SPCX Is Above Its IPO Price Again. 319 Million Shares Land Next Week, and I'm Still Not Buying

TL;DR

• SPCX closed August 12 at $146.15, up 9.65%, its second close above the $135 IPO price in three sessions. It closed below the IPO price in between, at $133.29 on August 11. This is a range, not a trend.
• The move came on a Morgan Stanley reiteration ($75 bear, $300 base, $600 bull) and a routine Starlink launch. No new financial information arrived. A 9.65% session on a maintained price target tells you what the positioning looks like.
• The AI segment tripled quarter on quarter, from $818m to $2,561m, and is now 33% of revenue against 17% three months earlier. It also consumed $15.83bn of capex and lost $1,257m from operations.
• Cursor is being paid for in SpaceX stock at a 7-day VWAP struck just before closing. At the August 5 low that bill was about 554 million shares; at Wednesday's close it is about 411 million. The rally has already cut it by roughly 144 million shares.
• My answer is no at $140. The trigger I want is the day-70 tranche coming and going with the stock holding $133.29. Below that I would rather wait for the $120s.

The Board

Six sessions, an average absolute daily move of 8%, and a share count that moves with the price.

Why Is SPCX Above Its IPO Price Again?

Morgan Stanley published a note on August 12 keeping its $300 target and arguing the market is undervaluing SpaceX's AI business, and SpaceX launched 24 more Starlink satellites from Vandenberg the night before. The stock rose 9.65% to $146.15.

Neither of those is new information about the company's earnings power. The Morgan Stanley note reiterated a target the bank already held, and Starlink launches happen constantly. What moved was the interpretation: at $139 the bank reckoned the market was valuing the AI business at roughly $12 a share, which it called low against listed AI infrastructure comparables, and it framed a $600 bull case conditional on the Cursor acquisition delivering.

The tape underneath is worth labelling properly, because the round-trip has been messy. August 10 was the first close above $135 since July 15, at $138.74. August 11 gave it back, closing $133.29. August 12 took it back with interest at $146.15. As I write on Thursday morning, August 13, the stock is around $140.75, down about 3.7%. Average absolute daily move across those five sessions: 7.96%.

That is a stock being repriced by argument rather than by disclosure, and arguments are cheap to reverse.

What Cursor Actually Costs, in Shares

Here is the piece I have not seen run anywhere, and it changes how the supply calendar reads.

SpaceX is buying Anysphere, which makes the AI coding tool Cursor, at an implied equity value of $60 billion. The consideration is SpaceX Class A stock, and the number of shares is set by SpaceX's seven-day volume-weighted average price immediately before closing. The deal is expected to close in Q3 2026, subject to regulatory approvals.

So the share count is a function of where SPCX trades in September, and the arithmetic runs the opposite way to intuition:

• At the August 5 closing low of $108.27, $60bn is about 554 million shares. That is very nearly the entire 555.6 million share base offering that created the public float in June.
• At the August 12 close of $146.15, it is about 411 million shares.
• The difference is roughly 144 million shares that SpaceX no longer has to hand over, purely because the stock went up.

Both figures need their denominator, because 411 million means two completely different things depending on what you divide it by. Against 13.18 billion shares outstanding it is 3.1%, which is trivial. Against the 1.55 billion shares that are actually tradeable today it is 26.5%, which is not.

Which denominator applies depends on whether the Anysphere shares are registered and sellable on closing or sit restricted for a period. I could not source that from the merger terms, and it matters more than anything else in this section, so I am flagging it as an open question rather than filling it in. If those shares are free to trade in Q3, the float maths for September and October are materially worse than the published lock-up staircase implies.

The AI Segment Tripled. It Also Ate $15.8 Billion.

The Q2 numbers are better than the share price behaviour suggests, and the capex is worse.

Revenue was $7.81bn, up 92% from $4.07bn, against about $6.93bn expected. The three segments, for the quarter ended June 30:

Segment	Q2 2026	Q1 2026	Q2 2025	Operating result
Connectivity (Starlink)	$4,291m	$3,257m	$2,588m	+$1,656m
AI	$2,561m	$818m	$737m	−$1,257m
Space (launch)	$962m	$619m	$746m	−$542m

The AI line is the one to sit with. Within it, "AI solutions and infrastructure" went from $475m in Q1 to $2,194m in Q2, a 362% quarterly step, and nameplate compute reached 1.4GW. Add $14.1bn of contracted cloud services sales signed in the quarter and the segment stops looking like a Musk side project.

