talk about the vibe table.
For sure. My name’s Ezra Feilden. I’m the co-founder and CTO at Starcloud. Before Starcloud, I was with Airbus, designing and building satellites, focusing on large deployable structures, which are a very critical piece of what we’re building here. Before that, I did my PhD in materials engineering.
Damn.
We’re standing in front of our vibe table, our very old, very beaten-up vibe table. We had to get this on an extremely short lead time. New vibe tables have a 26-month lead time and cost over a million dollars, and that didn’t work for our schedule. We had to do this test in-house for various reasons, so the only option available to us was to get a secondhand vibe table. This one, I can tell you, has had a life. It’s as old as I am. It came to us filled with water—
No.
—and with a horrendous internal leak. There’s very high voltage inside, and obviously it’s vibrating. There are high temperatures, and right in the heart of all of that, there was a very large leak, which is evidently the reason the previous owner sold it. We didn’t realize that until we got it in-house, essentially 2 days before the test.
Oh my God.
Through the weekend, evenings and weekends, we had to hustle to tear this thing down right to the core and repair the leak. We had an expert come—basically the country’s expert in vibe tables—and tell us the repair was not possible. We pressed on and got it limping along enough to complete our spacecraft-level vibe test.
So do you find this on Facebook Marketplace? Where do we find old vibe tables?
Yeah, it tends to be auction houses, surplus auction houses. When companies go out of business, they get rid of all their assets. There are some diamonds in the rough there, for sure.
It’s not easy to move. It’s about 60 tons.
Oh my God.
16 tons.
Sorry, 16 tons. Yeah, which is—
Yeah, it’s hefty. To power this thing, we need a generator, so we have a very large generator outside.
I know when I was pulling up, I thought, “Oh, we’re in kind of a warehouse office park. Which one could possibly be Starcloud?”
Yeah.
Probably the one with huge generators outside.
Yeah. When you have extremely tight timelines and you’re trying to do new things very quickly, you often find yourself in these kinds of situations where you have to get very creative, hustle, and basically not take no for an answer. You have to be incredibly scrappy. That’s our superpower as a startup, I think: the ability to turn these crazy scenarios around and make a success of them.
It’s a great point, because Starcloud went from founding to orbit in 21 months.
Yeah, which is a record. I think the fastest anybody had done that before us was 4 years, and that was basically because Adi and Ezra built the satellite by hand. We’ll speak with Adi in a minute, but—
Oh my God. I know this might surprise people. It definitely surprised me, because now I’m doing more of these tours, and the locations aren’t El Segundo or down in Long Beach. You guys are up here in the Seattle region. So why Seattle?
Maybe people don’t realize this, but 90% of all satellites launched in the last few years, by mass, were designed, built, and tested in Redmond, Washington, which is exactly where we are now. Ezra and I were literally sitting in El Segundo for the first month of the company, typing into LinkedIn, “Okay, spacecraft thermal engineer.” It was literally going, “Redmond, Redmond, Redmond, Bellevue, Seattle, Redmond, Redmond.” We were like, “Okay, shit, I guess we have to move up.”
About half our team formerly worked at SpaceX. The other half came from the hyperscalers, so we have AWS and Azure down the road. For doing data centers in space, this is literally the best place in the world you could be.
Okay, we’re also joined by another Starcloud co-founder, Adi. Adi, tell us a little bit about your background, and how did you end up here?
Before Starcloud, I was at SpaceX for about 2 years. I was the responsible engineer for a feature called Tracking Beams, which essentially gives Starlink to all in-motion vehicles, including aviation, cruise ships, and even Starship. If you’ve ever seen those cool Starship videos, I helped build that. Before SpaceX, I was at Microsoft for about 2 decades. In the last couple of years, I was working on large-scale GPU deployments in production.
We onboarded the very first large language models, before they were called LLMs, in a bunch of data centers, and worked very closely with NVIDIA on that. It was really rewarding to be part of 2 keynotes at GTC: one in 2019, where Jensen talked about conversational AI. That was—
Mm.
—what it was called back then. Our demo was part of that keynote. Then later this year, at the same GTC keynote in San Jose, Space-1 was introduced—the first GPU designed to work in space.
Damn. Yeah, Philip was telling me you’ve been mentioned in GTC talks multiple times.
Not me personally, but our team, yes. I helped build the demo.
That’s incredible. So we’re standing in front of Starcloud 2. Do you want to give us a little rundown on it?
People are busy at the moment. They’re about to fire the HDRM, which is essentially a latch that will deploy the solar panels on each side of the spacecraft. If you hear a loud pop, that’s it. We’re testing the satellite actively and testing various systems. There you go.
Okay.
So it’s a huge event.
Oh, really? I was like, something’s gonna explode in here.
Maybe not as climactic as we wanted, but this is actually the most important deployment event that happens in orbit. We use that to kick off the deployment of both solar panels in orbit and make sure we can get kilowatts of power to all our compute.
So I know this is a very new facility, but where does that get made? Where are all these components coming from? Is it vertically integrated? How do you think about engineering?
We try to vertically integrate as much as possible. Obviously, we’re not manufacturing the NVIDIA boards or other high-level compute systems, but we integrate everything here. We manufactured our radiators and put together the solar panels. A lot of the cooling system is being built internally here, which is kind of our secret sauce, if you will. Another HDRM firing. We try to integrate as much as possible in-house.
What’s been the biggest technical challenge coming over to Starcloud?
I think the cooling system and the radiators are by far the largest technical challenge. I think we’re very proud of what we did, how it works, and how compact the whole thing is. It’s mostly a focus on engineering costs rather than whether this is possible. We know it’s possible, but—
The question is how you make it super cheap and how you make it manufacturable at scale.
Yeah, I mean, that’s a big point. I was interviewing Sean Maguire, and he was talking about data centers in space. He was like, “I did the math on it, and it makes sense.” The math maths.
Yeah.
A large component of that is energy transfer and how much energy is used on Earth for Earth data centers versus in space. This is going to be a little bit of a different environment. It is very, very cold, but you still need radiators, cooling, all that kind of stuff.
Yes.
So could you explain that out a little bit more?
