SpaceX Stock (SPCX): What Actually Moves It

I was in the SPCX IPO. I trade it on the swings most days, and I carry a longer-horizon position in retirement accounts. So I am asking the same company two different questions at once, and neither one is on autopilot.

Those two questions want different things from the same ticker. The swing trade cares about the lockup calendar, positioning and whatever got posted overnight. The longer-horizon money cares about whether four engines fire. But long horizon does not mean I stop looking. With investing you need your tap dancing shoes on and your head out the window watching for the macro that can wreck even a good plan, and that applies to both sides of the account.

Most of what gets written about SPCX is either cheerleading or catastrophizing, and almost none of it explains the machine underneath the ticker.

Every number below comes from SpaceX's own SEC filings, mostly the Q2 2026 earnings release, or from primary reporting I could open and read. Not from takes.

I'm not a financial advisor and nothing here is financial advice. I was in the SPCX IPO, I trade it actively, and I hold it in retirement accounts, so I am a biased source by definition. This is education and research built from public filings. Always do your own research and size like you might be wrong.

So this is the thing I wanted to exist. It is long, and it is built to be read in sections.

What you actually own when you buy a share

Here is the part almost everyone skips. SPCX isn't a rocket company with a side business. It's three companies stapled together, and the stapling happened four months before the IPO.

On February 2, 2026, SpaceX acquired xAI in an all-stock deal that valued SpaceX at $1 trillion and xAI, which already contained X, at $250 billion. By May, Musk said xAI would stop existing as a standalone brand, with Grok and X moving under a unit called SpaceXAI.

So when the combined company listed on Nasdaq on June 12 under SPCX, buyers got rockets, satellites, an AI lab and a social network in one ticker.

The company reports three segments, and the gap between them is the whole story:

• Connectivity, which is Starlink plus Starshield: $4.29 billion of Q2 revenue, up 66% year over year, with $1.66 billion of operating income. The only segment that makes money.

• Space, which is Falcon and Starship: $962 million of Q2 revenue and a $542 million operating loss.

• AI, which is Grok, X and the data centers: $2.56 billion of Q2 revenue, up 247%, with a $1.26 billion operating loss and $15.8 billion of capex in ninety days.

Now the ownership fine print, and it matters more here than on any stock I have looked at.

SPCX has two share classes. Class A, the one you can buy, gets one vote. Class B gets ten, and only Musk, his family and certain entities can ever receive it. He controls roughly 82% of the voting power with about 42% of the economics. The company incorporated in Texas, requires arbitration for shareholder disputes, and restricts shareholder proposals.

Reuters put it plainly at IPO time. The only person who can fire Musk is Musk.

I am not making a moral argument about that. I am telling you what the security is. You're buying economic exposure to the machine below, with essentially no governance rights, attached permanently to one man's judgment. Price it that way or don't buy it.

The first 75 days, and what each leg taught

The history is short enough that we can walk all of it, and each leg taught something specific.

June 12 to 16: the euphoria leg

Priced at $135, opened at $150, closed day one at $160.95, and printed $225.64 intraday on June 16, four sessions in. At the peak the market was paying well over $2.5 trillion.

What it taught: with 638.9 million tradable shares against about 13 billion outstanding, that wasn't price discovery. That was a demand pile-up in a tiny float.

Mid-June to July 22: the give-back

Three straight down sessions off the high, then a grind to a closing low of $115.26 on July 22. That's 15% below the IPO price and 49% below the peak.

Short sellers built a $15.5 billion paper profit on the way down, and by mid-July SPCX was the most shorted new listing on Wall Street, with short interest near 34% of the tradable float.

What it taught: a small float cuts both ways. It amplified the ramp and then it amplified the bleed.

August 4: the first earnings print

This is the day worth studying, because it showed exactly which wires are live.

The stock ran 7.9% into the report and popped another 9.4% on the headline numbers, which were genuinely enormous. Revenue up 92%. Adjusted EBITDA up 191%. The AI segment swinging to positive EBITDA for the first time.

Then the market read one line. Capex of $18.4 billion in a single quarter, more than six times the prior year and roughly $5 billion above the average analyst estimate, with $15.8 billion of it in AI. Shares reversed 8.6% after hours to about $114.60.

What it taught: at this valuation growth is already assumed. The stock trades on the gap between growth and what it costs to produce.

August 5 to 12: the squeeze

Three days after that flush, SPCX ripped from an open near $115 to close over $133, then kept going and closed back above the $135 IPO price on August 10 for the first time in weeks. CNBC put the rebound at more than 40% off the low.

