SpaceX, AI Compute, and the New Rules of Space Capital

Zach Martin

Here's a number to hold onto before we do anything else. In a single quarter, SpaceX sold AI compute to Anthropic and Google at roughly twenty six billion dollars annualized. That comes from Space Capital's Q2 2026 opening letter, and the letter's own framing is that this more than doubled the revenue SpaceX took twenty three years to build. Sit with that for a second. A company that spent more than two decades building a launch and satellite business added more than that entire business in one quarter, and it did it by selling something that isn't a rocket ride at all. I'm Zach Martin, and this is Space Capital on KMKT IR Hub Radio. Real assets, real ideas, a brighter tomorrow. Today I want to do something a little less exciting than a victory lap and a lot more useful. I want to walk through what happens after the confetti, when a sector that just got its biggest public market moment has to run on unit economics, procurement rules, and physics.

Zach Martin

If you caught the last episode, Launch Plus and the Real Economics of Orbit, we covered the introductory pieces. The initial Q2 2026 Space IQ scoreboard, the SPCX debut on Nasdaq, and the formal definition of Launch+. I also left you with a warning, and it's the warning I want to pick up today. Public equity is a gift and a stress test at the same time. It hands you liquidity and a currency, and it also hands you a price that changes every second the market is open. So this episode moves past the definitions. It's about autumn 2026 and what the capital is actually running into. Think of it like a training camp. The tryouts are over, everybody made the roster, and now the real work is conditioning, because the season doesn't care how impressive the tryouts were.

Zach Martin

Let's set the facts down plainly. SpaceX debuted on Nasdaq under the ticker SPCX on June 12, 2026. It closed its first day up 19 percent, at a market cap north of two trillion dollars. Space Capital's letter calls it the largest IPO in history, and it makes a point I think is worth repeating. The company's S one filing doesn't describe a rocket company. It describes, and I'm quoting the filing as the letter quotes it, 'the only company building the integrated hardware and software infrastructure of the future across space, connectivity, and AI.' Space Capital's read is that the stock listed as an AI company, not a rocket company. That's their interpretation, and they're an investor with a long history in SpaceX, so keep that in mind. But the filing language is the filing language, and it tells you how the company wants to be priced.

Zach Martin

Now look at the sequence of moves, because the sequence is the story. According to the letter, SpaceX merged with xAI ahead of the IPO. It bought an option to acquire Cursor, and then exercised that option in a sixty billion dollar all stock deal days after the debut. It sold that AI compute to Anthropic and Google. It plans to make AI chips with Terafab to power its orbital data centers. And it launched a twenty five billion dollar notes offering, which the letter describes as following big tech into the debt markets to fund its AI buildout. Before this year, the letter says, SpaceX had made four acquisitions in its entire history, none larger than 525 million dollars. In Q2 alone it added two more, including that sixty billion dollar deal priced in its own record stock. Go from 525 million as your largest deal to sixty billion, and the lesson isn't that the company got ambitious. It's that the company got a new kind of money.

Zach Martin

Here's the mechanism, because this is where the earlier warning about volatility becomes concrete. When you pay for something in your own stock, you're not spending cash you earned. You're handing the seller a claim on your future, valued at today's share price. That works beautifully when the price is high and rising, because the currency feels free. Generally speaking, in an all stock deal the seller is taking your paper, so what they actually receive moves with your share price between the day you sign and the day they can sell. I don't have the specific terms of the Cursor deal, whether there's any protection built in for price swings, so treat that as the general mechanism and not a claim about this contract. The point is simple. A record valuation makes stock the cheapest money you'll ever spend, and a falling valuation makes the same deal quietly more expensive in real terms. Same shares, different weight.

Zach Martin

And the market didn't wait long to test the idea. Space Capital's own roundup of coverage lists a MarketWatch headline that reads, 'SpaceX ushered in a new era for investing in space. Why the stock is now cratering.' I want to be careful here. I'm relying on the headline as Space Capital lists it, and I don't have a figure for how far the stock fell, so I won't invent one. What I can say is that within weeks of a record debut, a major outlet was already writing about a decline, and that is precisely the volatility I flagged last time. It doesn't mean the thesis is wrong. A stock can swing hard and the underlying business can still be compounding. But it does tell you that the sixty billion dollar Cursor deal and the eight billion dollar Rocket Lab Iridium deal are not finished the day they're announced. They live inside a share price, and a share price is a moving target.

Zach Martin

Speaking of Rocket Lab, let's put the second big currency move next to the first. The letter reports Rocket Lab announced an eight billion dollar acquisition of Iridium, at 54 dollars per share in cash and stock, expected to close in mid 2027. In the same quarter, the letter says, Amazon paid 11.6 billion dollars for Globalstar. Notice the timing on the Rocket Lab deal. Mid 2027 is many months away, and the letter says Rocket Lab is buying Iridium partly with its own stock. Anything that's partly stock and closes far in the future is a deal whose real price will be discovered along the way. That's my interpretation, not a quote from anyone. The mix between cash and stock matters a great deal, and I don't have that split, so I'm not going to guess. What I'll say is that these are balance sheet engineering moves at a scale this sector has never seen, and balance sheet engineering is only as strong as the market's mood on closing day.