The engineering math is hard, but the economic math is the one that truly matters in orbit. What you want to do is dissipate every single watt you’re receiving from the sun. That’s the hard part: how to do it very efficiently. How do you essentially emulate a cooling system that mimics the one in data centers today on Earth?
As people probably know, 2025 was the year of liquid cooling in all the data centers. People have to switch to liquid cooling because that’s the only way to achieve the incredible densities of compute in next-generation data centers. We’re capitalizing on that and using the same technology in space as well.
Mm-hmm. For those who don’t know about the Fermi Explorer mission, can you explain that and why it’s significant for you?
1. The Fermi Explorer Mission
My co-founder, Zadi Hazra, and I have created a nonprofit with a few others, including Cameron Gefaren. We created a nonprofit to send the first spacecraft to the nearest star. The nearest star system to us is Alpha Centauri. It’s about 4 light-years away.
We realized after doing Starcloud 1 that you can do things very cheaply. For example, we had a quote from one of the primes. They told us it would cost $75 million to $100 million to do Starcloud 1, and we did it for about $2.5 million.
No way.
Yeah. So we realized you can do things very cheaply, and we found—
You can just put things in space.
Yeah, pretty much. We had a physical AI lab out of MIT run this analysis and find a trajectory that allows us to get a spacecraft on a trajectory to Alpha Centauri for about $10 or $15 million.
It won’t use any technology that doesn’t exist today. It’s going to use Hall-effect ion thrusters powered by solar energy. We’re going to spiral in toward the sun for 5 years, and then we’re going to fire the thruster at perihelion, or the closest point to the sun, for another 5 years to create these ever-expanding ellipses.
On the last firing, we’re just going to fling off in the direction. It’s not even anywhere close to where Alpha Centauri is today. Alpha Centauri is moving toward us, and it will intersect on a path in about 80,000 years. The hope is it will be both the first to leave on this journey to another star and also the last to arrive.
So, hopefully, in 50 or 100 years, there’ll be better propulsion technology. In 1,000 years, we’ll have much better propulsion technology. I hope stuff starts flying past us. By the time we arrive in 80,000 years, maybe there will have been a colony for 15,000 years, even if your propulsion technology was only 20% faster.
Yeah.
What I hope happens is they applaud our spacecraft across the finish line. They capture it and put it in a museum or something. From there, it’ll take 5 to 10 million years to settle the galaxy, even with just today’s technology. Then it’s about a billion years to get to Andromeda, our nearest galaxy. From there, it’s about another 5 billion years to get to the local group of 50 galaxies.
The purpose for me of doing it is that either all of that happens or it doesn’t, for some reason. There’s this idea of the Fermi paradox: the question of why we don’t see more intelligence in our galaxy or more obvious evidence of it.
You mean aliens.
Yeah. Or why our solar system isn’t part of a galactic empire.
Yeah.
It’s pretty quick: 10 million years to settle the galaxy is the blink of an eye in galactic time frames. If there had been intelligent life on any other star in our galaxy—there are a trillion, trillion planets in our galaxy—and there had been intelligent life on any other one in the last 10 billion years, then it would have taken 5 million years to settle the whole galaxy. The galaxy has been around for more than 10 billion years, so it’s really concerning that we don’t see more evidence of that.
To me, there are 2 options. Either we’re exceptionally rare—maybe we’re the first species or intelligence to get to this level of complexity in the whole galaxy, or maybe even in the local group of 50 galaxies—or intelligent life doesn’t last very long. By the way, I’m open to the idea that we’re exceptionally rare. I like that idea. Maybe we’re going to be the first to settle the galaxy, and that’s a great privilege.
Damn.
Maybe there’s been intelligent life on many other stars in our galaxy, and for some reason it doesn’t survive very long. Either way, I think we should know about that. One of the reasons to do this mission is to make people think more about our place in the galaxy, the Fermi paradox, and other things.
How did you get into space, and how did you get interested in this?
I’ve always been interested in space. I wanted to be an astronaut as a kid, like anybody.
But why? What exposed you to that?
I guess sci-fi movies when I was a kid. I read the Asimov novels, the Foundation series. There’s a phrase that people sometimes say: “We’re too late to explore the Earth, and we’re too early to explore the solar system and the galaxy.” I always thought that was really sad.
It was exciting to me to be an explorer, and the voyages across the globe were exciting to me. Realizing, okay, we’ve explored everything now, everything’s done, was kind of sad. Recently, I started thinking, “Maybe we can make our own effort. Maybe we can be the first to send a spacecraft to the nearest star.” What an incredible privilege that would be.
So if you wrote a sentimental note on Starcloud 1, what are you gonna put on the spacecraft?
I’m definitely gonna put the name of my fiancée on there.
So, for the low, low price of—
The low, low… Oh, you’re going to start selling logo spots?
Yeah, yeah.
Wow. We’ve got an advertising option.
Yeah, there you go. People can go to fem explorer.com/donate, and they can put whatever message they want on there. But what would I put on there? I’m excited to put lots of information, like records of what Earth is like. We’ll have some scientific payloads on there, and we’ll have the Golden Record, which is one of the artworks they put on one of the probes going out of the solar system. But we’ll also have a solicitation period for various artistic payloads, like artworks that people can put on there.
We’re drawing ideas from various places. One cool scientific payload—did you ever see this fruit fly that Google—or, there was a startup. They cut—
The brain.
They sliced the brain of a fruit fly.
Yeah. Mm-hmm.
Yeah, so we’ll put a digital copy of that, this fruit fly thing.
Whoever finds this is going to be like, “Humans, they’re tiny.”
Yeah, yeah. Yeah.
This—
They’ll be like, “How the hell did they build this?”
This made this?
Yeah, yeah, yeah.
That’s awesome. All right. I have to address this because every time… We’ve done plenty of hot takes now. We did one at the New York Stock Exchange. We did one in Paris at the Raise AI Summit. So I have to ask you now: What’s your hottest take?
2. SpaceX and Tesla Predictions
Every time I do a hot take with you, I get in trouble. Last time we did it, I said, “You should buy SpaceX stock,” and then Elon retweeted it.
But you were right, and it rocked.