Two mechanical things did it. The August 6 lockup released a large block, which expanded the tradable float and cut short interest as a percentage from about 34% to about 11%. And shorts who rode the July slide covered into strength.

What it taught: in year one of this stock, share supply can matter more than fundamentals over any given week.

August 13 to today: the fade into supply

From the mid-August highs the stock has drifted back to $136.84 as I write this, dipping under the IPO price intraday along the way.

On August 22 Musk posted that SpaceX would wind down Falcon once Starship flies reliably several times per week. The stock slid overnight and a rare sell rating with 27% downside made the rounds.

What it taught: his posts are a live input to this price in both directions, and the market has not decided whether retiring your profitable product line in favor of your developmental one is bullish or terrifying. It is arguably both at once.

Notice what is missing from that whole history. No competitor did anything. No customer left. No rocket blew up in a way that mattered financially. Every major move came from valuation, positioning, share supply, capex, or something Elon said.

Revenue line one: Starlink, which pays for everything

Starlink is 55% of revenue and the entire source of operating profit, so it gets the most scrutiny.

Q2, straight from the company's 8-K:

• Connectivity revenue: $4.29 billion, up 66% year over year.

• Consumer: $2.49 billion, up 44%.

• Enterprise and government: $1.81 billion, up 108%.

• Operating income: $1.66 billion, roughly a 39% margin, up 79%.

• Subscribers: 12.0 million at quarter end, exactly double a year earlier, up 1.7 million in ninety days. By August the company was citing over 13 million.

Then the number the bears correctly obsess over. ARPU was $85 a month in Q2 last year. It was $66 in Q1 this year and $66 again in Q2.

That's a 22% drop in a year, driven by international expansion and cheaper plans.

The bull read is that ARPU has now held flat for two straight quarters while subscribers doubled, which is exactly the trade you'd want. Give up price, take the planet. The bear read is that every new subscriber comes from a poorer market on a cheaper plan, so the mix keeps degrading and the doubling gets harder each year.

Both of those are true right now. Which one shows up next quarter is the question, and sequential ARPU is the single most important number in the report.

The piece I think most commentary underprices is enterprise and government, because 108% growth on a nearly two billion dollar quarterly base isn't a rounding error. Three things live in there:

• Aviation and maritime: in Q2 alone SpaceX signed American Airlines and activated service on Southwest, Virgin Atlantic, Iberia and Aer Lingus. Multi-year, high ARPU, and much stickier than a price-sensitive household.

• Starshield: the government-exclusive encrypted version serving the NRO, Space Force and other agencies. Roughly $1.8 billion of 2025 revenue, and in Q2 the company announced over $6 billion in new multi-year US government contracts, mostly two Space Force awards for LEO communications and sensing.

• Direct-to-cell and spectrum: direct-to-cell passed 10 million customers by February, and the FCC approved the EchoStar license transfer giving SpaceX 65 MHz in the US plus global mobile satellite spectrum. Carrier partners now include SoftBank, NTT Docomo and Spark New Zealand.

That spectrum is the part I keep coming back to. Satellites depreciate. Spectrum doesn't, and there isn't any more of it.

On competition, honestly stated: Starlink has around 11,000 satellites in orbit. Amazon's Kuiper has fewer than 400, missed its July deployment milestone and pushed its commercial timeline toward 2028. The Chinese constellations are real but will mostly win in markets Starlink was never getting anyway.

For the next few years the practical limit on Starlink isn't competition. It's capacity, which is why V3 satellites and Starship matter so much.

Revenue line two: launch, which is a factory and not a product

This is the part that confuses people who come to SPCX expecting a rocket company.

The Space segment did $962 million of Q2 revenue and lost $542 million from operations. For the first half, $1.58 billion of revenue against $1.2 billion of operating losses. The most famous launch business on Earth loses money.

The reason is that SpaceX is its own biggest customer by a mile. Of 78 launches in the first half, 17 were for customers. The other 61 carried SpaceX's own satellites. Of 1,041 metric tons put in orbit, 908 tons were internal.

When SpaceX launches for itself no revenue gets recognized. The cost lands in the Connectivity segment's asset base instead. So the Space income statement makes the business look worse than it is, because its real output, cheap orbital capacity, shows up as somebody else's margin.

That's why I think of launch as the factory. Falcon 9 is the most reliable machine in the history of spaceflight, and this week one booster flew its 37th mission. Its financial job is to make Starlink possible at a cost nobody else can touch.

