SpaceX: The Next Revolution—At What Price?

Full transcript · 30-second passages

1. 0:00

   Um, imagine running a business that generates, you know, over $1.6 billion in pure operational profit. Right. Total cash machine. Exactly. It's a wildly successful product. Customers absolutely love it. The margins are, well, they're fantastic. Yeah. Yet somehow at the end of the day, your parent company is still over half a billion dollars in the red. It's wild. It is. Today on the deep dive, we are looking at this ultimate financial paradox. We're talking about SpaceX or, you know, as it's known now on the public markets, Space Exploration Technologies Corp.

2. 0:30

   Trading under the ticker SPCX. Right. SPCX. And honestly, it is easily one of the most complex corporate structures on the public markets right now. Oh, absolutely. We are unpacking a really incredibly data rich report today. It's the AI investment edge from iPulse AI report on SPCX. Good stuff. Yeah. And to get to the bottom of this profitability paradox, we're cross-referencing their EP-023 report with, well, the latest SEC filings, the official launch logs, and a fascinating recent NVIDIA press release about orbital AI.

3. 1:00

   Right. Which really serves as the missing puzzle piece here. Definitely. Okay. Let's unpack this because the scale of the money moving around inside this company is, I mean, it's just hard to wrap your head around. It really is. How can their connectivity arm, Starlink, be such a cash generating machine, yet the group as a whole is just burning through capital at an unprecedented rate? Like, where is all that cash actually going? Well, to figure that out, we really have to look under the hood of the newly public entity, because since listing in June 2026, SpaceX essentially operates as three distinct businesses sharing one roof.

4. 1:30

   Okay. Break those down for us. Sure. First, there's space. This is the legacy business, right? Selling rocket launches and spacecraft services to both commercial and government clients. The classic rocket business. Exactly. Second, there's Starlink, which sells satellite internet to households, airlines, ships, and provides secure communications for governments. Right. And then there's the third arm, the one that's really warping their balance sheet.

5. 2:00

   The AI division. Yeah. This segment sells AI products and massive compute infrastructure. It absorbs Grok and X from the February XAI merger and, you know, Cursor, which they just acquired in August. Right. Because they have these three distinct operations sharing resources, analyzing them requires a bit of a mental shift. It does. The best way I can think to clarify the dual use nature of their reusable rocket fleet is with a delivery truck analogy. Oh, I like that. Yeah. So imagine owning a massive fleet of delivery trucks. Half of them are doing external jobs, hauling freight for outside customers and bringing in actual cash.

6. 2:30

   Like revenue generating. Exactly. But the other half of your fleet, they are strictly hauling materials to build your own massive internal warehouse network. Yeah. They aren't bringing in a dime of outside cash. They are purely an expense building your infrastructure. That is the perfect way to visualize it. I mean, the rockets launch customer payloads for external revenue, but they also launch internal Starlink satellites to build out their own network.

7. 3:00

   Right. So you have three divisions with entirely different margins and totally different scales of investment. It's completely skewed. Yeah. If you just look at the top line revenue, you're missing the fact that a highly profitable connectivity engine is essentially trapped inside a much wider, massively loss-making infrastructure program. Because they are basically building this massive internal warehouse and it completely distorts their financials. Let's look at the actual numbers from Q2, 2026 to see the damage.

8. 3:30

   Okay. So the Q2 results ending June 30th show incredible top line momentum. Revenue hits $7.814 billion. Wow. Yeah. That is up almost 92% year over year. That's massive. It is. But despite that near doubling of revenue, the group still posted a $541 million GAAP net loss. And when you break down the margins of those three divisions, the source of the bleeding becomes pretty obvious. Very obvious.

9. 4:00

   Connectivity, which is Starlink, operates at a phenomenal 38.6% margin. Right. Bringing in $1.656 billion in operating profit. Incredible. But the space division lost $542 million. Okay. And the AI division. That segment recorded a staggering $1.257 billion operating loss. Wow. I have to jump in here actually and push back on a specific accounting phrase the report uses for that AI division.

10. 4:30

    Okay. Go ahead. Because it notes the AI segment has, quote, positive adjusted EBITDA. Ah, yes. The magic adjusted numbers. Exactly. And I want to be crystal clear with everyone listening about what that actually means. Saying you have positive adjusted EBITDA by adding back your depreciation, your stock compensation, and your restructuring costs. Right. Well, it's like claiming your personal budget is in the green only because you, quote, added back your rent and your car payments. That's exactly what it is. Right. It is a legitimately disclosed non-GAAP reconciliation, sure.