I was totally right. I was 100% right.
Okay.
What was my hottest take? Let’s stick with the theme. Elon posted that they’re going to be worth $100 trillion in 10 years. I think they might be worth more than $100 trillion in 10 years.
You do?
Yes.
If all goes to plan and there’s no World War, you believe—
That’s my hottest take.
—SpaceX will be worth more than $100 trillion.
Correct.
Okay. Well, that’ll be good for business.
Yeah, exactly.
What would make that up, though? Break that down. How do you think about it?
Actually, a good chunk of it will come from the fact that Tesla’s Optimus will be up and running. Once Optimus is building Optimus factories, you have this incredible exponential growth. The rate of growth for all of manufacturing will increase a lot once Optimus is producing Optimuses. It’s a bit like what’s happening now with recursive self-improvement, but for the physical world.
Do you think Tesla will merge with SpaceX, which will help with that?
This isn’t a hot take. This is just a fact.
Okay.
SpaceX is going to buy Tesla, and they have to wait until the stock unlocks, because if they do it now, everyone’s going to say the price wasn’t fair. At the moment, you’re not going to see too many announcements, but I think there’s a Roadster announcement tonight.
Mm.
Beyond that, I don’t think you’re going to see any announcements about Optimus for the next year. Then once they do the merger, we’ll hear a lot more.
I know this will come out after the Roadster announcement, but what’s your prediction?
What I’ve heard is they have these cold-gas thrusters from the spacecraft. They have a compressor that’s compressing air, essentially, and then that’s being fired out the back, which gives you absurd acceleration and cornering and everything else. And also, I think—
Lift?
Lift. Yeah. I actually think you can get lift on these things.
Oh my gosh.
What do you think?
I kind of thought it might be an eVTOL.
That would be wild, but I don’t think it is.
You don’t think it is?
We would know by now if it’s an eVTOL.
Think you could just skip the whole car thing and go straight to eVTOL, call it a Roadster?
That would be amazing. That would be incredible, but I’d be extremely surprised if it’s an eVTOL.
All right. Stay tuned for our predictions. Oh my God. I don’t know. This one is a little bit more than length.
Length and girth. Length and girth.
That’s crazy.
Yeah, so—
So what do you use this for? What—
This—
What do you use this for?
This is for screwing in bolts—
Yeah.
—at the very—
That’s crazy.
Anyway.
Where did you find this?
I went to a Halloween store this morning because I was—
Oh, okay.
I was like, well—
Well, that’s actually pretty impressive for a Halloween store.
I can’t be sword mugged by the rest—
Yeah.
—of your guests.
That’s wild.
Okay, so we just finished the vibe table. We finished looking at Starcloud 2. Now you want to go into—
We’ll go to the fab shop so we can see where we’re welding and fabricating various aluminum and steel parts, some for the actual spacecraft, and most for test equipment and other things. Maybe Ezra can talk a bit about the fab shop.
3. Building Starcloud 3
Yeah, this is a big upgrade from our last facility. We’ve literally just moved in here over the last few days, so everything’s a little bit chaotic. We moved from this tiny office building into this lovely manufacturing space.
This area is what we call the fab shop. This is where we do dirtier work. The cleaner work goes on in there; the dirtier work happens here. We’re talking sheet-metal work, welding, and general fabrication. We have lots of fab-shop equipment here: MIG welding, TIG welding, laser welding, and a shear for sheet metal.
Right here we have the manufacturing demonstrator for Starcloud 3. Starcloud 3 is our largest platform. It’s a 2-ton vehicle, approximately. We wanted to get some experience building a primary structure at the scale of our Starcloud 3 vehicle. This vehicle is designed to go through the PEZ dispenser door for Starship. It’s around 2 tons fully loaded, 235 kilowatts, and we can deploy up to 50 of them per Starship launch.
My God.
So this is the Starcloud 3 manufacturing demonstrator. Per layer, you get 2 of these.
Wide, and then—
Two wide, and then 25 tall.
Oh my God.
So it’s from here to the ceiling, with these stacked in two rows.
So right now we’re in the middle of Starship. We’re inside of it.
Yeah. Yeah.
Okay.
Yeah, that gives you an idea of the diameter of Starship, the fact that this can fit in there. It’s an enormous vehicle.
Where have you guys been launching to date? What are the launches? I know you have 3 launches coming up as well.
Mm-hmm. Our first launch was with Falcon 9, a Falcon 9 rideshare mission. Our upcoming launches are all with Falcon. It’s really hard to beat the benefits and economics of rideshare when you’re producing single missions like we are at the moment. The future is definitely Starship, and that’s what we’re gearing up for here.
What’s the process like for booking a ride?
Booking rideshare is unbelievably easy. It was basically the first thing we did as a company: we booked the launch for Starcloud 1.
Do you have to book it really far in advance? How does that work? You’re gunning to get on the vehicle, but how does that work?
We booked about 21 months in advance. The initial booking is—you can pay for it with a credit card and get a confirmation straight away.
You could.
You could.
Could.
Yeah.
Falcon 9 rideshare is going away, though. These were the rosy days, when it was much easier to get a single payload launched. We’re in a transitional period at the moment where Falcon 9 is being wound down and Starship has yet to be wound up, so we’re definitely in a crunch time when it comes to launch.
What’s the process for that, though? Do you need insurance? What happens if it doesn’t go to plan?
Insurance is relatively inexpensive for Falcon 9 because it has such a strong track record: 300 consecutive launches without a failure. You can insure against a failure on launch—if the rocket explodes or the payload doesn’t separate, that kind of thing. It’s much more difficult to insure against things going wrong with the spacecraft, since this is the first spacecraft of its type to be built.
Mm.
We got insurance for the launch, and if something goes wrong, that can give you a bit of a financial payout. But what it can’t compensate you for is lost time, which is really the most valuable resource for a fast-moving startup. I’m slightly on the fence about the value of insuring payloads, but we’ll take it case by case.
We were just at Rocket Lab, and we did a tour. We did a series there, and the one topic that keeps getting delayed is Neutron.
Mm-hmm.
How do you think that will or won’t affect the market if and when it starts to be serviceable?