Zach Martin

Zoom out to the flow of private capital, because the numbers around this are staggering and they deserve to be stated carefully. Space Capital reports 55.3 billion dollars of private investment in the space economy in 2025, anchored by a 17.0 billion dollar fourth quarter across 135 rounds. Then Q1 2026 set a record, 36.0 billion dollars across 148 companies in a single quarter. Q2 followed with 31.6 billion dollars across 129 companies. Put those together and the first half of 2026 comes to more than 67.6 billion dollars, which is more than all of 2025 in six months. Cumulative investment since 2009 now exceeds 488 billion dollars. One detail I don't want to skip, because it's the kind of thing that gets lost. Q2 was lower than Q1 in both dollars and number of companies. Both were enormous by historical standards, but the pace didn't climb in a straight line. SpaceNews ran a headline, as Space Capital lists it, that read 'Satellite investment sets annual record halfway through 2026.' Records, plural, and a bumpy path to them.

Zach Martin

Then there's the part the sector waited a decade to hear. For years the knock on space was exits. Space Capital's letter says that argument ended in Q2. The SpaceX IPO, they write, is the largest liquidity event in venture capital's history, surpassing the full year exit value of 2021, the previous high water mark, and generating more exit value than every venture backed IPO of the past decade combined. HawkEye 360 listed. Blue Origin acknowledged for the first time that it will need outside capital to scale, and didn't rule out going public. Space Capital itself announced its fourth fund in March, crossing one billion dollars in assets under management, and noted that only 17 percent of venture firms ever reach a Fund IV. In May it partnered with Nasdaq, with quarterly briefings filmed at the MarketSite. All of that adds up to institutional attention arriving at scale. The firm's founder, Chad Anderson, has called this a Netscape moment. I'd gently add that a Netscape moment is a statement about attention and access. It's the moment a category becomes investable. It isn't a promise that every company in the category gets paid.

Zach Martin

That brings me to the word that anchors this episode. In the Fund IV announcement in March, Anderson wrote, and this is the exact line, 'The skepticism phase of the space economy is over. We have now entered the industrial phase.' I like that sentence, and I want to push on it a little. Ending skepticism doesn't end scrutiny. It changes the kind of scrutiny. When people doubted the category, the question was whether anyone would ever pay. In the industrial phase, the question is whether you can build the thing, at a cost that works, on a schedule a customer will accept, using money that's still there when the market has a bad week. Those are factory floor questions. So here's how the next hour is going to run. We'll look at five filters that decide which of these capital flows survive. First, what you own once you reach orbit. Second, whether you own a factory or a design. Third, whether the government is a customer or a bystander. Fourth, whether the physics and the price tag agree. And fifth, whether a mission can become infrastructure. Let's start with the plus.

Zach Martin

Let me start with a statistic that reorders how you think about rockets. Space Capital reports that pure launch has attracted 18.3 billion dollars of investment since 2009. That sounds like a lot until you hear the rest of the sentence. It's roughly four percent of all capital invested in the space economy. Satellites took 81 percent. If you had to explain where smart money has actually gone, it's a one line answer. It went to what sits in orbit and earns money every month, not to the vehicle that gets it there. A rocket is a toll bridge at the edge of town. Useful, sometimes essential, but the real estate that appreciates is on the other side. Space Capital has said for years that launch alone doesn't make a great business, and the Q2 data, in their words, delivered the proof of the corollary. Access to orbit is a strategic advantage that lets launch companies expand into more lucrative markets.

Zach Martin

That's the foundation for the category Space Capital introduced in the Q2 report, Launch+. Since we defined it last time, I'll keep this quick and move to what's new. Launch+ is an industry combination for companies operating across launch plus at least one other space industry. Space Capital describes two forms. Integrated operators combine launch with other space businesses. The write up puts SpaceX as the fullest expression, with launch and satellites, plus Dragon crew and cargo service to the ISS, a Starlink based space situational awareness system called Stargaze, and a cargo return capsule called Starfall for in space manufacturing, with lunar ahead. Blue Origin combines launch, satellites, and lunar. Rocket Lab combines launch, satellite manufacturing, and satellite communications with the pending Iridium deal. Firefly pairs launch with lunar landers and orbital vehicles. The second form is digital engineering companies, like Prometheus, Nominal, Sift, and Revel, which supply the design, testing, and operations software that launch vehicles and satellites are built on.

Zach Martin

Here's the part of the data that has capital allocators sitting up. Space Capital says Launch+ led all Infrastructure investment at 14.6 billion dollars year to date, and drove 71 percent of the quarter's top Infrastructure deals, anchoring what they call the largest Infrastructure quarter on record. Read that as a statement about where the big checks are going. They aren't going to rocket makers that only make rockets. They're going to companies that use access to orbit as the first move in a longer game. I do want to add a cautionary note about the source, because that's part of doing this honestly. Space Capital created the Launch+ category, they invest in many of the companies we're discussing, and a framework you invented will naturally show you what it was built to show. The data is useful and the logic is coherent. But it's one firm's taxonomy, and it's worth holding it the way you'd hold a coach's scouting report, as a strong opinion with a point of view.

Zach Martin

Now the best illustration of the plus is Rocket Lab's quarter, because it reads like a case study in turning a launch business into something bigger. The letter says Rocket Lab joined the Nasdaq 100. It broke its own launch turnaround record by more than ten hours on a U.S. Space Force mission. It closed its acquisition of Mynaric, adding laser optical terminals and its first European entity. Then came the headline, the eight billion dollar Iridium acquisition. Notice the logic of that progression. First, prove you can launch quickly and reliably for the customer who cares most about speed. Then buy the optical communication hardware that links satellites. Then buy a network. Each step takes the company further from selling a ride and closer to owning something that earns money after the ride is over. If you wanted to design a path from launch provider to communications company, you could hardly draw it more clearly.