11. 5:00

    But we need to be very clear here. It is neither GAAP profit nor actual segment cash flow. Yeah, absolutely. The mechanism of adding those costs back might show you an underlying operational trend, but, you know, it doesn't pay the actual bills. No, it does not. And the real GAAP operating loss for AI remains over $1.2 billion. Right. And those bills are enormous, especially when we look at the capital expenditure. The actual physical cash going out the door to build things. What are we looking at there? In Q2 alone, their total CAPEX was $18.369 billion.

12. 5:30

    Wait. Nearly $18.5 billion spent on infrastructure in just three months? Yes. Let me guess. The vast majority of that is going to those massive new data centers. You'd be right. Out of that total, $15.828 billion went strictly to the AI segment. Wow. That means 86% of their entire capital expenditure went purely to AI infrastructure. That's insane. And the report stresses that this is the whole AI segment allocation.

13. 6:00

    They aren't itemizing single named project budgets for us. Right. Connectivity only took about $1.4 billion and space took just under $1.2 billion. The AI build out is absolutely consuming the parent company's capital. So if I'm looking at a company spending almost $16 billion on AI infrastructure in a single quarter, my immediate question is, how on earth are they funding this? Well, they are leaning heavily on the massive war chest they built from their IPO and from debt funding.

14. 6:30

    Okay. As of June 30th, they have roughly $100 billion in combined liquidity. That's cash, equivalents, and marketable securities. That's a huge buffer. It is. Bounced against that, they carry $39.364 billion in current and long-term debt, plus finance leases. So the runway is undeniably huge. Roughly $100 billion in the bank against $40 billion in debt. Right. But I imagine with a burn rate this high, even $100 billion can evaporate pretty quickly.

15. 7:00

    Oh, it absolutely can. The report derives a deeply concerning metric for the first half of the year. What is it? When you take their operating cash generation and subtract the massive purchases of property, plant, and equipment, you are left with a six-month cash investment gap of negative $25.010 billion. A $25 billion hole in just six months. That is astonishing. Yeah. And because this is a loss-making group, you can't just look at a conventional price-to-earnings ratio to figure out if the stock is a good deal.

16. 7:30

    You have to anchor it to highly specific market tracking. Exactly. To do that, the iPulse AI report anchors its entire valuation model to a very specific reference price for SPCX Class A shares. Okay. What's the price? $146.37. Wow. That level of precision, $146.37, tells me this isn't just a daily average. What's the context there?

17. 8:00

    It was an after-hours alpaca IEX trade on Friday, October 16, 2026, executed at precisely 16.59.47 New York time. So almost an hour after the bell. Exactly. That is almost one full hour after the 16:00 regular-session close. Tracking off-exchange after-hours trades down to the second like this, it's pure forensic institutional analysis. Right. It tells us where the absolute most committed money is clearing trades when the public markets are closed.

18. 8:30

    And anyone buying at $146.37 needs to understand a harsh reality. Investors in SPCX are getting zero regular common cash dividends. Zero. The modeled future dividend cash is also zero. So your return relies entirely on the share price going up. Yeah. It's all appreciation. So what does this all mean for the forecasts? If I buy at $146, what do the models actually say? Well, the evidence-weighted base targets predict a classic J-curve.

19. 9:00

    Things get slightly worse before they get significantly better. Okay. They project a slight dip of negative 1.1% by September 18, 2027. Right. But looking further out, they project a powerful rally, hitting positive 48.1% by September 18, 2031. A J-curve makes perfect sense. I mean, you have a year or two of slight dips as they burn cash to build the infrastructure, followed by an aggressive climb once those assets start paying off. Exactly.

20. 9:30

    But what are the underlying assumptions driving those views? The bull and bear cases rely heavily on qualitative mechanisms right now. The bulls are betting on the engineering. Naturally. They believe that faster commercial starship reliability will drastically lower the cost of putting mass into orbit. Right. If it costs a fraction of the price to launch a satellite, the margins just explode. Exactly. They also bet that AI compute utilization will quickly outpace these massive upfront investments. But the bears must be terrified of shareholder dilution.

21. 10:00

    Oh, absolutely. I mean, if you have a negative $25 billion cash gap every six months, at some point you have to issue more shares to raise money, which shrinks the slice of the pie for everyone else. Exactly. The bears worry that these heavy recurring network investments are going to drag on for years. Yeah. They worry that AI pricing will disappoint. And they are incredibly focused on the reality that raising more capital weakens the per share economics for existing owners. Let's pivot slightly and look at the ordinary business drivers that are actually working today.

22. 10:30

    Because the connectivity side is genuinely booming. It is. Starlink doubled its subscriber base from 6 million users a year ago to 12 million by June 30th. That is massive adoption. Huge. But there is a friction point in the data. Even as they doubled their users, the average revenue per user, the ARPU, dropped from $85 down to $66. Right. Now, on the surface, bringing in less money per customer sounds bad, but their operating income actually rose.

