SpaceX filed its S-1 this morning, and we devoted nearly our entire episode to picking it apart.
How the Filing Compared to Our Model
The headline for us was how little the filing changed our model. After plugging in our pre-filing assumptions and building a delta sheet, we found 2025 revenue within 5%, customer launch within 5%, and Falcon cadence and active satellites essentially confirmed (though Vlad downplayed the latter as public data anyone tracking closely could assemble). V3 Starlink deployment in the second half of 2026 and orbital data centers in 2028 both matched our timelines. The misses were instructive rather than alarming: Starlink subscriber count was off about 30% (we back into subscribers from ARK's broadband demand curve rather than modeling them directly), ARPU ran high while customer count ran low (the two cancel out in our method), and the cash position was off by 50%, which we argued is immaterial since free cash flow from Starlink and orbital compute makes today's cash balance irrelevant within a couple of years.
Competitive Position: A Multi-Year Lead
On competitive position, the filing confirmed SpaceX handles roughly 80% of all mass to orbit globally. We pegged the lead at a half decade on technology and closer to a full decade on operational scale. Starlink now accounts for 75% of Falcon 9 launches, with customer share holding steady as Starlink volume accelerates.
The ARPU Myth
The most useful reframing concerned ARPU. Online commentary has fixated on a falling figure (the consumer line moving from roughly $99 toward $66), reading it as a collapse. We argued the opposite. ARPU is dropping because Starlink is expanding into less developed economies where it is often the only option and terminals are selling out, not because pricing is eroding within a single market. As prices approach terrestrial fiber, Starlink shifts from a niche product to a genuine substitute for anyone outside dense urban cores. Aaron added nuance from listeners: in markets like Kenya, Starlink remains priced out even after cuts, and the still-elevated pricing looks characteristic of a capacity-constrained business that could drop prices further if it had the bandwidth.
Reading the Financials
On the financials, Vlad pushed back on bearish net-income takes. EBITDA exists precisely to strip out the noise of a heavily reinvesting business, and with Starlink early, Starship pre-revenue, and V3 not yet launched, large depreciating capex is exactly what you would expect from a healthy growth company. Across the three business lines, Starlink shows strong margins and growth, Launch runs at roughly half Starlink's revenue but will not grow as fast (much of it is consumed internally), and AI is deeply free-cash-flow negative. That AI figure, though, reflects 2025 only and excludes the single most interesting reveal in the filing: the Anthropic deal, worth $1.25 billion per month. The full contract totals $45 billion through 2029 but sits outside the reported backlog because it renews quarter to quarter. Aaron's view is that the intellectually honest treatment is a probability curve weighted toward the high end, given how scarce alternative compute is.
The Staggered IPO Lockup
The second standout reveal was the staggered IPO lockup. Roughly 7% of an "early release" program unlocks in steps over about 70 days, with two larger tranches (around 20% combined) tied to the Q2 and Q3 earnings releases, while Elon's founder shares are explicitly carved into a longer lockup exceeding a year. We read the staggering as deliberate supply-demand smoothing against index-fund inclusion, and as a signal that the early-release pool is large, likely covering most non-Elon holders.
Other Threads: Leaseback, Strategy, and the Idiot Index
Several other threads ran through the discussion. A "failed sale leaseback" tied to selling Colossus assets to Valor Equity Partners is an accounting treatment, not a red flag. On strategy, we expect SpaceX to push Starlink Mobile direct to consumers rather than wholesale through carriers, paralleling an Amazon "Prime plus internet" bundle with a SpaceX "AI plus internet" bundle. And on the "idiot index," Aaron noted it cost roughly $15 billion to build the entire Starship program and city versus $20 billion for xAI to bring about one gigawatt of compute online, pointing to enormous margin still to be squeezed out of the AI hardware stack.
Starship Economics
On Starship economics, the reassuring conclusion is that even a never-reusable upper stage paired with a reusable booster lands around $600 per kilogram on the first flight and $300 by ten booster reuses, already a 3X improvement over Falcon 9, before counting the gains from Starship's larger fairing. The worst case is already better than today's paradigm.
What to Watch: Flight 12
The episode closed on Flight 12, the first launch from Pad 2 and the debut of Raptor V3 engines and the V3 Starship and booster. The worst outcome would be a pad explosion, which historically means a roughly three-month delay; anything past the pad is recoverable. With weather a coin flip and us expecting the market to overreact regardless, the message was patience.