Zach Martin

So why Iridium? Space Capital's answer is spectrum and services, not sentiment. And spectrum is the single most important idea in this chapter. Spectrum, the letter says, is the scarcest asset in the space economy, because no one can make more of it. Iridium's L band licenses are globally harmonized, meaning they work everywhere on Earth rather than country by country. Add roughly 2.5 million subscribers and decades of government and safety of life contracts, and you get an installed base in the markets the letter calls heating up fastest: direct to device, positioning, navigation and timing, and aviation. Here's the analogy I'd use. Building a satellite constellation is like building a hotel. Spectrum is the deed to the only beachfront lot in the world, and you can't pour more beach. Anyone can build a hotel somewhere else, but the lot is the lot. That's why Amazon paying 11.6 billion dollars for Globalstar in the same quarter isn't a coincidence. Space Capital reads it as the same logic, and I find that reading persuasive. The value of reaching orbit is what you own once you're there.

Zach Martin

But here's where I want to slow down and apply the industrial phase lens, because owning scarce spectrum and monetizing it aren't the same thing. Space Capital published an analysis back in September 2024 of direct to cell and direct to device service, and it's a sobering counterweight to the excitement. It explains that the rules for spectrum are set through the International Telecommunication Union, along with regional regulators like the FCC. The ITU meets every four years, it takes another four years to put an item on the agenda, another four for voting, and two more for countries to implement. Add it up and you get 14 years. As of that 2024 analysis, they said we were four years into the cycle, and experts indicated it would be another ten years before global direct to cell spectrum rules are in place. I can't tell you precisely where that clock stands today, but the structure of it hasn't changed. The scarce asset is scarce partly because the process of allocating it moves at the speed of treaties.

Zach Martin

And the same analysis offers the size of the prize that makes the patience worth it. Space Capital's projection is that the global market potential for direct to device technologies could reach up to 168 billion dollars over the ten year adoption timeline, with a projected market penetration of five to fifteen percent. It forecasts that satellite based direct to device users could grow to 386 million within six years of the 2024 piece, generating cumulative revenues of about 66.8 billion dollars for the satellite sector over the decade. Please read those as forecasts from an interested party, made in 2024, not as reported results. They're an argument for why the money is chasing spectrum, not proof of how much will be earned. The honest summary is this. The prize is large, the rulebook is slow, and a company that has the licenses already, like Iridium, doesn't have to wait for the rulebook to catch up with the technology. That's a real asset and it's priced like one.

Zach Martin

So now connect this back to the stock currency we talked about. Both Rocket Lab and SpaceX are using richly valued equity to buy things that don't depend on the equity staying rich. Iridium's subscribers pay whether or not a rocket stock is having a good month. Government safety of life contracts continue in either market. That's actually a defensible reason to use stock to buy cash generating assets. You're converting a volatile paper price into durable revenue. The risk is on the other side of the trade. If the acquirer's stock falls between signing and closing, the target's shareholders, or the acquirer's own, may feel the deal got more expensive than it looked. Again, that's general mechanics rather than a statement about the terms. And there's a second risk that doesn't show up in a ticker. Integration. Merging a launch operator's engineering culture with a mature communications network, or with an optical terminal maker in Europe, is a lot of operational work. The deal announcement is the easy day. The hard days are the eighteen months after.

Zach Martin

What does this mean if you're trying to think clearly about where capital goes next? Space Capital's prediction is blunt. Standalone launch will atrophy. They expect more high value M&A, more teaming agreements, and more launch companies expanding into services as they reach orbit. They also describe a hypothetical that I think is clarifying. If an Applications company like Anduril acquired or built its own launch capability to vertically integrate its defense platform, it would become a Launch+ company at the Applications layer. In other words, the arrow doesn't only point from rockets outward. Customers may reach backward toward launch. Either way, the common thread is the same. The business is no longer a single stage. It's the full stack, and the full stack has to be financed, integrated, and defended against a falling share price. That's the first filter. Ask what the company owns once it gets there, and then ask how it's paying for it. Next up, the question of who builds the hardware, and whether owning the factory is a strength or a trap.

Zach Martin

Let me tell you about a company with thirty employees and no factory that wants to put whole constellations in the hands of customers who could never build one. It's called Eclipse Space, and it came out of stealth on stage at the Space Capital Summit at Nasdaq MarketSite. Space Capital led its seed round, so the firm has skin in this one, and I'll flag that again. Space Capital published the conversation on August 4, 2026. On stage were Derek Huerta, Eclipse's CEO, and Tom Whayne, a Space Capital operating partner who was formerly CFO of OneWeb and chief strategy officer at Maxar. The session took place two weeks after SpaceX delivered the largest IPO in history, which is not an accident of scheduling, because the whole conversation is a response to what SpaceX built. And their argument is a provocative one. Vertical integration, the thing the industry copied from SpaceX, was never the source of the advantage.

