SpaceX Isn't a Moonshot Anymore — It's a Cash-Flow Problem the Market Hasn't Solved

The market went through the full motion on SpaceX in seven weeks. It opened at $135, surged to $225, and collapsed back through the IPO price — down about 40% from that June peak. Dan Ives initiated coverage at $190, then publicly acknowledged the "valuation support is not there" as the stock traded near his entry point. The headlines settled on cash burn, speculative AI ambitions, and a company that raised $86 billion just to lose money faster.

The old story is a rocket company pretending to be a tech business. The newer evidence says something different: the cash engine exists, the contracted backlog is real, and the $18.4 billion quarterly capex cycle — ugly as it looks — is the kind of infrastructure wall that eventually flips. The question isn't whether SpaceX can grow revenue. It's whether the free cash flow hole closes before the market loses patience.

What Actually Happened in Q2

SpaceX reported its first public earnings on August 4. Revenue came in at $7.81 billion, a 92% year-over-year jump that beat the $6.93 billion consensus by roughly 13%. Earnings per share landed at a loss of $0.09 versus an expected loss of $0.26. Adjusted EBITDA — a rough proxy for operating cash generation before depreciation and non-cash charges — climbed 191% to $3.54 billion.

Every segment beat. Connectivity (the Starlink satellite internet business) generated $4.29 billion. Space (launch services) brought in $962 million. AI infrastructure contributed $2.56 billion, up 247% year-over-year.

The company ended the quarter with approximately $100 billion in cash and investments and a $47.5 billion backlog of contracted business. By almost any measure of top-line momentum, the operating trajectory is steep.

Here's what the tape didn't celebrate. Capital expenditure hit $18.37 billion for the quarter. Most of that — $15.83 billion — went to AI data center buildout. Free cash flow, the metric that separates growth stories from compounding businesses, came out sharply negative. Analysts project FCF losses could approach $100 billion next year. That's not a speed bump. That's the mountain the entire thesis has to climb.

The Starlink Engine

This is the one segment where the numbers actually work. Starlink now has 12 million subscribers and generated $4.29 billion in Q2 revenue. It's the only part of the business producing an operating profit. The adjusted EBITDA from connectivity alone ran $2.41 billion.

Average revenue per subscriber has fallen from $85 a year ago to roughly $66, which reflects the move into lower-priced international markets and the introduction of cheaper plans. That trade-off — more subscribers at lower margins — is a known path. It's how terrestrial broadband operators scaled, and it's the same playbook SpaceX is running from orbit. The addressable market is still massive: less than 1% of global telecom and broadband, by the company's own filings.

Starlink is the reason this company doesn't need to raise more capital in the near term. It's also the reason the $100 billion FCF hole isn't existential — it's a bridge question, not a solvency question. The cash pile is large enough to fund the burn cycle while revenue doubles.

The AI Backlog and the Capex Wall

The AI segment is where the market's skepticism is concentrated, and for good reason. $15.83 billion in quarterly capex is not a number the market likes. Google, Amazon, and Microsoft all went through similar hyperscaler buildout cycles, and their stocks took months to digest the FCF impact. But SpaceX's projected FCF losses are larger than what any of those companies faced at their peak buildout.

There's a counterbalance that the headlines have barely touched. Management disclosed $14.1 billion in contracted AI cloud services that haven't been recognized as revenue yet. These are compute leasing agreements with Anthropic, Google, and Reflection AI — firms paying SpaceX to use its Colossus data centers in Memphis. Once recognized, that backlog pushes meaningful recurring revenue into the pipeline.

The durability risk is real: these contracts are cancellable within 90 days. They aren't the multi-year lock-ins you see from AWS or Azure. That makes the AI revenue less stable than the Starlink subscription base. But the revenue is contracted, which means the buildout has a demand floor even if the execution doesn't hit every milestone.

Why the Stock Keeps Falling Anyway

The supply mechanics are doing heavy lifting right now. About 911 million insider and early-investor shares unlocked on August 6, releasing roughly $106 billion in potential float from an IPO that sold only about 4% of the company. That's not a fundamental signal — it's a mechanical overhang that creates months of selling pressure. The stock has traded down to the $105–$116 range, well below the $135 IPO price.

Ives left Wedbush to launch his own firm and has been transparent about the disconnect between his long-term conviction and the near-term price action. He's pointing at a three-year-to-five-year arc. The stock is being judged on a three-month FCF print. That gap between timelines is where the inflection setup lives — if the operating trajectory holds long enough for the lockup supply to be absorbed.

The Bridge

This is the version of the story that matters over the next 12 months:

Starlink doubles its subscriber base while the AI contracted backlog begins recognizing revenue. Capex stays elevated in the first half of 2027 but then begins to taper as the initial data center wave completes. Free cash flow doesn't turn positive immediately, but the quarterly FCF loss narrows from the projected $100 billion annual run rate toward something the market can price. If that path plays out, the stock at $105–$116 is well below where the business should trade.

The counter-argument is simple. If capex doesn't taper, if AI contracts get cancelled at the 90-day window, or if the Starlink ARPU decline outpaces subscriber growth, the FCF hole widens instead of narrowing. Then the $100 billion projected loss becomes permanent structural burn, not a transitional infrastructure cost. And no amount of backlog justifies that.

I'm not putting a specific target on this yet. Simple forward multiples work when the cash-flow bridge is explicit, and it isn't clear enough right now. The capex cycle hasn't peaked. The AI revenue hasn't been proven durable. The FCF trajectory is still a projection, not a demonstrated pattern.

What I would watch is the next two quarterly reports. If capex stays above $15 billion per quarter and FCF losses widen past the current estimates, the setup breaks. If capex begins flattening and Starlink continues its subscriber ramp, the FCF bridge gets real enough to build a target around.

The tripwire is straightforward. A quarter where combined revenue growth falls below 50% year-over-year while capex stays above $15 billion would signal that the revenue ramp can't keep up with the infrastructure spend. That's the condition where the old story — cash-burning rocket company with speculative ambitions — starts looking like the right one again.

Until then, the market is pricing the FCF mountain as if it's permanent. The evidence so far says it's transitional. Whether the market lives to see that distinction is the whole trade.