23. 11:00

    I want to explain why lower ARPU and rising profits can coexist here. Go for it. It all comes down to the mechanics of scale. When you build a satellite network, the massive costs are fixed, right? Building the satellites, launching them, setting up ground stations. Yeah, the upfront capital. Exactly. Whether you have one customer or a million, those satellites are up there. So when you jump to 12 million users, you spread those massive shared fixed costs over a much larger pool. Right. Even if each customer pays a little less, your overall profitability surges.

24. 11:30

    That is the magic of operating leverage. But as the report points out, you can't just scale infinitely for free. Right. There are limits. There are finite beam and network capacity costs. Unlike broadcasting a radio signal that anyone can tune into, satellite internet requires actual physical tracking and dedicated bandwidth per region. Right. Every new continent they dominate requires expensive gateway expansions and localized support.

25. 12:00

    And there are also market frictions beyond the technology. Yeah. There are potential stock lockup expirations. Right. The early shares. Early investors and employees hold conditional shares that are currently locked. Depending on the conditions. When those expire, a massive new supply of stock could hit the market and pressure the price. What about the extreme edges of the forecast? Like the massive upside or downside tails? Well, on the upside tail, if mobile connectivity becomes a globally dominant business, or if orbital logistics and industrial services become highly economic, this company reaches a totally different valuation tier.

26. 12:30

    But the downside tail is severe, I imagine. It is. The biggest risk is a serious satellite debris event, specifically Kessler syndrome scenarios. Oh, Kessler syndrome. For those unfamiliar, Kessler syndrome is essentially a catastrophic chain reaction in space. Right, like a runaway collision. Exactly. Imagine two satellites colliding and shattering into thousands of pieces of shrapnel traveling at orbital speeds.

27. 13:00

    Those pieces hit other satellites, creating more shrapnel, until orbit becomes an impassable minefield of debris. It's terrifying. It could literally lock us out of space. Yeah, it's an existential threat to their entire business model. The other major downside risk is extended starship failures, which would completely derail their timeline for reducing launch costs. Here's where it gets really interesting. When you synthesize all this data, the central gap in the research isn't about whether the rockets actually fly or people want satellite internet.

28. 13:30

    Right, we know that part works. We know the engineering works. We know the demand is there. The alpha gap, the space where investors can actually find an edge in the market, is entirely about all the data. It's entirely about operational growth versus the actual cash available to owners after these massive investments. If we connect this to the bigger picture, it is exactly the same dynamic we discussed in our previous deep dive on Alphabet. Yeah, totally. Just like Alphabet, the key test for SpaceX isn't whether they can build incredible AI models or cool technology.

29. 14:00

    The test is whether their current operating cash can increasingly carry that ambitious infrastructure bill. And the analysts looking at SPCX are fiercely divided on when that cash inflection point will actually happen. The iPulse AI September panel highlights this perfectly. We have 14 different AI research contributions using nine framework families across three model versions. Okay. And what is fascinating is that 12 of those 14 contributions use the exact same underlying Gemini engine. Wait, wait.

30. 14:30

    If 12 of these models are using the exact same underlying Gemini engine, looking at the exact same financial data, Yeah. why is there such a wide disagreement on when this company turns cash flow positive and what the terminal value is? Because the math changes entirely based on how you weight the future. The panel splits into two distinct camps. Okay. Who are they? The highest weighted view, holding approximately 10.2% and 9.3% weights, is the polymath forensic scrutiny camp.

31. 15:00

    Let me guess. They are the accountants. Precisely. They focus heavily on the strict realities of current valuation. Right. They heavily discount future cash flows because of current interest rates, and they are obsessed with tracking the share counts, the cash investment burns, and those lockup expirations. Okay. Opposing them is the visionary engineering optimism perspective, holding a 4.6% weight. And they are the engineers who believe the J curve is going to spike. Yes. In their models, they heavily weight the exponential cost drops of reusable rockets

32. 15:30

    and the scale of the expanding global connectivity network. Right. They believe the engineering breakthroughs will entirely erase the short-term cash burn. Well, we have an immediate catalyst coming up that will test both of these camps. If you are tracking this, mark October 20, 2026 on your calendar. Definitely. Q3 results arrive after the U.S. close with a webcast scheduled for 1630 New York time. That report is going to be vital. It will test AI revenue.

33. 16:00

    It will show us the Starlink ARPU trends. And critically, it will reveal exactly how much of that $100 billion liquidity pool they burned through in Q3. Right. And this leads us to the concrete quarterly turning points mapped out from that 146.37 base. Walk us through the exact J curve path so we know what to watch for. Okay. The consensus models project an early dip. December 2026 at negative 1.2%. Okay. Followed by a deeper dip in March 2027 to negative 4.6%. Makes sense. But then the recovery begins.