Then read the cash. SpaceX spent $18.37bn of capex in the quarter, about 2.35 times revenue, and $15.83bn of that went into AI infrastructure. So the AI segment spent roughly six times its own quarterly revenue on capital, and still lost $1.26bn at the operating line. That is the actual reason the market prices this business at $12 a share rather than the $50 Morgan Stanley thinks it is worth, and I do not think the market is being stupid about it. It is being asked to underwrite a build.

Connectivity, meanwhile, is carrying the company: 12.0 million Starlink subscribers, doubled year on year, throwing off $1.66bn of quarterly operating income. The blemish there is price. ARPU fell from $85 to $66, down 22%, as SpaceX pushed into cheaper international plans. Subscriber growth is real and it is partly being bought.

What Happens Next, by Date

• Around August 20: the day-70 tranche, roughly 319 million shares, per the full lock-up staircase.
• Around September 10 and September 25: two more 7% tranches.
• September: the index rebalance, where a bigger float should mean a bigger Nasdaq-100 weight. That bid has still not arrived.
• Q3, date unconfirmed: Cursor closes, and roughly 411 million Class A shares are issued at whatever the VWAP says.
• October 10 and October 25: two more tranches. Q3 earnings: about 28%. December 8: the remaining 180-day shares.
• June 2027: Musk and selected insiders unlock.

Consensus sits a long way above all of this. 35 analysts average a $231.40 target, and the stock trades at roughly 80 times trailing sales at $140.75.

Would I Buy It Here?

No. The stock is 30% off its low in five sessions on an analyst reiteration, into a dated supply event five sessions away, in a name delivering 8% daily swings. Every part of that sentence argues for waiting rather than for pessimism.

The bull case is genuinely stronger than it was a month ago, and I want to say so plainly: the unlock that was supposed to break this stock produced a green candle, which I did not expect, the AI segment is compounding faster than the multiple assumes, and the Cursor bill shrinks as the stock rises. That last one is a real reflexive support that the bears, including this site, have been ignoring.

What I want before paying up is evidence that supply gets absorbed at a higher level than last time. The day-70 tranche around August 20 is that test. If SPCX takes 319 million shares and still holds $133.29, the August 11 low of this range, then the pattern is established across two tranches at rising prices and I would rather own it than watch it. If it breaks that, the next level I care about is the $120s, and I would be happy to wait there through the September tranches.

What would make me wrong: Q3 AI revenue doing again what Q2 did. At a 362% quarterly step in AI infrastructure revenue, 80 times sales stops being the relevant number very quickly, and no amount of entry discipline would save me from missing it.

The Options Angle

I could not source a live SPCX chain or a current implied move, so I am not going to put a number on the weekly straddle. What I can quote is realised, and it is the more useful figure here anyway: 7.96% average absolute daily move over the last five sessions, with individual days of −13.6%, +15.8% and +9.7%.

That kills the reflex play. The setup looks made for selling premium into a dated event, and this site's July scorecard is a long record of that reflex failing in exactly this regime. A cash-secured put at $130 into the August 20 tranche would collect a good premium and get run over by one 13% session. When realised is printing 8% a day, the conservative-looking side of this trade is the leveraged one.

The options structure I would actually want is a call spread bought after the tranche clears. Buying volatility ahead of a known date in a name already moving 8% a day means paying for a move the stock is delivering for free.

Trade log

#	Stance	Structure	Strikes and expiry	Cost or credit	Spot at writing	Implied move	Breakeven
1	Pass	Long shares	Spot entry, no expiry	n/a	$140.75, Aug 13, 11:01am ET	Not sourced; realised 7.96%/day over 5 sessions	n/a
2	Bullish, conditional	Long shares	Triggers only on a close above $133.29 after the ~Aug 20 tranche	n/a	$140.75, Aug 13, 11:01am ET	Same	Trigger level $133.29
3	Pass	Short cash-secured put	~$130 strike, first expiry after Aug 21	Premium not sourced	$140.75, Aug 13, 11:01am ET	Same	~$130 less premium

Row 1 gets scored against where SPCX trades once the day-70 tranche has cleared. Row 3 is the one most likely to embarrass me if the tranche passes without incident, and it is logged precisely so that it can.

The One-Line Read

SpaceX is above its IPO price on an analyst's opinion rather than a new number, with 319 million unlock shares due next week and 411 million more owed for Cursor. I want the next tranche absorbed above $133 first.

Related

• Is SPCX a buy before the August unlock?, the decision hub this piece updates
• The full lock-up calendar through December 8
• SpaceX's first earnings report as a public company