On August 5, SpaceX stock closed at $108.27, a new all-time low, roughly 20 percent below its $135 IPO price and more than 50 percent off its June peak. The next morning, the largest single lockup expiration in the company’s brief public history — roughly 911.5 million shares, more than SpaceX sold in its entire June IPO — became eligible for sale. Every structural argument this publication has made about SPCX since July said this was the moment the stock should break.
Instead, it closed up 6.1 percent that day. It kept climbing. By August 12, SPCX had reached $146.15, a rally of roughly 35 to 40 percent off its low, briefly reclaiming the $135 IPO price it had spent a month trading below. Short interest — which had sat near 34 percent of the float, the largest short position against any major new listing this year, larger in dollar terms than the bet against Tesla — collapsed to roughly 11 percent in the space of a week. S3 Partners’ Ihor Dusaniwsky put it plainly: “Shorts that wanted to short are out of bullets.”
This is the resolution this series has been building toward since the pre-inclusion rally first reversed in early July. It did not resolve the way the mechanics alone predicted. Understanding why requires walking through what actually happened on August 4, and what the lockup ran into that nobody — bulls, bears, or this publication — had fully priced.
What SpaceX Actually Reported
SpaceX’s first-ever public earnings report, released August 4, beat consensus by a wide margin on the top line. Revenue came in at $7.8 billion for the quarter, up 92 percent year-over-year, against a consensus estimate near $6.9 billion — a beat of nearly $1 billion. Net loss narrowed to $541 million, or $0.09 per share, well inside the roughly $0.23 to $0.24 loss analysts had modeled and a sharp improvement from a loss near $970 million to $1 billion a year earlier. Adjusted EBITDA reached $3.5 billion, up 191 percent year-over-year, and operating loss narrowed to $143 million from $970 million.
The segment breakdown is where the quarter’s real story lives. Starlink — reported as the Connectivity segment — generated $4.3 billion, up 66 percent year-over-year and ahead of the roughly $3.83 billion analysts expected. Subscribers reached 12 million, doubling year-over-year, though average revenue per user fell 22 percent as the company expanded into international markets with cheaper plans. Launch services, reported as the Space segment, brought in $962 million, up 29 percent year-over-year and 55 percent quarter-over-quarter, but posted a $542 million operating loss reflecting heavy Starship development spending across 78 total launches and 1,041 tons delivered to orbit in the first half of the year.
The segment that reframed the entire valuation conversation was neither of those. SpaceX’s AI compute business — the SpaceXAI operation this series has tracked since the Colossus data centers first entered the picture — generated $2.6 billion in revenue, up 247 percent year-over-year, with new cloud-compute agreements with Google and Anthropic contributing $1.6 billion of that growth within the quarter alone. The AI segment’s adjusted EBITDA turned positive at $1.146 billion. Nameplate compute capacity reached 1.4 gigawatts, up from 1.0 gigawatt in the first quarter, with management targeting 2 gigawatts by year-end and 10 to 15 gigawatts by the end of 2027, running exclusively on Nvidia hardware. CFO Bret Johnsen told analysts the company is on pace for a $100 billion annualized revenue run rate by year-end. JPMorgan, in its post-earnings note, flagged a path to $100 billion in AI revenue alone in 2027 — not total company revenue, AI revenue specifically.
The Number That Actually Moved the Stock Down
None of the revenue beat is what sent SPCX to a new all-time low. Capital expenditure did. SpaceX spent $18.4 billion in the quarter, up from $10.1 billion the prior quarter, with $15.8 billion of that directed at AI infrastructure buildout. A company whose stock had already spent a month sliding on lockup-supply anxiety now had to defend a near-doubling of quarterly capex against a market that had grown broadly warier of open-ended AI infrastructure commitments across the entire technology sector through the summer.
The reaction was immediate and severe. SPCX fell as much as 8 percent in after-hours trading the night of August 4. By the afternoon of August 5, it was down 13 percent intraday, trading near $109, and closed the session at $108.27 — the new all-time low, set one trading day before the lockup opened. Elon Musk’s personal net worth reportedly fell by roughly $80 billion on paper that single session, a reminder of how much of his wealth remains concentrated in a stock still finding its footing four months after listing.