Zach Martin

Huerta isn't a bystander making that claim. He spent seven years at SpaceX as a satellite payload engineering manager, leading the team behind the phased arrays that link Starlink satellites to the ground. According to Space Capital's write up, roughly half of Eclipse's thirty employees came from Starlink, thirteen of them from the years when the program was still called Satellite Development, and they work out of Redmond, Washington, where SpaceX builds Starlink satellites. Huerta's stated motivation is that the technology is, in his words, 'too important to leave in the hands of just these trillion dollar companies.' Whayne set up the scale problem with a striking fact. In fifty years of satellite communications, only two companies have organically reached a billion dollars of connectivity revenue. Inmarsat needed thirty years. SpaceX needed a handful. So the question for anyone who wants to compete is obvious. What exactly did Starlink get right?

Zach Martin

Huerta named two things, and neither was owning every factory. The first was treating the user terminal, the spacecraft, and the launch vehicle as one system, so engineers could make trades across the entire chain rather than optimizing inside one slice. If a change to the satellite makes the dish cheaper, you want the freedom to make that change. The second was importing mass manufacturing discipline from industries that had already solved it, designed in from the start, not retrofitted after the fact. Think about how a home cook and a restaurant chain approach a recipe. The home cook perfects the dish. The chain redesigns the dish so a trained but ordinary crew can produce it ten thousand times the same way. Huerta's point is that the second skill is what mattered, and that skill lives in the design. In his words, 'You don't actually need to own a factory.' That's a bold claim, and it's worth noticing that he's making it as an engineer who has seen the factory from inside.

Zach Martin

Here's the operating model he proposes. Eclipse runs the way Apple runs with Foxconn and NVIDIA runs with TSMC. In the Space Capital write up, the company owns the design, architecture, integration, and on orbit operations, and it engineers the spacecraft so that building it doesn't demand highly skilled labor. Someone else owns the plant. That sentence is the whole strategy, and it's the same sentence from the exact source evidence: own the design and the supply chain, let someone else own the plant. In the semiconductor world, that division of labor created whole industries. Companies that couldn't afford to build a chip fabrication plant could still design chips and win. Eclipse is betting the same division of labor can work for satellites. And the argument doesn't just save money. It changes who gets to be a customer.

Zach Martin

Now let's talk about why the factory becomes an anchor, because this is the part I'd encourage you to carry with you into any conversation about space industrials. Here's the logic from the write up. A factory obligates you to fill it. SpaceX and Amazon can fill their factories because they generate their own demand. SpaceX needs Starlink satellites for Starlink, and Amazon needs satellites for its own constellation. Everyone else builds a plant first and then goes hunting for customers. Idle capacity raises the cost of every unit that does get built, because the fixed costs get spread across fewer satellites. Higher unit cost means a higher price, and a higher price drives away the very customers who would have filled the plant. It's a spiral. Think of a gym that signed a long lease on a huge building. The rent is the same whether the machines are busy or empty, so the less it's used, the more each workout costs, and the more members leave. The moat, in Space Capital's phrasing, becomes an anchor.

Zach Martin

Underneath the argument is one metric that I want you to remember. Dollars per throughput delivered in orbit. That's the number Eclipse says it's driving down. And the claim, as the write up describes it, is that if you push it far enough, serving a rural user from orbit costs less than serving that user with a tower. At that crossover point, the billions telecom companies spend every year on radios and towers could shift upward into space. Eclipse says it starts with direct to device service because that's the hardest version of the problem, and the stack that solves it carries over to broadband and to orbital compute. I'd mark that as a thesis, not a measurement. Nobody has published the crossover figure, and the first customer hardware only ships later this year, with a demonstration spacecraft following in 2027. The company is a design on paper and a set of relationships today. What I find compelling is the shape of the argument, not a proof.

Zach Martin

And there's a geopolitical layer that I think explains why capital might care. Huerta picked up a point from NASA's Deputy Administrator earlier that same day, that space inverts the usual order. Commercial operators built the infrastructure, and governments became the tenants. Countries are discovering they depend on capability they neither own nor control. Huerta described two existing options. China's answer, exported along the Belt and Road, is to build and finance your constellation and keep the strings. The American answer has been leased capacity on someone else's network. His third option is infrastructure built in country on what he described as an ITAR free platform, which the customer owns and operates. ITAR refers to U.S. export controls, and I'd note that an ITAR free claim is the kind of thing a buyer's lawyers will want to examine closely. But the demand it points to is real. A government or a telecom operator that wants a constellation it actually owns has, until now, had very few places to go.

Zach Martin

Let me give you my own caution here, and I'll label it as interpretation rather than anything from the stage. Going fabless shifts the risk rather than deleting it. You avoid the capital cost and the idle plant problem, but you inherit the supply chain. If your contract manufacturer has a bad quarter, a capacity crunch, or a quality problem, your launch date moves and you don't control the fix. The story Space Capital tells at the end of the write up is telling on that point. Before Eclipse had a logo, a website, or an office, Huerta had someone on a plane building the supply chain, and he had met heads of state. An advisor told him he was doing it backwards. But the supply chain, Huerta concluded, was the hard part. The logo could wait. I love that, because it admits that the supposedly asset light model is really a bet on relationships and execution in a part of the business that's not glamorous.