Yeah. Just yesterday, they publicly announced that the first flight will be in 2027. Composite manufacturing is really hard. I know from personal experience that very large composite structures often take a bit longer than you expect, especially when you’re producing them and running the tooling for the first time. I have a lot of faith that Rocket Lab can get Neutron flying regularly. They’ve proven they can scale, with 100 launches of Electron. We’re definitely rooting for them as well.
All right. So now we’re walking around here. You have a motorcycle. We’re going to go look at that.
We can look at it if you want.
Whose is that?
I think it is—I think it’s Bailey’s.
It was a joke, by the way. Does that go into space?
Yeah, maybe. Depends how fast they’re going.
Everyone’s laughing.
I’m permanently terrified that somebody’s going to kill themselves on a motorbike.
Oh, no. Yeah, that is a liability.
Half our team is—yeah. So this is kind of interesting. This is a half-built, zero-gravity sort of offloading rig. We have these runners that run in two axes, and the deployable for Starcloud 2 runs down this. This is one half of it.
Oh, wow.
The other half goes on that side. It’s a 120-foot-long deployable.
Oh, my God.
That small package you saw there extends down the length of this.
What?
And then down the length of that on the other side.
Oh, my God. Okay, so—
The first few segments of that are a radiator segment. That’s the radiator we designed and fabricated in-house. As Adi was mentioning, our design is way more efficient than any other. When we launch in January, it’ll be by far the largest commercial deployable radiator in space. The nearest comparison is the International Space Station. Our radiator has at least 10 times less mass per watt of dissipation than the ISS radiator, and about 500 times less cost per watt of dissipation. The ISS is an easy benchmark to beat on cost, but on mass, it’s quite an achievement.
One of the biggest topics, and I know we talked about this with Adi, is the economics of this. You’re making more economical satellites going out into space, and the vehicle itself. I’d love to break down the economics there, and then further drill down into the difference between terrestrial data centers and orbital ones.
4. Orbital Data Center Economics
The whole thing is actually contingent on the launch cost. If the launch cost gets below around $500 a kilo, then it will be much cheaper to do this in space than terrestrially. A good comparison is if you build a new solar project on Earth to power a data center, because solar is the cheapest form of energy we have terrestrially. The way you can think about it is that solar terrestrially has 3 big costs. The biggest cost is the cost of permitted land, depending on where you are. The second-biggest cost is battery storage and backup power, and the last cost is the solar cells themselves.
In space, we don’t need permitted land, so your biggest cost is gone. We don’t need batteries and backup power because we’re in the sun 24/7. And we need 8 times fewer solar cells, because 1 square meter of solar panel in space produces about 8 times the energy of one in space. So we have all these savings in space. The main additional cost we have is the launch cost, and that’s where the trade-off comes in. When you run those numbers, you come out to about $500 a kilo, where it makes more sense to do it in space than terrestrially.
Of course, right now with Falcon 9, we’re at about $5,000 a kilo. Starship is designed to produce launch costs—internal launch costs for SpaceX—of around $20 to $40 a kilo. So as long as they pass that launch cost through to us at anything less than, say, 10 times their marginal launch cost, or even more, that makes sense for us. Right now, on Falcon 9, they charge 2 or 3 times their marginal launch cost, so that seems reasonable.
Yeah.
Yeah.
Another thing I listened to—I listened to a handful of your interviews—is that one of the other very concerning points is chips and chip failures. How does that work into the model?
For sure. It’s actually a super relevant point. I think it’s even more relevant and pertinent than cooling, which people think is the biggest challenge. We’ve basically solved the cooling challenge. The other part we’re working on is radiation shielding and testing all the chips in different particle accelerators to make sure they can survive the radiation environment in space.
The reason we put so much time into that is that if the chips have a 10% or 20% higher failure rate in space than they do on the ground, then the whole economics of the model fall apart. The chips are by far the most expensive part of what we’re doing. There are a few ways you can work around that. Essentially, we’re not going to have access to the satellites once they’ve launched, so they’ll operate a bit like Starlink satellites do: you need redundancy on critical systems, like pumps, and you over-provision things that fail over time. For example, you’re going to lose a few percent of the solar panels per year, so you slightly over-provision that. On the chips, we have a longer burn-in period on the ground.
Why is that?
I don’t know. It’s basically what they call infant mortality of the chips.
Oh, no.
Yeah.
Oh, my God.
I know, it sounds kind of terrible.
That’s horrible.
Yeah. What you want is to get through the infancy period and find out about all those failures. For Starcloud-2, we had a 3-month burn-in period. It basically means you run the chips hot for 3 months so you can find all those failures early. If the chips can survive 3 months at full power, the chance they’ll have a failure from that point onwards diminishes very dramatically, compared with the chance they’ll have a failure in the first 3 months.
So you have 5 chips up in space right now, including one H100?
Yes.
What are they doing?
We have 3 chips from NVIDIA, 2 from Arm, and we also have some CPUs from AMD on Starcloud-1.
And what they’re doing is, at the moment, we used them to do some interesting workloads, like train the first model in space: nanoGPT from Andrej Karpathy. We ran the first version of Gemini. But now they’re actually doing useful workloads for government customers.
They’re mainly in a test phase, but we’re doing inference on satellite imagery, including synthetic aperture radar data and hyperspectral data. We’ll make an announcement about some of the results, but we’ve got some very interesting results.
The use case is, let’s say, the government has a bunch of military satellites. They collect enormous amounts of mostly imagery, but also other types of data. They don’t want to have to wait for a ground station to downlink all of that data. If they can run inference on it at the edge, we can then send down the insight much quicker.
For example, if they’re monitoring a patch of water, they might want to know if there’s a vessel crossing some boundary. At the moment, they take an enormous amount of imagery and then have to wait at least an hour and a half to get to a ground station. Then they try to downlink as much as they can, but usually that’s not very much.
Once we’re in space, we have a bunch of optical terminals on our satellite. They can send data to us in real time. We can then run the inference workloads on our satellite and downlink the coordinates of a vessel, for example, in minutes rather than hours or days, which is the current way this works.
That’s wild.
Yeah.