Starship is where the thesis either compounds or dies. The company states in its own earnings release that it believes Starship will cut the cost to orbit by 99% or more versus the historical average. Follow that through each segment:

• Connectivity gets V3 satellites: far more bandwidth, launched in bulk, which answers the capacity ceiling and the ARPU problem at the same time.

• Space gets a payload business nobody else can quote: the customer line stops being a rounding error.

• AI eventually gets compute in orbit: which sounds like science fiction until you read that Musk's own pay package includes a tranche tied to 100 terawatts of space-based data center compute.

Where Starship actually stands, without hype or doom:

• Flight 12, May: debuted the V3 vehicle, landed the ship precisely, lost the booster, triggered an FAA mishap review.

• Flight 13, July: hit every objective. Deployed 20 production V3 Starlink satellites, relit a Raptor in space, and executed the softest ship splashdown yet with an intact heat shield.

• Flight 14: targeted for around September 15.

• The tower catch of the ship: slipped to later flights. Musk says the catch is probably a few months out and the first ship reflight is end of this year or early next.

Revenue line three: AI, which is the furnace

Fastest growth, biggest losses, rawest nerves.

Q2 AI revenue was $2.56 billion, up 247% and more than triple Q1. But read what is inside it. Advertising, which is basically X, was $367 million, down from $426 million a year ago. The social network inside this company is shrinking.

All the explosion came from AI solutions and infrastructure at $2.19 billion, driven by Cloud Services Agreements, which are contracts to sell access to SpaceX compute. The company signed $14.1 billion of contracted CSA sales in the quarter and recognized $1.6 billion of incremental infrastructure revenue from them. Compute capacity hit 1.4 gigawatts, up from 0.4 a year ago.

The quarter also included a $60 billion all-stock agreement to buy Cursor, the AI coding company, expected to close in Q3. Grok 4.5 shipped in July and was trained alongside Cursor's tooling. Whatever you think of the price, the intent is legible. He's assembling an enterprise AI stack, not a chatbot.

Two numbers should keep any owner of this stock awake:

• Capex: $15.8 billion of AI capex in one quarter and $23.6 billion in the first half, heading toward a full-year number that will likely exceed the company's entire 2025 revenue. This is the line that took 18% off the stock in an afternoon.

• Contract durability: analysts have flagged that the cloud agreements are cancellable on 90 days notice after December 31, 2026. The revenue is real and the $47.5 billion backlog is real, but a big slice of the AI story sits on contracts the counterparties can walk away from.

The fair read on the other side: AI segment adjusted EBITDA actually turned positive in Q2 at $1.1 billion, from negative $609 million in Q1. The operating loss halved sequentially. And the company ended June with $100 billion of cash and marketable securities after raising $85.7 billion net in the IPO and pricing a debut $25 billion investment-grade bond at a 5.855% weighted average rate.

The furnace is enormous and it is fully funded. The bet is whether demand for compute stays ahead of the depreciation curve, which is the same bet the whole AI economy is making, concentrated here in one ticker with a rocket attached.

What actually makes this stock go up

From 75 days of evidence plus the structure of the business, ranked:

1. Starlink subscriber adds with stable ARPU: the doubling to 12 million with ARPU flat at $66 for two quarters is the load-bearing wall under the entire valuation. Every quarter that repeats, the profitable core gets bigger and the multiple gets easier to defend.

2. Enterprise, government and Starshield contracts: the 108% growth line. Airline signings and Space Force awards are the highest quality revenue in the company and each one lands as a headline the market can price that day.

3. CSA signings and AI revenue beats: $14.1 billion of contracted compute in one quarter re-rated that segment from cash bonfire to bonfire with invoices. More of them, especially with longer non-cancellable terms, attacks the biggest bear argument directly.

4. Starship milestones actually achieved: not announced, achieved. Flight 13's clean sweep mattered. A successful tower catch of the ship is probably the single biggest pure-headline upside event on the board, because it makes the 99% cost claim visible.

5. Supply mechanics running in reverse: short interest at 34% of a small float produced a 40% rally in under two weeks once the float expanded. Squeezes aren't fundamentals, but in year one they're tradable weather and they'll recur around every unlock.

6. Management raising its own bar: Musk pulling the $1 trillion revenue target from 2031 to 2030 on the Q2 call, with a non-zero chance of 2029, is a template. When he raises the bar while the segments are beating, momentum money treats it as confirmation.

What actually makes it go down

The lockup calendar first, because through December it is the most important thing on this list.