Zach Martin

Tie it back to the capital flows. For years the question was who could afford to build a constellation. A vertically integrated giant can, because its internal demand fills its plants. If the fabless argument holds up, a much broader set of buyers can afford one, and the capital requirement moves from heavy fixed assets to design talent, supply contracts, and customer commitments. That's the second filter. Ask whether a company's fixed costs are waiting for demand to arrive, or whether demand arrives first and builds the fixed costs. It's a coaching principle as much as a financial one. You don't buy the biggest equipment before you've built the habit of showing up. Space Capital's overall pattern, as it describes it, is that infrastructure gets built, becomes invisible, and then becomes the foundation of the next commercial layer. GPS, then geospatial intelligence, now satellite communications. Eclipse's bet is that this time the infrastructure layer itself becomes something you buy rather than a program you run. Next, a customer that is trying to do exactly that.

Zach Martin

Let's talk about the one customer whose checks don't care what the stock market did last Tuesday. Space Capital reports that the FY2026 U.S. defense budget is roughly one trillion dollars, with a record 1.5 trillion dollars proposed for FY2027, and space, cyber, and AI explicitly prioritized. Note the word proposed. That's a request, not an appropriation, and I'd treat it as a signal of direction instead of a deposit in anyone's bank account. Space Capital's thesis from the Fund IV letter is that space technologies are increasingly recognized as critical infrastructure, which makes them countercyclical, with defense spending providing what they call a structural floor. A floor is a comforting image, so let me test it. At the Space Capital Summit, Eric Romo, President and COO of Impulse Space, offered a cautionary note. Confidence in long term production, he said, will ebb and flow with news cycles and OMB reports. So the floor exists, but it isn't concrete. It's more like a trampoline with a very strong frame.

Zach Martin

The session I want to walk through was published August 25, 2026, and it's called Shield and Spear: The Age of Maneuver. Space Capital's Chad Anderson interviewed Marcia Holmes, Deputy Director of Golden Dome for America, and Romo. Holmes is the architect, defining what the largest defensive space architecture the country has ever attempted is made of, and who builds it. Romo is the builder, whose company makes vehicles that move. Holmes didn't soften the picture. She said the threat is real, complex, and growing. She said Russia has fielded hypersonic systems and used them on the battlefield in Ukraine, and that China is projected to field over a thousand intercontinental ballistic missiles, doubling its arsenal. 'Our homeland is no longer a sanctuary,' she said. Her conclusion is that ground based systems can't see the newest threats early enough, and so the country has to go to space to see them from launch through their entire flight, and to position interceptors before a missile deploys its warheads or decoys.

Zach Martin

Why is that possible now? Romo's answer starts where nearly every space conversation starts, with launch cost. The last time the country seriously discussed intercepting missiles from space was the 1980s, when reaching orbit cost a fortune. A constellation of a thousand spacecraft in low Earth orbit was a pipe dream until Starlink and others made it routine. He added something that connects directly to our last chapter. The manufacturing base has matured. 'I'm not doing this for the first time,' he said. 'I'm doing it for the second or third time and scaling up.' That sentence is the industrial phase in miniature. The first time you build something, you're paying for learning. The second or third time, you're paying for production, and production is where the margins and the schedules either work or don't. Keep that in mind, because it's the same thing that justified the Eclipse team's confidence about designing for ordinary labor.

Zach Martin

Then Romo made the argument that gives the session its title, and it's a physical one, not a political one. The Pentagon's strategy for low Earth orbit is proliferation: put up too many targets to take out. But proliferation has limits. Medium Earth orbit and geostationary orbit are, in his words, 'literally, a million times bigger than LEO.' Blanketing orbits that vast with enough satellites to cover every point at every moment is, as he put it, 'completely untenable.' And those are the orbits where a lot of the most critical assets live. GPS is in medium Earth orbit. Missile warning and secure communications are in geostationary orbit. Most U.S. investment has historically gone to those high energy orbits, and those assets are now under pressure for the first time. He pointed to recent reports of three Russian craft maneuvering around a U.S. imaging satellite over Ukraine. If you can't defend with numbers, you defend with movement. Everything up there, he argued, now has to be able to move.

Zach Martin

Impulse is flying toward that capability. Two vehicles, Victus Surgo and Victus Salo, both space domain awareness missions, are in the clean room now. They're built for fast, high energy repositioning, one bound for a geostationary transfer orbit and one for low Earth orbit. And in December, with Starfish Space, the company completed a fully autonomous rendezvous in orbit, one vehicle approaching another using a single camera with no joystick. Romo was frank about how hard the next step is. Reaching out and touching an asset when it wants to be touched is one thing, he said. When it does not want to be touched is another. Closing velocities run to thousands of meters per second, and hitting a target is not so much, in his words, a bullet hitting a bullet as maybe a hundred times harder than that. I'll take his word on that, since he builds the vehicles. What it tells you for investors is that maneuver isn't a feature you add later. It's a different class of difficulty, and the hardware that proves it will be expensive to develop.

Zach Martin

That brings us to price, which is where the industrial phase either works or doesn't. The session cited a recent Congressional Budget Office estimate that space based interceptors cost roughly 22 million dollars a shot. Romo noted that in historical context that isn't outrageous, since ground based interceptors like SM 3 and GBI run from 30 million to nearly 100 million dollars. But here's the arithmetic that matters, and let me be clear this is my own back of the envelope math and not a program estimate. If a defense needed a thousand interceptors at 22 million dollars each, you're at 22 billion dollars. If it needs several thousand, you're in the tens of billions. That's the problem Romo flagged, that multiplying the number by the thousands of interceptors a real architecture would need makes the total get very large, very fast. Holmes was direct about the constraint. General Guetlein, who leads the program, has said publicly that if space based interceptors cannot be bought affordably and at scale, they will not be part of Golden Dome. That's a program leader telling the industry that cost is a pass or fail test.