Analyst reaction split along a predictable line. JPMorgan raised its price target to $240 from $225, explicitly citing the company’s “extreme vertical integration” and the AI-revenue trajectory toward $100 billion. Wells Fargo cut its target to $215 from $230, citing capex intensity and lingering questions about AI monetization durability. BofA’s Ronald Epstein captured the split most directly: a genuine beat on revenue and profitability, undercut by a capex figure and monetization uncertainty large enough to explain the stock’s continued weakness even after strong results. The market, in the 24 hours between the report and the lockup, was pricing a capex story, not a revenue story.
The Lockup, and What It Ran Into
The mechanics were exactly as this series had mapped them. On August 6, roughly 911.5 million shares — more than the entire 639 million shares SpaceX sold in its June IPO, and something on the order of 7 percent of total shares outstanding — became eligible for sale simultaneously, the largest single unlock event in the company’s short public history. The lockup arithmetic this publication laid out on July 20 was not wrong about the scale of the event. What it could not have known in advance was what condition the stock, and the surrounding narrative, would be in when that supply actually arrived.
The stock had already found its low the day before. On the day the largest supply shock in its trading history landed, SPCX opened weak, dipped as low as roughly $105.11 intraday, and then closed up 6.1 percent. It kept rising in the sessions that followed: $133.11 by August 7, holding near $133.29 through August 11, then a 3.23 percent gain to $146.15 on August 12 — the strongest close of the entire post-IPO period since the pre-inclusion rally. A modest pullback to roughly $142.31 followed on August 13, but the trajectory from the August 5 low to the August 12 high represents a 35 to 40 percent rally in a single week, executed while the largest tranche of sellable shares in the stock’s history sat in the market.
Elon Musk’s own holdings — roughly 6.4 billion shares — remain locked until June 2027 and were not part of this tranche; he did not sell. Broader insider selling activity was reported anecdotally rather than aggregated: one early SpaceX employee, a former NFL player who had purchased a stake in 2022 for roughly $150,000, told CNBC he intended to sell his entire position. No comprehensive tally of total insider selling into the unlock has been disclosed. A second lockup tranche, roughly 319 million shares, is understood to follow in the coming weeks, with further staggered unlocks through September and October — the total lockup schedule remains a live, multi-month story, not a single resolved event.
Why the Facebook Path, Not the Snap Path
The July 20 analysis in this series laid out two precedents for how a large-float unlock into a weak stock can resolve. Facebook’s November 2012 unlock — a comparable scale event against a then-depressed post-IPO price — saw the stock rise on the unlock day itself, because the event was fully priced in and the company’s own earnings trajectory had, by that point, given institutions a reason to hold rather than sell into the supply. Snap’s 2017 unlock ran the opposite way: insider selling met no institutional bid, and the stock extended its decline for months. The determining variable in both cases was not the mechanics of the unlock. It was whether the company’s most recent earnings report gave the market a reason to want more of the stock, not less, at the moment supply arrived.
SpaceX lived the Facebook path, and the August 4 earnings report is the reason. Every previous article in this series treated the lockup as a pure supply event running against a stock whose demand case rested on rocket launches, satellite subscriptions, and an unresolved question about whether Starship’s development costs would ever be justified by commercial return. That demand case, on its own, was thin enough that a 900-million-share supply shock looked close to fatal. What actually arrived on August 4 was a different demand case entirely: a $2.6 billion AI-compute segment growing 247 percent year-over-year, with named hyperscale customers, a positive EBITDA margin, and a credible path to $100 billion in standalone AI revenue by 2027. Investors who had never been interested in owning a satellite-and-launch company found, four days before the largest lockup event of the year, a reason to want exposure to a vertically integrated AI infrastructure company that happened to also build rockets. That is a different buyer than the one who bid SPCX up ahead of its Nasdaq-100 inclusion in July, and it is the buyer who absorbed the lockup.
Short sellers had positioned for the mechanical scenario — supply overwhelming a stock with a thin, listing-psychology-driven demand base, regardless of what the earnings report contained. That was a reasonable bet through most of July, when the demand base genuinely was thin and mechanically driven. It stopped being reasonable the moment the earnings report reframed what the stock was a claim on. A short position sized for a satellite company facing a supply glut does not survive contact with a stock the market has just decided to reprice as an AI infrastructure play, and the speed of the short-interest collapse — from roughly 34 percent of float to roughly 11 percent in about a week — is the market’s own record of how fast that repricing happened.
