SpaceX stock closed at $146.23 on August 17, up 5.66 percent on the day, extending a rally that had carried it more than 30 percent off its early-August low. Two sessions later, it closed at $139.65, down between 2.6 and 3.1 percent depending on the source, after opening weak and drifting through a session compressed between $136.72 and $145.01. Today, Thursday August 20, roughly 319 million shares — the second tranche in SpaceX’s staggered post-IPO lockup schedule — become eligible for sale.
The stock fell into this event the same way it fell into its first one. Two weeks ago, this series documented what happened when 911.5 million shares came free on August 6: SpaceX had crashed to a fresh all-time low the day before, on capex-driven earnings jitters, and the market’s working assumption was that the largest single unlock in the company’s history would compound the damage. Instead, the stock closed up 6.1 percent that day and kept climbing, short interest collapsing from roughly 34 percent of the float to roughly 11 percent inside a week. The pattern-match to today is obvious enough that it barely needs stating: sliding price, lockup date, a market bracing for supply pressure. What is less obvious, and worth working through carefully, is how many of the specific conditions that produced the August 6 squeeze are actually present again.
A Smaller Tranche, Arriving on Larger Gains
The most basic difference is scale. August 6 released approximately 911.5 million shares — more than the entire float SpaceX sold in its June IPO. Today’s tranche, at roughly 319 million shares, is smaller by a factor of nearly three. Depending on which share count is used as the denominator, that represents somewhere between 2.4 percent of SpaceX’s basic share count and a larger fraction when measured against the free-trading float specifically — sources diverge on the precise percentage, and that divergence is itself worth flagging rather than papering over with a single confident number. What is not in dispute is the direction: today’s supply event is meaningfully smaller than the one the stock already absorbed.
A smaller supply shock landing on a stronger price base cuts in a specific, less-discussed direction. The shares eligible to sell today were, by definition, held through the entire run from the August 5 low near $108 to the August 17 peak near $146 — a gain of roughly 35 percent for anyone who held the full stretch, and a considerably larger one for anyone whose cost basis predates the IPO entirely. Analysts covering the stock have described this as a “sterner test” than August 6 for exactly that reason: holders facing today’s unlock are sitting on substantially larger unrealized profits than holders faced at the first unlock, when the stock itself was near its lowest point since listing. Larger embedded gains create a stronger, not weaker, incentive to realize some of them, which is the opposite of what a naive “smaller tranche, smaller impact” reading would suggest. The size of the event has shrunk. The incentive to sell into it may not have.
A Named Bear Case, Timed to the Date
The analyst dispersion this series has tracked since the post-quiet-period initiations — a spread this publication has repeatedly noted runs unusually wide, from outright sell ratings to price targets implying multiples of the current price — sharpened into a specific, dated argument this week rather than remaining a background statistic.
Phillip Securities analyst Glenn Thum reiterated a Sell rating with a $75 price target, implying roughly 46 percent downside from current levels, and explicitly timed the call to this week’s unlock. Thum’s core argument is not primarily about the lockup mechanics at all — it is about the durability of the revenue segment that drove the August 6 squeeze in the first place. The AI-compute business that reframed SpaceX’s valuation story after its first earnings report, Thum argues, is the least durable part of the company’s revenue base, resting on cloud-compute agreements that carry roughly 90-day cancellation notice periods. That is a materially different critique from the capex-anxiety selloff that preceded the first unlock. It does not dispute that the AI segment is real or growing; it disputes that its customers are committed for long enough to justify pricing the stock as if that revenue were as durable as the launch and satellite businesses SpaceX has run for two decades.
The bull case remains intact and well-represented against that argument. Clear Street’s Brian Dobson reiterated a Buy rating with a $217 target; UBS’s John Hodulik reiterated Buy at $210. Consensus across covering analysts sits at Moderate Buy, with roughly 28 buy ratings against two sells and an average target in the low-to-mid $200s — upside north of 60 percent from current levels, if the average holds any predictive value at all against a spread this wide. The dispersion this series flagged as unusually stark in July has not narrowed. If anything, Thum’s specifically-timed Sell call and the bulls’ unmoved triple-digit targets illustrate how far apart the two camps remain even after a full earnings cycle and a completed squeeze that, on its face, should have given the market more information to converge on.
What the Analyst Split Reveals About the Company, Not Just the Stock
It is tempting to read a 28-buy, two-sell analyst spread with a $75-to-$217 target range as simple noise — a sign that Wall Street has not done its homework, or that SpaceX is too new a public company for coverage to have matured. A more useful reading treats the dispersion as a direct reflection of a genuine structural ambiguity in what SpaceX has become since its June listing.