Zach Martin

So how do you build something affordably when the incumbent system for buying defense hardware has been cost plus for fifty years? Holmes's answer is the procurement itself. Because Golden Dome is net new, she said, it can start from a clean sheet. She described prize based competitions, rewarding demonstrated performance instead of paperwork, and dealing directly with builders to compress timelines. Romo, who would be one of those builders, was a fan. 'You set an ambitious goal, and if you get it, you get paid a bunch. If you don't, sorry.' It's the same instinct we'll see at NASA. Pay for outcomes. Let the contractor carry the risk and the reward. The program is meant to field real capability by 2028. And Holmes described an ecosystem hub, meant to be a single front door for companies to find where they fit, with primes subcontracting to non traditional companies inside its command and control consortium rather than the other way around. The flow of money, in other words, is being redesigned along with the hardware.

Zach Martin

I want to bring in one more defense relevant story, because it shows what a government customer pivot looks like at the company level. Legacy GPS sits in medium Earth orbit, about 20,000 kilometers up, and its signals are weak and unauthenticated. That makes them vulnerable to jamming and spoofing in contested regions, and the brief I'm working from points to Ukraine, the Middle East, and commercial aviation routes. Enter Xona Space Systems, founded by Stanford GPS Lab alumni and former SpaceX engineers, with Brian Manning as CEO and Dr. Tyler Reid as CTO. Xona's constellation, called PULSAR, is a dedicated commercial navigation and timing network in low Earth orbit, planned for 258 satellites. Operating about twenty times closer to Earth delivers signals 100 to 170 times stronger than GPS, with encrypted ranging and a patented distributed timing approach that avoids costly onboard atomic clocks. Those are the company's technical claims as summarized in my research, and I'd encourage you to treat them as claims until independent testing says more.

Zach Martin

The capital trail here is a nice illustration of how a defense backed company gets financed in stages. Xona raised a 92 million dollar Series B in June 2025, which included a 20 million dollar SpaceWERX STRATFI award. Then it closed a 170 million dollar Series C in March 2026 to scale its manufacturing plant in Burlingame, California. It contracted Aerospacelab to build eight microsatellites for initial commercial rollouts. Its prototype, Pulsar 0, completed commissioning and broadcast what's described as the first fully authenticated navigation signal in history from low Earth orbit, and it has achieved enterprise integration with Trimble and the UK's National Physical Laboratory. Notice the mix. Government money to prove the technology, venture money to build, and commercial integrators to prove the market. And then the policy shift that matters most. My research says the U.S. Space Force cancelled its billion dollar internal Resilient GPS program in early 2026 and formally pivoted toward procuring existing commercial navigation services. Build became buy. I'd add a caution. A stated pivot toward commercial services isn't the same as signed contracts at scale, and revenue shows up when the contracts do.

Zach Martin

So here's the third filter. Is the government acting as a customer, or as a competitor to its own suppliers? When Golden Dome pays for demonstrated performance, when the Space Force chooses to buy navigation instead of building it, when NASA, as we'll see next, becomes an anchor tenant, the demand signal changes character. It goes from a research budget to a recurring procurement stream that a lender or an investor can underwrite. Space Capital's summary of the session is that the signal to watch is not press releases but contract awards, test campaigns, and the steady expansion of the space budgets that quietly underwrite it. I'd put it in training terms. Don't judge a program by the speech at the kickoff. Judge it by what shows up on the schedule every week. Now, let's turn from the sky overhead to the two stories where physics and price tags are in a very public negotiation: data centers in orbit, and the Moon.

Zach Martin

The pitch for orbital data centers is seductive, and Space Capital's August 20 write up of its summit session starts by saying so. Near limitless solar power, an infinite heat sink, and not a single county permit. As terrestrial AI runs into power shortages and water fights, the idea of moving compute above the grid has gone from fringe to boardroom fast. Elon Musk says all of it is going up. Jonny Dyer, CEO of Muon Space, put it this way on stage: 'It may very well be that Elon's right, and in three years all of our data centers will be in space. You may not be right about that timeline. But in some ways it doesn't matter, because we're on this secular trajectory where that's where we're headed.' Then he sat down with Space Capital partner Tom Ingersoll to separate physics from hype. His read was that the physics works, the economics do not yet, and almost every obstacle in between is an engineering problem and not a physics problem. That's a very useful sentence. It tells you where to look for the bottleneck.

Zach Martin

Let's give the case for orbit its due, because a lot of it is real. Dyer said you can pick orbits where the sun is baseload power, always on in a way that fundamentally can't be true on Earth, which means solar without the storage a ground array needs to survive night and weather. Cooling is the second advantage. Deep space is a heat sink near absolute zero, with none of the water, real estate, or permitting fights cooling triggers on Earth. Third, he said, there are no county councils in orbit, and a regulatory approach there is, literally, a global one. The catch is that capturing any of this, in his phrase, is not physics, it's engineering. And demand isn't the issue. The compute shortage, he said, is potentially unbounded in size as AI grows, while the grid, especially in the United States, is falling behind. Google alone, he noted, is spending something like 100 billion dollars on data center capital expenditure this year. A very small fraction of that goes a very long way above the atmosphere.