The scheduled releases:

• August 6, done: released up to 20% of eligible insider shares, as much as 911.5 million.

• September 9: a fixed 90-day tranche of about 319 million shares.

• August through October: additional percentage tranches every two to four weeks, a steady drip.

• Early November, with Q3 earnings: the single largest release, roughly 28% of the 180-day block, about 1.3 billion shares.

• December 8: full 180-day expiry.

• June 12, 2027: Musk's own 6.4 billion shares unlock.

Until December, every rally runs into scheduled sellers. That's the chart, not a defect in it.

The rest:

• Capex surprises: proven live on August 4, where a $5 billion capex beat outweighed 92% revenue growth for about three days.

• Anything denting the compute demand story: the 90-day cancellation clauses after year-end are the structural weak point. One large CSA cancellation, or credible reporting of one, would hit harder than any rocket failure.

• Starship setbacks: Flight 12's booster loss barely moved the stock. But the company has now publicly tied Falcon's retirement to Starship's reliability, so a serious failure now reads as a threat to the whole roadmap instead of one program.

• Rates and the AI complex: SPCX fell hardest in July when the entire market was repricing AI capex against elevated rates. This is a leveraged expression of that trade whether shareholders want it to be or not.

• ARPU erosion resuming: if $66 breaks meaningfully lower while subscriber growth slows, the profitable core's math degrades and no valuation framework survives it intact.

• Key-man and governance events: 82% voting concentration means market-moving decisions can arrive by post, at any hour, with no board process you can see. The August 22 Falcon post is the small version.

What Elon says, and what his words have historically been worth

Since so much of this stock's daily movement is literally his sentences repricing in real time, his forecast record isn't gossip. It's data, and it's the closest thing to an edge a retail investor has here.

One published scorecard of his major public predictions found roughly a 20% on-time delivery rate, rising to about 45% when partial deliveries count. A separate analysis of 28 completed SpaceX predictions found the timelines consistently ran long.

The rule I use, and I would put it on a sticky note before buying a single share: he is usually right about the what and wrong about the when.

The right-about-the-what half is why this thing is worth two trillion dollars. Falcon booster reuse was widely called impossible and one booster just flew its 37th mission. Starlink was mocked as economically absurd and it has 13 million subscribers at a 39% operating margin. Those were not consensus outcomes. He called them, funded them and delivered them, years late and undeniably.

The wrong-about-the-when half is just as consistent. Starship reached orbit a year behind target. Full reuse has been promised on aggressive timelines for years and still has not happened. Mars windows have slipped serially.

Bears keep betting the outcome never arrives. Bulls keep paying today for the year he named. Both of those are ways to lose money on the same correct observation.

So here's the current promise ledger with dates on it, so you can grade it yourself as it resolves:

• $1 trillion in annual revenue by 2030, pulled forward from 2031 on the Q2 call, with a non-zero chance of 2029. Apply his historical slippage and a mid-2030s arrival is still extraordinary. Paying for 2029 is ignoring twenty years of data.

• Starship flying at least once a day probably a year from now, said August 4. His own operation currently targets roughly one flight a month.

• A tower catch of the ship in a few months and a first ship reflight end of 2026 or early 2027, said in August. These are the nearest-term falsifiable claims on the board and the ones I'd actually watch.

• Winding down Falcon once Starship flies reliably several times per week, posted August 22. Undated and conditional, which by his patterns means real intent on an unknowable clock.

• A million-person Mars colony, which is embedded in his pay. The board tied his largest tranche, 200 million supervoting shares, to a $7.5 trillion market cap plus a permanent Mars settlement of a million people, with a separate 60.4 million share tranche tied to valuation milestones and 100 terawatts of space-based compute.

Read that last one as the company constitution. The pay package tells you the endgame is Mars and orbital compute, and everything between now and then, including your shares, is a funding mechanism.

Now the part I did not enjoy writing.

Yesterday I promised, in public, that my daily market newsletter would publish before the opening bell. It published at 11:10 in the morning because the pipeline broke on the art step and I fixed it live while the market was open. Directionally I did the thing. Chronologically I was two and a half hours late and had to say so in the post.

I spent weeks building a system whose entire job is to catch me being wrong, and it caught me on the exact failure mode I was about to write four hundred words about in somebody else. (In fairness to me, I did not have a rocket to build.)

So when I say his timelines slip, I am not saying it from above. I am saying it as a guy who missed a deadline by 150 minutes on a blog post, watching a man try to land a skyscraper on a tower.