Oh my gosh. So are the initial use cases for these satellites and what they’re doing right now the satellite operators and that kind of data? When do you think it’ll reach out into other parts of the economy?
It’s entirely dependent on the launch cost hitting $500 a kilo. The business as it is today runs economically with the current launch price.
Mm.
In order to be competitive with terrestrial data centers on energy and infrastructure cost, we need the launch cost to be $500 a kilo or less. My expectation is that we’ve already seen Starship deploying its first payloads. We’ll see reusability of Starship probably within a year, or at least we’ll see the capture of the upper stage. Reflight might take a little bit longer.
From 2028 onwards, we’re hoping to see SpaceX selling capacity to commercial customers. If they don’t, we’re also considering other options, and we’ve booked on a bunch of other launch vehicles, not just Falcon 9. We announced Stoke and Firefly already, and we’ll announce some others soon.
Congrats.
Yeah.
Well, if the goal on Earth is to get to a one-gigawatt facility, a five-gigawatt facility, what is realistic for space, and what does that look like in mass?
Very large data centers on Earth are usually used for training large models. You need a very highly internally networked cluster to train a frontier model. In space, for a long time, we’re probably not going to be docking things together. Most of the workloads we’re going to be doing in space are inference, usually for things like coding agents, back-office business-processing agents, and ChatGPT queries.
The roadmap is that we launch a bunch of things like this. For the 200-kilowatt one, we can launch 50 per Starship, so you’re talking about close to 10 megawatts of new compute capacity per Starship launch. We’ve filed with the SEC for a constellation of 88,000 of those.
That’s crazy.
Yeah. And that’s just the first 88,000.
Yeah.
So that—
Just the first 88,000.
Yeah, yeah. I mean, we’re not stopping there. That enables us to deploy the first 20 gigawatts.
Mm-hmm.
We think we can fit about 10 terawatts in this dawn-dusk sun-synchronous orbit, so that’s about 20 times the entire U.S. power grid. Once we run out of that, we can go to cislunar, MEO, beyond GEO, and the Lagrange points. Essentially, if you want to scale much more compute, then you really need to do it in space.
So will we have access to sunlight on Earth if we have so many solar panels up?
Yes.
We will?
And the reason is—
’Cause I saw—was it Starcloud-3? It’s four—
Yeah, yeah.
—square kilometers—
Yeah.
—in orbit on his—
Yes. Starcloud-3 will take a while before we can start docking together these large structures, but even when we do, they won’t ever cast a shadow on the Earth. The reason is they’re flying over the poles. So if this is the Sun, and this is Earth, Earth’s spinning on its axis, and they fly over the poles like this. The Earth will never create a shadow on our satellite, and our satellite will never create a shadow on Earth. They never go behind or in front of Earth, basically.
The more you know.
The more you know.
Science. So, Philip, we’re back at the vibe table and we’re about to do some testing. Do you want to explain this?
5. Testing Spacecraft for Launch
Yeah. We need to simulate the launch environment. In order for SpaceX to allow us to launch on a rocket with multiple tenants, they want to make sure that nothing’s going to fly off any individual satellite during launch. They also don’t want any satellite to resonate at a certain frequency and throw the whole stack off.
The point of the vibe table is to put the satellite through every different frequency and amplitude it’s going to experience during launch, and also through this shock test. That way, we make sure nothing flies off and that we don’t have resonance at a certain frequency.
How long do you run it for?
It depends on the test, but the launch environment itself is about seven minutes. There are some tests we would run for a few minutes. Most tests we run for a shorter period. We run random vibe a bit longer, but the shock test we run for a shorter period.
Have you ever—
Sat on a vibe table?
Sat on that?
I have never sat on that, I have to say. People—
Had to ask. I’ve touched them before, but—
Oh, yeah?
We were at Zipline, and we went into their testing room. They had a vibe table and these RC propeller fans to test the drones with. They would fly them in front of it and see if anything would get fragile or break, if there were any fractures. It ended up just being me walking through RC fans.
No way.
Which, I think they go, like, 80 miles an hour.
Wow.
Yeah.
Crazy. Oh, there you go.
Oh.
So this is a very low-frequency test. We’re doing it at a low frequency mainly so that you can see what it looks like when it moves. When it moves faster, you won’t be able to see it moving. They’re moving at least 50 hertz, so 50 times back and forth per second.
This is basically a huge solenoid, a huge magnet that’s pulling this thing in and out. It’s speeding up now. I guess—
Should we walk in?
Yeah, yeah. We’re good.
The hardware we have on the table at the moment is part of our thermal management system, called the accumulator. We have liquid cooling on Starcloud 2 and all of our future vehicles. As the liquid heats up, it wants to expand, so you have to have a component that allows that expansion to take place. Otherwise, you’ll burst your cooling system.
That’s what the accumulator does. It allows for the thermal expansion of the fluid and manages the pressure inside the cooling system. What we have on the table right now is the accumulator for Starcloud 2.
Yep. Yep.
We good? Yeah, go for it.
Yeah. There you go.
Oh my God.
Right?
That’s crazy.
Yeah, so the test it’s doing right now is what’s called a sine sweep. It goes through every different frequency and amplitude that the satellite will experience during launch. It starts at a very low frequency. We’re doing it at a low frequency. Normally we wouldn’t even start this low, but we wanted to be able to see that it’s moving here. As it gets faster and faster, you won’t really be able to see it moving.
What is the worst thing that a vibe table has done?
The worst thing that a vibe table—what?
Has a vibe table, like, ever—has a magnet shot out or anything?
No. The actual vibe tables themselves are pretty safe. The worst thing that can happen with a vibe table is whatever you’re testing can explode. For example, underneath Starcloud 2, we have this xenon tank for the thrusters. If the xenon tank explodes, it’s about half the explosive power of a hand grenade.
Great.
So, yeah.
That’s awesome. That’s really good to know.
Yeah, yeah.
Okay.
That’s why we’re not doing that test today. So—
Yeah.
Normally—
So does everyone leave the building when you test that?
Yeah.
Okay.
Literally everybody leaves the building. We put this entire satellite on this. By the way, you can see it’s in two axes, so the solenoid can flip up and we can test it like this, in different axes. When we fire that test up, everybody has to leave the building.