Zach Martin

Now the honest part, which is why the post is titled the honest case. At today's launch prices, Dyer said, a large orbital data center runs roughly eight to ten times more expensive than its terrestrial equivalent. Closing that gap means getting launch cost down about five times and cadence up about ten. Falcon 9 flies something like 100 to 200 times a year, and the math wants 500 to 1,000. And he made a neat observation. That's also the scale of launch you need to put gigawatts of data centers in space, so the two numbers move together. It all comes back to the cost of launch. Now I want you to hold that eight to ten times figure in your mind next to the first chapter, where SpaceX said it will make AI chips with Terafab to power its orbital data centers. I'm drawing a connection here, and it's mine, not a source's. A valuation that prices in orbital compute at scale is a valuation that assumes someone closes that eight to ten times gap, and the gap closes through launch economics that, today, only a few companies control.

Zach Martin

Dyer's refrain, engineers not physicists, is the most useful lens on the debate, so let me walk through his list. Communications is the objection people reach for first. Moving data between spacecraft is easy, he said. The throttle is the regulated, bandwidth limited link down to the ground. But for AI workloads, most of the heavy traffic stays machine to machine. Starlink's inter satellite links already run about 100 gigabits per second, roughly an order of magnitude short of the terabit fabrics inside a data center. A gap, not a wall. Radiation has quietly inverted, he said, because terrestrial chips are already hardened against cosmic ray bit flips since data centers hit them at scale on the ground. His favorite proof is that Google's fleet effectively became, in his words, the world's largest cosmic ray telescope. These problems are getting easier over time, not harder. And the Starlink satellites, he said, look a lot more like a rack in a data center than they look like a traditional satellite. Cheaper launch and commodity style manufacturing feed each other.

Zach Martin

What he won't wave away is how hard space still is, and I appreciated this part. 'It's still a hostile environment,' he said. 'These are autonomous robots that we're sending in space that have to take care of themselves. If you don't do that right, they will fail, and we can't go up and fix them.' The deeper challenge is integration, stitching hardware, sensing, compute, and software into one system that works in orbit. 'Integration is one of the hardest problems in technology,' he said, 'and we just haven't done it enough in space to get it to the point where it's muscle memory.' Muscle memory is exactly the right image. You can't think your way into it. You earn it by repetition, which is why cadence matters so much. And he gave the timeline honestly. 'We're not going to flip a switch tomorrow and suddenly there's gigawatts of data centers in space. It's going to be a progression.' National security workloads, and data that's simply better kept in orbit, will pull the first systems up early.

Zach Martin

Where is the near term money actually going? My research brief says industrial scale orbital data center farms remain an engineering and cost challenge for the years after 2027, and that's my summary of the brief, not a date Dyer gave. What's flowing now is edge compute on sensing spacecraft, and Muon's FireSat constellation is the example. It pairs orbital sensing with onboard AI compute and SpaceX optical links that, in Dyer's words, take a satellite from dial up to broadband. It was built with Cal Fire, with backing from Google and the Environmental Defense Fund, and it's meant to give crews something they have never had, full situational awareness with overhead views of where the fire is, how hot it's burning, and how fast it's moving, updated hourly, day and night. Space Capital also links to a Bloomberg report that Muon raised 250 million dollars with Google and Salesforce backing. So here's the pattern. Put the compute where it saves the customer something today, in this case minutes during a wildfire, and let the heavy infrastructure follow the proof.

Zach Martin

Now the Moon, and I want to start with a number that sounds trivial and isn't. In 2021, NASA Administrator Bill Nelson presented a company called Lunar Outpost with a ten cent check, representing ten percent of its one dollar winning bid to collect a sample of lunar regolith. Space Capital's profile of the company, published August 5, calls it the first commercial payment for space resources in human history. Lunar Outpost was founded in 2017 by brothers Justin and Julian Cyrus, with Forrest Meyen joining shortly after, and AJ Gemer as chief technology officer. Space Capital backed it at the seed stage in 2022, so the same disclosure applies. The founders' critique of the old model is the sharpest sentence in the whole profile. Cost plus contracting, they concluded after working at Lockheed Martin, rewarded delay and perfection over learning. The Apollo rovers were single use. Each supported a single mission of roughly three days and was abandoned. No American vehicle returned to the lunar surface for fifty three years.

Zach Martin

The institutional change that makes Lunar Outpost's model possible came out of NASA. At the Space Capital Summit, Deputy Administrator Matt Anderson, a month into the job, told CNBC's Morgan Brennan, 'We cannot do our job without commercial companies.' Space Capital's post published July 28 frames the agency as anchor customer, not sole builder. The dates are hard. A crewed lunar landing by 2028. A commercial handoff of low Earth orbit by 2030. Anderson described the contracting philosophy as incentivizing excellence. Deliver on cost, schedule, and performance and you get repeat business. Miss, and the contract gets canceled and the money redeployed. Schedule, which NASA treated as flexible for decades, is now non negotiable because the policy dates are non negotiable. Asked whether the agency prefers public or private companies, he said, 'It's just outcomes. How you're funded is irrelevant to us.' And the cadence matters. Lander and rover contracts went out in June, more followed that week, and the goal is a rhythm of awards so regular it stops being news. A predictable recurring customer is exactly what makes commercial business models financeable.