The Analyst Consensus This Series Tracked, Revisited
It is worth returning to the analyst dispersion this series flagged as unusually wide back in July, because the August 4 report is the first genuine test of which camp had the better model. At the time, coverage ranged from CFRA’s Sell rating and a Street-low target near $115 to Raymond James’s Street-high target near $800 — a spread wide enough, this publication argued at the time, to indicate the analysts were not disagreeing about a number so much as disagreeing about what kind of company SpaceX actually was.
The August 4 disclosures resolved a meaningful part of that disagreement, though not all of it. The bears who priced SPCX as a 49-times-revenue launch-and-satellite business with a decade of insider liquidity about to hit the market were correct about the lockup’s scale and correct that capex was a legitimate risk — Wells Fargo’s post-earnings cut reflects that camp’s model essentially intact. The bulls who priced in a vertically integrated AI-compute conglomerate got direct confirmation: a $2.6 billion AI segment growing 247 percent, with a credible path to $100 billion in AI revenue alone by 2027, is close to the thesis Raymond James and Macquarie had been running since the post-quiet-period initiations. JPMorgan’s target increase to $240, still well below both extremes, represents something closer to a synthesis — acknowledging the AI upside while keeping the capex risk and monetization-durability questions in the model rather than dismissing them. No single analyst call from July was fully vindicated. The market’s actual behavior — a sharp initial selloff on capex, followed by a sharp recovery on AI-segment evidence — looks, in retrospect, like the market itself working through the same bull-bear tension the analyst spread had been describing for a month, compressed into a single trading week.
What the Squeeze Does and Does Not Prove
It is worth being precise about what a short-interest collapse of this speed and size does and does not establish, because the mainstream financial press has, in the days since, moved quickly to “trillion-dollar short squeeze” framing that overstates the certainty of what happened.
What it establishes: the market’s aggregate judgment, formed rapidly after August 4, is that SpaceX’s AI-compute business materially changes the company’s earnings trajectory, enough to outweigh a near-doubling of quarterly capital spending and the largest lockup event in the stock’s history occurring in the same week. That is a real, market-tested conclusion, not a speculative one — it was tested against genuine, simultaneous downward pressure from both a capex scare and a supply shock, and it held.
What it does not establish: that the capex concern was wrong, that AI-segment revenue at this growth rate is durable, or that the remaining lockup tranches through September and October will be absorbed as cleanly as the first one was. The AI segment’s growth in this quarter was substantially driven by two named customers — Google and Anthropic — whose future compute demand and contract renewal terms are not disclosed in any way this publication can verify. A business generating $2.6 billion from two large customers carries meaningfully more concentration risk than a business generating the same revenue from a diversified base, and concentration risk of that kind has ended AI-infrastructure growth stories abruptly before, in other companies, when a single customer’s spending plans shifted. Wells Fargo’s caution — a price-target cut on the same day JPMorgan raised its target — reflects a genuine, unresolved disagreement about how much of this quarter’s result is durable versus how much reflects the timing of two large contracts landing in the same three months.
The Mechanics That Made the Squeeze Possible
The speed of the short-interest collapse deserves its own explanation, because a squeeze of this size in a single week is not simply a story of shorts changing their minds. It is a story of mechanical forced buying interacting with a thin float in exactly the way this series has described since the stock’s earliest sessions.
A short position is, structurally, a borrowed-share sale with an open-ended obligation to buy the shares back eventually. When a stock a fund is short begins rising sharply, the position generates a mark-to-market loss that grows every session the rally continues, and at some threshold — set by the fund’s own risk limits, margin requirements from its prime broker, or simple unwillingness to keep bleeding capital into a name moving against it — the fund is compelled to buy back its borrowed shares regardless of what it believes about the stock’s long-term value. That buying-to-cover is itself additional demand pressing against the same thin float that made the original short position so large a percentage of shares outstanding in the first place. In a stock with a normal, deep float, forced short-covering of this kind barely registers in the price. In a stock where short interest reached roughly a third of all tradable shares, the covering itself becomes a significant fraction of the buying volume that drove the rally — which is precisely why a 35 to 40 percent move in a single week, on top of an earnings report and a lockup, is plausible without requiring an implausibly large amount of genuine new long-term capital to have arrived.