I doubt the toy manufacturer that has storage across the street has any idea what’s going on in here.
They have no idea.
They have no idea.
But they’re far enough away that we know they’re good.
They’re gonna be okay.
They’re safe. Yeah, yeah, yeah.
Yeah. Like, why is everyone running out?
Yeah. That’s first mode. That’ll be the first mode. Yeah. Oh, okay, okay. No. So what they’re testing for is this first mode. It’s the first frequency at which it’s resonating, and when you can hear it making this clicking sound, that means we’ve hit the first mode of this thing here. What SpaceX requires is that you don’t have a first mode below a certain frequency. I think it’s around 40 hertz.
Okay.
So it cannot resonate at anything less than 40 times a second. The reason for that is, let’s say it would resonate at twice a second, then the whole thing’s gonna go back and forth like this.
Yeah.
And that will move all the other payloads in the stack.
That’s a really good point. So what is the process to get your spacecraft, or any kind of components, tested for space to make sure that it’s viable for Falcon 9 or Starship?
Yeah. As we mentioned, booking is just putting your credit card down. But then that kicks off a whole long process. On the regulatory side, we have to file with the FCC for the orbit and the spectrum we’re gonna use, and we have to have a de-orbit procedure and all these kinds of things. SpaceX has a whole laundry list of requirements, mainly to make sure that your spacecraft isn’t gonna destroy any other spacecraft on the stack. If you book a fully dedicated Falcon 9, the requirements are less stringent than if you’re booking a rideshare, because they need to make sure you’re not gonna move the whole stack or have things flying off and killing other spacecraft.
Damn.
Yeah.
So how do they actually verify your approvals or your checklists?
We have to do vibe testing, and then we send them the vibe reports—
Okay.
—after we’ve done the vibe testing.
Has anybody ever lied?
I don’t think so. It would be quite hard to lie. Reproducing these vibe reports would be quite a challenge.
I don’t know why you would do that unless you’re an op.
Unless you’re an op.
Unless you’re an op.
There you go.
Geez. I think we just uncovered something. All right, so we’re walking through to the laser lab.
Yeah.
How many people work at Starcloud?
Right now it’s 30 people at Starcloud.
Wow. That’s it?
That’s it. Yeah, yeah.
Oh my God.
Only 2 are non-engineers. We’ll probably have 50 by the end of the year, maybe 100 by the end of next year, but we very deliberately keep the team extremely lean. As I say, we have the most kickass engineering team on planet Earth. This will become a full laser lab, so we’ll use this for cutting and welding aluminum and things. We have a little demo for you.
Make things go zap.
Yeah, yeah.
Hi.
This is Camilo, who’s—
Do we wanna give you a mic and you can talk through it? Or do you wanna talk?
I can just do the, you know—
Yeah, yeah. You can have a little word.
Demo.
So this is our CNC xTool laser cutter. We use this to do quick prototyping on different types of materials. It gives us a lot of flexibility to get things done as fast as possible. Right now I just have a little demo with some 30-thou stainless. Sorry for the compressor. It’s a little loud. You’ll see how fast this can do very complex cuts, and it’s extremely accurate. So I’ll—
I’m nervous.
—run a quick demo here.
I always get nervous for these.
If you wanna get the camera in there, you can probably—
Here, you can go in there, Brett.
—you can get a shot, yeah.
Yeah, I’ll hand this back to you.
Oh, thanks.
All right. Ready? Go.
Fire. Yeah, I can’t see. There’s too much reflection.
You’ll see it here, yeah.
And then it’s done.
Oh my God. It’s so cute.
Yeah.
It’s a little present for you.
That’s awesome.
Little samples. I got some here that are ready. Stuff like this helps a lot to move quickly, and it gives us a lot of flexibility. This is just a start. We’re gonna get much larger professional machines.
Wow. Is this your favorite tool?
One of them.
One of them?
The next one’s gonna be my favorite.
What is it?
It’s gonna be an AMADA fiber laser cutter.
Mm-hmm.
So that we can fit material that’s up to, I believe, 5 by 10 feet. It’s a professional-grade laser cutter, so it’s gonna be significantly more capable than this one.
Damn. That’s fun.
Yeah.
Thank you.
Of course.
And as I understand, you yourself—
Oh, really? Okay.
Yeah.
I get to keep it.
I can take it if you want me to put it in my pocket. I’ll take—
Here, I’ll give it to Brett. All right, so we walked through—
The fab shop, the assembly area—
—the fab shop, assembly area, laser.
We’ve got—we’ll be assembling a clean room, which I don’t think we showed you, but that’s on the other side. As I understand, you yourself are actually a proficient welder. Is that true?
I made a very bold claim, but I did learn how to weld in college, and—
I think we should put this to the test.
I’m down to try.
I think your viewers would find this very exciting.
I am down to try. I was called a prodigy, so we’ll see if that’s still the case.
We’ll find out. We’ll find out. We have some welding materials back here. Shall we get you in a—
Sure. Do I have to change or anything?
You might. Yeah, you need to change.
Okay. Then I’ll change. All right. We’ve now advanced. We’re now with Miguel, the engineering operations lead here at Starcloud. What are you gonna show us today, and what’s your background?
Yeah, so we’re gonna do a little bit of traditional welding. We do a lot of laser welding here at Starcloud, but we can also fabricate pretty much anything out of steel, aluminum, titanium, and stainless steel. We’re set up for MIG welding right now, so we’re gonna have you run some beads and make some sparks.
Before here, I was the welder for Helion Energy, where we built fusion generators. I ran their fabrication shop for over a year. Before that, I’d been working at a lot of different fab shops, doing architectural and structural welding and fabrication. Before that, I spent a decade in the US Army, deploying overseas.
Oh my God. How did you get into welding?
I wanted to do something very different. I wanted to do something that was neutral, in the sense that I was contributing to society in a different way. So I decided to get into fabrication and welding after listening to Mike Rowe, the guy from Dirty Jobs.
He said—
Yeah.
The United States really needs welders, and it needs people to build stuff with their hands. I was like, “Well, I can build stuff with my hands.” So that led me to this career path, which has been very rewarding.