Zach Martin

The mission architecture behind that urgency is extraordinary. Anderson said Artemis III, slated for 2027, will put the three largest rockets in the world, SLS, Starship, and New Glenn, into orbit within days of each other, summarized as three rockets, two dockings, and one splashdown. It's a rehearsal in low Earth orbit for the 2028 landing. The destination is the lunar south pole, because of water ice. Propellant produced on the Moon changes the economics of everything beyond it. And Anderson was candid that the early outpost, in his words, is going to look like a junkyard. I love that candor. It signals an industrial buildout, not a highlight reel. Space Capital's lunar post adds that Administrator Jared Isaacman announced the Ignition plan in March 2026, a ten year, three phase commitment to a permanent Moon Base, and that Artemis II flew its crewed lunar flyby in April. Under the plan, Artemis IV is targeting 2028 as the first crewed landing, and the orbital Gateway has been paused.

Zach Martin

Here's how Lunar Outpost earned its place in that plan, and it's a lesson about failure as a product. Its first big mission, Lunar Voyage 1, launched February 26, 2025, aboard Intuitive Machines' IM 2. On March 6, the Athena lander touched down at an off nominal attitude, and the mission window closed within twenty four hours, before the rover could roll onto the surface. A bad outcome, by headline. But Space Capital's profile says the rover validated its navigation computer, thermal control, stereo cameras, and deployment system. It stood up the first commercial 4G and LTE network on the lunar surface with Nokia. And the Stargate command and control software reached TRL 9, the highest readiness level NASA assigns, meaning it was proven in the environment it was built for. Everything short of the drive itself got tested. That's how a company turns a missed landing into a credential, and notice that it depends on a customer who values learning. A cost plus program might have treated that day as a disaster. An outcomes based buyer treated it as data.

Zach Martin

Then the contract. On May 26, 2026, NASA selected Lunar Outpost and Astrolab to deliver the first human rated lunar terrain vehicles, the first time NASA selected commercial companies to provide astronaut surface mobility as a service rather than build it in house. The award was a 220 million dollar firm fixed price task order to advance Pegasus, a lighter vehicle for two astronauts that drives crewed, teleoperated, or fully autonomous. Per Space Capital's profile, speeds of more than nine miles per hour, a 900 kilometer range, a full year on the surface, and compatibility with both the Blue Moon Mark 1 and Starship landers. Partners include General Motors, Goodyear, MDA Space, and Leidos, and the vehicle has been through suited testing with former astronaut Dr. John Grunsfeld. Lunar Outpost is charting a path to deliver the flight vehicle to NASA in November 2027. The broader program has a 4.6 billion dollar ceiling through 2039, structured as per mission task orders, not one fixed contract. A ceiling isn't a guarantee, and the structure means each mission has to be won, but it's a recurring revenue shape, which is what lenders understand.

Zach Martin

Mobility alone won't carry the business, and the company knows it. It has built a software suite. Stargate is the command and control layer. Starweave is swarm software that lets different robots work as one fleet. Spark is an AI mission operations assistant for telemetry analysis and anomaly detection. Space Capital says a recent 30 million dollar Series B is funding those teams. And the swarm engine, developed under Air Force Research Laboratory and Space Force SpaceWERX contracts, gets its first orbital test this fall. The MARS 1 mission launches in autumn 2026 aboard a SpaceX Falcon 9 via an Exotrail transfer vehicle, and two small spacecraft will run decentralized swarm operations in low Earth orbit. Meanwhile Lunar Voyage 2 is set for the second half of 2026 aboard IM 3, bringing the first onboard GPU compute to the Moon with NVIDIA edge processors. Look at that, and tell me it isn't the same story as the data centers. Compute pushed to the edge, where waiting on a signal from 384,000 kilometers away isn't an option. A mission turns into infrastructure when the software, the recurring contract, and the hardware all start compounding on each other.

Zach Martin

Let me pull the five filters together, because this is the takeaway, and it's less glamorous than a two trillion dollar debut. One, what do you own once you're in orbit, and how are you paying for it? SpaceX paid for Cursor in stock, Rocket Lab is paying for Iridium partly in stock, and a falling share price makes that paper heavier. Two, does your factory wait for demand or does demand build your factory? Eclipse says design it, don't pour concrete. Three, is the government a recurring customer paying for outcomes? Golden Dome, the Space Force's pivot to commercial navigation, and NASA's task orders all say yes, with qualifiers about budgets and contracts still to be signed. Four, do physics and price agree? Orbital compute says physics yes, economics eight to ten times off today. Five, does a mission become infrastructure? Lunar Outpost's path from a ten cent check to a 220 million dollar task order says it can, but only through repetition. And a final caution. Most of the evidence today comes from Space Capital, an investor with positions in many of these companies. This is not investment advice. It's a map, and maps are best used by people who check the terrain.

Zach Martin

I'll leave you with the coaching version. Every great run begins with a moment that feels like arrival. The IPO is that moment. But the athletes who last aren't the ones who celebrate the loudest. They're the ones who show up for the next rep, and the one after that. Capital in this sector will flow to the companies that can do the boring things, like hitting the launch cadence, pricing the interceptor, filling the factory, landing the rover, and paying the notes, again and again, in a market that's allowed to change its mind every morning. The industrial phase isn't a prize. It's a practice schedule. Keep watching the contract awards, the test campaigns, the closing dates, and the unit costs. They're the quiet numbers that decide who's still standing when the noise fades. I'm Zach Martin. This has been Space Capital on KMKT IR Hub Radio. Real assets, real ideas, a brighter tomorrow. Thanks for spending the hour with me, and I'll see you next time.