This dynamic cuts against the shorts in a way worth stating plainly: the same float thinness that made SPCX vulnerable to a mechanical demand collapse in July — the listing-psychology mechanism this series opened with — made it equally vulnerable to a mechanical demand surge in August, once the direction of forced buying and selling flipped. Thin floats amplify whichever mechanical flow is currently dominant. In July, that flow was index-inclusion buyers exhausting themselves, then sellers with nowhere to absorb their supply. In August, it was short sellers forced to buy back borrowed shares into a stock the market had just repriced upward. The float structure did not become more forgiving between those two episodes. The direction of the mechanical pressure simply reversed, and a float this thin transmits either direction with equal violence.
The Calendar Is Not Finished
Two further events keep this story open rather than closed. The second lockup tranche, roughly 319 million shares by the schedules currently circulating, is expected in the coming weeks, with additional tranches following through September and October — a testing ground for whether the demand shift this earnings report produced is durable enough to absorb repeated supply events, or whether the first unlock simply landed at an unusually favorable moment in the news cycle that will not repeat.
Separately, Musk confirmed on the earnings call that Starship Flight 14 is targeted for the end of August, from Starbase — the first attempt at reaching orbital velocity, the first attempt at catching the ship itself with the launch tower rather than a splashdown, and the first deployment of operational-generation Starlink V3 satellites into a stable orbit rather than the suborbital demonstration profile Flight 13 flew in July. A successful Flight 14 would be the first Starship mission in the program’s history to combine a genuine technical milestone with a company that the market has, in the two weeks since its first earnings report, begun pricing as an AI infrastructure business as much as an aerospace one. A second consecutive booster failure, or a scrubbed catch attempt, would test whether that AI-driven demand base is resilient to setbacks in the rocket program that gives the company its name and its remaining execution risk.
What the Series Got Right, and What It Had to Learn
The listing-psychology thesis this series opened with in early July has held up on its own terms. Mechanical demand ahead of Nasdaq-100 inclusion inflated SPCX above what fundamental buyers alone would have supported; that demand exhausted on schedule; the stock corrected through its IPO price and continued lower, exactly as the mechanics predicted, all the way to a floor. Each of those steps ran on the timeline the framework anticipated, sometimes faster.
What the framework could not anticipate, because it was not yet disclosed, was the composition of the business underneath the stock price. A pure listing-mechanics analysis treats a company’s demand base as roughly fixed and asks how mechanical buying and selling moves price around that fixed base. SpaceX’s August 4 report demonstrated that the underlying business itself was not fixed — it had, over the same months this series was tracking IPO mechanics, quietly become a materially different company, with an AI-compute segment large enough and fast-growing enough to constitute a second, independent investment thesis layered on top of the original rockets-and-satellites one. The lockup did not fail to depress the stock because the mechanics were wrong. It failed because the stock, by the time the mechanics ran their course, was no longer a claim on the same company the mechanics had been describing.
That is the finding this series closes on, provisionally: SPCX’s post-IPO trajectory was governed by listing psychology through July, and has been governed by a genuine reassessment of the underlying business since August 4. Both can be true in sequence. The next test — the second lockup tranche, and Flight 14’s outcome — will show whether the market’s new read on SpaceX holds, or whether August 12’s peak turns out to be its own kind of overshoot, waiting for a mechanical correction of a different sort.
The watch list into September, in order of likely consequence: whether the second lockup tranche, expected in the coming weeks, produces anything resembling the volatility of the first, or is absorbed quietly by a market that has already recalibrated; whether Starship Flight 14 succeeds at both the orbital-velocity milestone and the tower-catch attempt, or hands the bears a fresh execution-risk argument at the precise moment the AI-revenue thesis needs continued credibility; and whether the Google and Anthropic compute agreements that drove this quarter’s AI-segment growth show any sign of expansion, plateau, or renewal risk in the next disclosed period. A stock that moved 40 percent in a week on the strength of one earnings report can move just as far in the other direction on the strength of the next one — the mechanics that produced August’s squeeze are, by their nature, symmetric, and this series will be watching for which direction they run next.