It’s very satisfying.
Yeah, it’s a lot of fun.
It’s a little addicting.
Yeah, it’s a lot of fun. A lot of fun.
That sounds like you.
I know.
You posted recently on X. Somebody was like, “America needs more welders,” and you were posting stuff about how good you are at welding.
Yeah.
So now we’re gonna find out.
Now we’re gonna find out. I used to be very good. For background, this isn’t at all in this world, but I went to college for studio art. I grew up as an artist, and I was also very entrepreneurial, so I ended up going that path. But I went to college for art, and the first year of studio art, you do every discipline, every practice. In one of them, we did metalwork, woodwork, glassblowing—
Nice.
—all of the fun flame stuff. It was so fun.
It’s a lot of fun, yeah.
And then I thought, well, making this for fun isn’t really a job, you know? You might as well be doing that at trade school. I don’t really understand that, but it was so much fun.
Nice.
And I was called a prodigy, so.
We will see.
We’re about to find out.
So what we’re gonna do is give you a little bit of PPE so the sparks can’t burn you. Just put this jacket on, and we’re gonna give you some gloves.
Should I tie my hair back?
If you keep your hair—
I don’t know how to tie my hair back, but—
—in the jacket, you might be fine.
Oh, I’ll just keep it in, yeah.
Yeah, as long as it’s behind you, you’re fine. And make sure you button at least a couple of the buttons.
I’ll button all of them.
As long as you have all this on and closed-toe shoes, you’ll be just fine. If you feel anything hot, just stop and step back, but you should be perfectly fine.
Okay. Yeah, I remember we weren’t allowed to wear any polyester.
Mm-hmm. Yeah, perfect. And then you can do your sleeves as well.
You haven’t done any of this? Why? Because you—
Busy guy. He’s never here.
Oh my God.
Traveling the world.
That’s true.
Oh, yeah.
He’s always on stage somewhere. Why aren’t you in the shop?
So we’ll have you put this on first. The way you put it on, it goes over your head, and you click this. You can leave this up, right? Then when it’s time to weld, you can put it down. You can see through it just fine, and the second that you start welding, it’ll auto-darken.
Okay.
So you don’t have to worry about doing anything. All you have to do is put it on, crank this down—
Okay. Let me—
—and you’ll be ready to go.
—try to do this part.
Do this one first. Yep. I’m gonna put on some PPE, too.
Do you have any kind of gloves? I’m gonna be quite far back, but—
You can use the same one right there.
This one?
Yep, you can put that on.
Okay, there you go.
Who is she?
Very nice.
That one, my hand skipped a little bit.
Yep, that’s fine. Still—
Beautiful.
—counts. We can do one more if you want.
Okay, one more.
All right, last one. This is one that counts.
All right.
Stay nice and tight.
A little bit closer.
Closer. Nice. All right, you’re hired.
Not bad.
Bang.
Good job.
Let’s go.
Nice.
America needs welders.
Woo!
Nice job.
Cool. Thank you, Miguel.
You’re very welcome.
This was so much fun.
Good job.
That’s my real reaction. On a scale of 1 to 10, what are we talking about here?
Thanks, Miguel. Appreciate it.
You’re a solid five.
Wow.
I’ve seen guys who’ve been making welds for 10 years and don’t do a job like that.
Damn.
Damn. There you go.
That’s pretty good.
You guys pay attention and take good instructions.
That’s right, we do.
Damn.
That’s great. Cool, thank you.
They just lifted the satellite. I don’t know if you want to—maybe it’s too late to get footage of that, but if you want, we just—
Yeah, let’s go over there. So we just did the welding, and—
You crushed it.
—I’m a five out of 10.
Yeah.
But—
He didn’t say out of 10. It could be out of five.
He did say out of 10.
Did he?
Yes.
I thought he said, “You’re a solid five.”
No, he said, “You’re a solid five out of 10.”
Did he?
Yeah.
Oh, okay, okay, okay. Yeah.
It’s like, damn. That’s pretty good. Okay, I was wondering, how does the flow change in terms of how many people are working on the floor versus up in the offices through the different stages and processes of building out Starcloud 2?
Yeah. Most people you can see are engineers, and we have all the engineers make sure that they’re hands-on with the satellite as well. Half the time these people will be upstairs designing stuff on CAD; the other half of the time they’re down here fiddling with wires and testing stuff. We also have a technician team. That was Miguel. We’ve got four technicians—
These are rough numbers.
—some more coming on soon.
But this will be—
Maybe five now. I don’t wanna get it wrong. Their job is welding and fabrication, actually producing the satellite. Miguel’s team produced and fabricated the whole compute module here, an aluminum box, in-house.
And I am contributing to that.
You are contributing a huge amount.
Whenever you need me, let me know. I can come on.
For your listeners, it is official: you have now done more welding at Starcloud than I have.
Wow. You gotta really visit Miguel. I think he’s waiting for you to put in some work.
Yeah, yeah, he will.
You’re not doing enough.
Yeah.
Bailey, tell us a little bit about yourself. I know we’re in front of Starcloud 2, and something just happened, and you guys are testing things, but we’d love to learn more about you and how you contributed.
You bet. I’m the engineering manager here at Starcloud. What we’re doing right now is a leak check of the prop system. We have a fully fueled electric propulsion system—an ion thruster—and we’re testing it after vibe to make sure that we didn’t introduce any cracks or other leaks during the vibe test.
How did you come here? How did you get to Starcloud?
It was a bit of a long journey. I started my career in aerospace at SpaceX. I spent about 4 years there. Then I did a couple of other side quests in biotech and consumer product design. Those didn’t really stick. I moved on to fusion energy, which was really cool, but at the end of that 3-year stint, I realized I’m a space nerd; I need to be working on space to be happy. Fortunately, living in Seattle, there are a ton of cool companies in the area. But Starcloud really stuck out. Back then they had a different name, and I got in contact with the founders super early on. The rest is history.
What’s your favorite thing about Philip?
You can’t do that. You can’t do that.
I know, I know. What do you hate about him?
What’s your least favorite thing about him?
Yeah, what do you hate about him? What’s your least favorite— We don’t say hate.