The valuation, in plain numbers

Shares outstanding as of June 30 were roughly 7.6 billion Class A plus 5.6 billion Class B, about 13.2 billion total. At $136.84 that's a market cap near $1.80 trillion. Net of about $100 billion of cash and marketable securities against roughly $39 billion of debt, enterprise value is around $1.74 trillion.

Against what?

Q2 revenue annualizes to about $31 billion. The company says it expects to exit December above $100 billion in ARR, but that number leans on contracted compute, some of it cancellable, so I treat it as a management claim rather than a run rate.

On the annualized Q2 number you're paying about 56 times revenue. On the most generous full-year estimates, somewhere north of 45 times. The company loses money on a GAAP basis so there's no earnings multiple to anchor to, and first-half capex of $28.5 billion means free cash flow is deeply negative by design.

What you have to believe for today's price to be cheap:

• Starlink keeps compounding subscribers with ARPU holding.

• Starshield becomes a durable multi-billion dollar defense franchise.

• The CSA book renews and extends instead of cancelling.

• Starship works well enough, soon enough, to collapse costs across all three segments.

Four engines all need to fire. For the price to be expensive, any one of them stalls.

That asymmetry is why 32 analysts have targets spread from $75 to $800 around an average near $228. A consensus buy rating sitting on top of the widest target dispersion I have ever seen on a mega cap is everyone telling you the same thing in different words. Nobody knows.

The calendar from here

• Around September 15: Starship Flight 14. Watch booster performance after Flight 12's loss.

• September 9, September 24, October 9, October 24: scheduled lockup tranches, a few hundred million shares of drip supply around each.

• Early November, likely the 3rd through the 5th: Q3 earnings. Capex, ARPU, the CSA book, and the roughly 1.3 billion share unlock the report itself triggers. The most loaded date on the board.

• December 8: full 180-day expiry. After this the scheduled supply overhang is done until Musk's shares unlock.

• Ongoing: the Cursor close, the Starship catch attempt, Starshield and airline contracts, and whatever gets posted at 2 in the morning.

What I'm actually going to do

Trading this daily since June, while carrying a longer-horizon position in it, has narrowed what I actually watch down to two short lists.

For the longer-horizon side, two numbers carry more weight than everything else in the quarterly report, and I put them side by side every time: sequential ARPU and sequential capex.

If ARPU holds while capex growth slows, the profitable core is winning and weakness is worth buying into. If ARPU slips while capex accelerates, the furnace is winning, and shareholders are the fuel.

Everything else in the release is texture. Those two decide whether the profitable core is outrunning the spending or the other way round.

For the swing side it is the supply calendar above, and nothing else comes close until December. Scheduled sellers on a known schedule are the most predictable thing about this stock right now, which is exactly what makes them tradable. And on the man himself: trade the direction, fade the dates. That framework has been right for twenty years and the first year of SPCX has not given me a reason to change it.

The daily version of this, where setups get published and graded the next morning, runs every weekday in The Morning Bell, and the system that produces it is a Claude scheduled task if you want the mechanics. If you want one built for your own business, reach out directly.

So a real question for anyone else in this name: what is the macro print or headline that would actually make you cut, and have you written it down anywhere? Mine is the rate path, because every one of the four engines above gets repriced by it, and a thesis with no invalidator is just a feeling.

One more time for the people in the back: not financial advice, education only, based on public filings and on what I may or may not be doing personally.

Sources:

• SpaceX Reports Second Quarter 2026 Results, U.S. Securities and Exchange Commission

• SpaceX IPO takeaways: SPCX closes at $161, jumping 19% after record debut, CNBC

• SpaceX short sellers are running out of bullets as stock rebounds more than 40% off low, CNBC

• The world’s most valuable IPO, SPCX, is now Wall Street’s most shorted new stock, Invezz

• Elon Musk delivers totally nuts plan for moon robots and $1 trillion revenue target, but capex tanks SpaceX on debut earnings, Fortune

• Analysis: SpaceX IPO gives Musk sweeping power and curbs shareholder rights, Reuters

• Elon Musk’s pay package reveals what SpaceX actually is, Fortune

• SpaceX SPCX Lock-Up Expiration Dates: Employee Stock Release Schedule 2026, StockAlarm

• FAA requires SpaceX-led mishap investigation before resumption of Starship launches, Spaceflight Now

• SpaceX scales back plans for next Starship launch, SpaceNews

• The Musk Prediction Scorecard: Twenty Promises, Exposed to the Record, Tesorb

• 6 Charts on SpaceX’s Pre-IPO Financials, Morningstar