34. 16:30

    September 2028 jumps to positive 8.2%. Nice. September 2029 accelerates to positive 20.5%. And it finally hits that positive 48.1% target in September 2031. To hit those 2028 and 2029 numbers, they really have to prove the visionary camp right. Yeah. And that brings up Flight 14, the September 28th orbital flight that successfully delivered 26 Starlink V3 satellites. Right. A massive win for the visionary engineering camp.

35. 17:00

    It is. But I want to push back on getting too excited about single launches. One successful orbital delivery is an amazing spectacle. Yeah. But the real financial test for those quarterly turning points is repeated economical operations. It's about rapid, full reuse of the vehicles. Turning a rocket around like a commercial airplane is what builds a business. Launching it once just proves the physics work. That tension between spectacle and economics is exactly why the research panel evolved the way it did.

36. 17:30

    How so? From July to September, the composition expanded from 12 contributions on a single version to 14 across three. Oh, I see. The shifting methodology reflects a growing realization that they have to balance the infrastructure potential against the sheer financial weight of building it. It naturally reminds me of episode 22, our deep dive on Tesla. Tesla is the perfect last public stock baseline for this. Completely agree. Obviously, they have totally distinct assets and end markets.

37. 18:00

    You aren't cross-shopping a Model 3 with a Falcon 9. No, definitely not. But Elon Musk's companies share the exact same capital allocation question. How do you balance breathtaking engineering spectacle and insanely ambitious AI investments against the eventual need to return actual cash to the shareholders? Which brings up the crucial question for anyone following this stock. What would actually invalidate the concerns of the forensic cash scrutiny camp?

38. 18:30

    Right. What proves the bears wrong? The bear thesis collapses if operating cash persistently catches up with investment, while the per share economics improve. Okay. If Starlink's enterprise margins hold and the AI customer concentration diversifies while revenue grows, the massive CapEx becomes justified. But I want to warn you about how to read those CapEx numbers. Revenue growth paired with an ever larger need for outside funding absolutely supports the bear concern. Yeah, it does.

39. 19:00

    It means you are growing, but you are bleeding to do it. Also, be careful if you see a sudden drop in CapEx in these upcoming reports. Lower CapEx is only a good thing if it means they have achieved efficient working capacity, meaning the factories and data centers are built and running smoothly. Exactly. It is very bad if CapEx drops because deployment has stalled or they hit a technical wall. Durable cash economics over time matter far more than a single profitable quarter. And just as we evaluate their terrestrial data center build out, we have to look at the next frontier.

40. 19:30

    Right. The long-term vision that could rewrite all of these models entirely. While the current $15.8 billion AI segment CapEx is terrestrial infrastructure, their SEC-filed strategy reveals a forward business opportunity called Starmind. Starmind. Starmind. It sounds like science fiction, but this is a very real, verified development program aimed at putting AI computing directly in orbit. Exactly. And the industry is taking it seriously. On August 24th, NVIDIA explicitly announced that SpaceX is adopting optimized NVIDIA Vera CPUs to accelerate this massive-scale agentic AI.

41. 20:00

    Agentic AI meaning artificial intelligence systems that don't just answer prompts but actively take actions and manage complex systems autonomously. Right. But putting data centers in space is incredibly difficult. Oh, the engineering challenges are immense. You have to figure out orbital power generation, radiation shielding, and crucially, thermal management in a vacuum. People often forget that space isn't just cold, it's a vacuum.

42. 20:30

    Yes. Cooling a massive heat-generating AI server in space is like trying to cool down a boiling laptop wrapped inside a thermos. Without air to carry the heat away through convection, managing temperatures is a monumental physics problem. It is. But if they solve it, they unlock an entirely different stage of the business compared to today's terrestrial build-out. Starlink's connectivity profits are real, and the SpaceX group still needs much better cash conversion after its enormous AI and space investment.

43. 21:00

    Right. The question is whether operating cash can grow faster than the investment bill while per-share economics improve. That is the test to carry into the next results. Explore the research, primary sources, four financial findings, and full forecast path at iPulseAI.com. Listen, watch, or subscribe for more AI Investment Edge. Stay tuned for Novo Nordisk next on AI Investment Edge, where we'll explore its medicines, growth, and the price investors pay for that growth. And I want to leave you with one final provocative thought about StarMind.

44. 21:30

    Right now, here on Earth, the biggest bottleneck capping AI growth isn't a lack of microchips. It's the energy grid. Absolutely. We simply don't have enough power plants to run the massive data centers required for the next generation of AI. But in space, you have uninhibited 24-7 solar access. If power, thermal management, and reliability can be solved in the vacuum of space, could off-world data centers eventually bypass the massive energy grid bottlenecks currently capping AI growth here on Earth?

45. 22:00

    It fundamentally changes the ceiling of what AI can achieve. Thank you.