Got it.
What’s your biggest challenge with Philip over here? Does he not give you enough attention? What’s going on?
No, definitely not that.
Let’s work this out on camera.
It would be… It’s a two-sided coin, or whatever, but the office is very busy with tours and things like that all the time.
We gotta go.
Yeah, there you go. There you go.
The interview’s over.
There you go.
The interview’s over.
Yeah.
He’s like, “I have to work.”
It’s good and bad.
Stop interviewing me.
That was the politest way to tell you to fuck off.
My God. Damn, Bailey, give me the mic.
Yeah.
Oh, that’s funny. Okay.
Appreciate it.
So what are you most looking forward to? What is your favorite part of the process?
I think the next thing I’m looking forward to most is obviously the launch, for sure.
I’m going to be tremendously proud of this. A lot of blood, sweat, and literal tears went into this thing.
Mm.
As far as the whole design process goes, I’d say testing is my favorite part. That’s where the rubber really hits the road. The recent vibration test was a really big one, and the cooling system was another. When we were in Colorado not long ago, that was our first full-up, integrated test of the whole cooling system, which was my baby from day one. It was really cool to see that up and running, too.
Everybody always asks what the hardest thing about Starcloud is, and it’s the thermal system, these radiators in particular. Bailey was responsible for designing and fabricating the radiators, and Bailey is also responsible for the satellite being here. In the end, we only had 3 weeks to do the final assembly of the satellite, and Bailey went down and basically built that bus from scratch.
Damn.
It was fun.
Where’d you build it?
In Colorado, in our bus provider’s office.
Oh my God. All right, so we’re just leaving—
So, yeah.
—the main facility. We’re now going over here.
Yeah. We’re going to—
You’re buying up the entire—
6. Scaling Starcloud’s Manufacturing
We’re building a bit of a campus here. This is definitely the place where you need to be to build AI satellites, basically. SI satellites.
Yeah.
We’ve got this middle section, which will come online in November, and then we have access to the third one in February next year.
Mm-hmm.
This enables us to scale manufacturing for Starcloud 3. One of the reasons we need so much space is that Starcloud 3 has 100-meter—well, 400-foot—deployables that we need to test the mechanisms on. But also, to manufacture something like 88,000, we’re going to need a way bigger facility.
I just think Brad’s going to trip. I’m so sorry.
It’s fine. It’s fine.
To manufacture 88,000, okay?
Yeah, yeah, yeah.
Yeah.
To manufacture 88,000, we’re going to need a whole bunch more space.
So how many do you think you’ll be able to produce a day?
We’ll be up to 100 a week from what we’ve got here, so 10 a day, or more than 10 a day after a year. This is one of the 3 spaces, and we’ve got about 100,000 square feet that we can fit out for manufacturing. We’re going to remove this wall, and we’ll remove that wall—
This is huge. Oh my God.
Yeah. It’s going to be quite a sight to behold when this is up and running.
You should throw a Halloween party in here with these—
Oh, yeah. We’re going to do an opening party sometime soon, so you’re very much welcome.
Thank you.
But I know you’re on the road all the time.
You never know where I land.
Who knows? Maybe Woodinville in 3 weeks.
I know. Okay.
So in here, it’s going to look much different than the previous facility?
It would look pretty similar, actually, to the previous facility.
It looks the same?
Except with a manufacturing line, like a proper—
Right.
—in the end state, we want it to be like an auto manufacturing line. If you see a Tesla factory, you see cars going down a line. It’ll be exactly like that, except with Starcloud 3 chassis going down the line, with various things being added.
How does Starcloud 3 get to Starship?
We load them on the back of a truck and then drive them down the motorway.
That’s it?
That’s it, yeah.
Oh.
It’s very similar to the way they do it at the full Starlink facility, just down the road from here. They literally load all those on the back of a truck and drive them across the country to Florida.
Damn.
Yeah.
Okay. Wow. All right, so one thing we haven’t talked about yet is Starcloud has raised $450 million in total. What are you using the money for? How is the capital being utilized?
Yeah. Right now, launch is very constrained, especially from 2028 and 2029 onwards. A portion of that cash is going to reserving launch slots. A bunch of it is going to fitting out these places, and then hiring and materials and everything else. We’re making about a million dollars a month in interest right now.
Oh my God.
But soon—
Do you use Brex?
We do use Brex, actually.
—we’ll be spending.
7. Securing Satellites in Orbit
So one major topic of discussion, as global competition heats up and satellites become the greatest asset for national security, is what happens if someone tries to shoot one down, or our adversaries get in contact with it, or maybe there’s a listening device. Apparently, there was one at OpenAI. How do you think through the national security concerns when it comes to chips and data centers in space?
Yeah, it’s a great point, and one we think a lot about. It’s actually a lot harder to shoot down a satellite than it is to, for example, blow up a data center in Virginia, which is surprising. The reason is that satellites are traveling at about 17,000 miles an hour, so targeting and having some kind of kinetic impact on a satellite at that speed is very difficult.
We have a pretty good case study from the Starlink satellites. Starlink is being used in Ukraine, and also Starshield, and the Russians aren’t particularly happy about that. In terms of both cyber and the physical security of satellites, we can see the tactics people are already using on Starlink. We’re using the same kind of cryptography to make sure the satellites are secure from that perspective.
Anything that approaches a satellite would be very visible. Anything bigger than a one-inch cube could be tracked in space, so we would see if somebody was tampering with our satellites, and it would be considered an act of war. There’s the Outer Space Treaty, which governs what you can and can’t do, and that would be considered an act of war on the U.S. That’s the purpose—the primary purpose—of Space Force. One of the key purposes is to develop interceptors and things to dissuade offensive action.
As we just finished walking through the HQ, seeing the tracking of Starcloud 1, the development of Starcloud 2, and some of the early innings of Starcloud 3, what are you most looking forward to in the next 12 months?
Definitely the thing I’m most looking forward to is the launch of Starcloud 2. That’s going to happen in January, and it’s going to be a huge proof point, with the radiators and everything else we’ve designed in there.
Exciting. Well, thank you so much, Philip.
Thanks so much for coming.