The company that priced its IPO was, on paper, a launch-services and satellite-broadband business with a well-understood, if capital-intensive, cost structure and a couple of decades of operating history to model against. The company reporting earnings six weeks later disclosed an AI-compute segment growing at a rate, and a scale, that had no comparable operating history to model at all — a business unit that went from a rounding error to $2.6 billion in quarterly revenue inside a single fiscal year, built on customer contracts whose renewal economics are simply unknown to outside analysts. Valuing the first kind of company is a familiar exercise with familiar comparables — other launch providers, other satellite operators. Valuing the second kind, bolted onto the first inside the same ticker, requires analysts to essentially build two separate models and then decide how much weight each deserves, with no established convention for doing so. A $75 target and a $217 target are not necessarily evidence that one analyst is right and the other badly wrong; they may equally reflect two analysts assigning defensible but very different weights to a genuinely unprecedented corporate structure, in the absence of enough quarters of AI-segment data to discipline the disagreement.
The Question This Series Cannot Answer Cleanly
The single most important variable in whether today repeats the August 6 pattern is also the one this research could not pin down with confidence, and it deserves to be stated that way rather than forced into false certainty: nobody has produced a clean, dated read on where short interest actually stands heading into this unlock.
The headline figure from the August 6 story — short interest near 34 percent of float in late July, collapsing to roughly 11 percent within a week of the squeeze — is well documented and was central to explaining why that unlock produced a rally rather than a crash: the mechanical buying-to-cover from a short base that size, forced into a thin float, did much of the work. What has happened to short positioning in the three weeks since is genuinely unclear from available reporting. FINRA’s official short-interest settlement reports lag by roughly two weeks and the most recent readable figures span a range from around 110 million shares (measured against total shares outstanding, a very different and much smaller-looking denominator) to figures closer to 165 million shares in less official trackers — numbers that are not directly comparable to each other, let alone to the pre-squeeze 219-million-share figure from late July. Whether traders have rebuilt meaningful short positions specifically targeting today’s tranche, or whether the post-squeeze retreat left the short base too depleted to matter this time, is an open question this piece cannot resolve with the data currently public. It is worth watching the next FINRA settlement report, expected within roughly two weeks, as the first hard data point that will actually answer it.
Why “Eligible to Sell” Is Not the Same Question as “Will Sell”
It is worth pausing on a distinction that has run through every lockup piece in this series and that matters as much today as it did on August 6: a lockup expiration is a change in what is legally permitted, not a forecast of what will actually happen. The 319 million shares becoming tradeable today do not become sell orders by default. They become eligible sell orders, held by a specific, identifiable population of shareholders — early employees, venture investors from SpaceX’s pre-IPO funding rounds, and participants in secondary-market tender offers the company ran in the years before listing — each facing their own tax situation, liquidity needs, and view on where the stock goes from here.
That population’s behavior is genuinely harder to model than the mechanical short-covering that drove August 6, because it depends on individual decisions rather than a forced, rules-based trigger. A short seller facing a margin call has limited discretion once the loss crosses a threshold; a long-tenured employee sitting on a decade of paper wealth has considerably more latitude to time a sale around tax-year planning, personal liquidity needs, or simply a view that the stock has more room to run. Multiple outlets covering this week’s unlock have explicitly framed it this way — as a test of revealed preference among holders who, unlike the shorts in the first unlock, are not being compelled to act by anything except their own judgment. That framing cuts against easy prediction in either direction. A wave of selling would say more about this specific holder population’s risk tolerance after a 35 percent rally than about the company’s prospects; an absence of selling would say more about conviction than about any legal or mechanical constraint being lifted, since the constraint is, as of today, already gone.
SpaceX Went Looking for an AI Coding Company, and Got a Public No
A separate story landed in the same window that bears on the same underlying question the lockup and the Thum bear case both raise: how far, and how fast, is SpaceX trying to extend its AI ambitions beyond compute rental into software itself.
Bloomberg reported August 19 that SpaceX had approached Cognition, the AI coding startup, about a potential acquisition. Cognition’s chief executive, Scott Wu, denied the report publicly the same day, stating flatly that the company was “not for sale.” The denial resolves the immediate question — Cognition is not being acquired, at least not on any terms it has agreed to — but the report itself, landing the same week SpaceX disclosed $18.37 billion in quarterly capital expenditure, is a data point about appetite rather than about outcome. A company already defending a capex figure that grew more than sixfold year-over-year, from roughly $2.83 billion to $15.8 billion on the AI-infrastructure line alone, reportedly probing an acquisition in adjacent AI software territory is not the profile of a company pulling back to consolidate its existing bets. It is the profile of a company still actively expanding the scope of what it is trying to become, even as analysts like Thum argue the market has not yet properly priced the risk in what it has already built.
The fuller capex picture, disclosed in follow-up coverage this week, sharpens the stakes on both sides of the debate. Revenue for the quarter rose 92 percent year-over-year to $7.8 billion, and the net loss narrowed substantially, from roughly $1.01 billion a year earlier to $541 million — genuine operating improvement, not a story of losses widening to match the spending. Active AI data-center capacity, the company has said, is targeted to grow from 1.4 gigawatts currently to 10 gigawatts within the next year, a scale-up that would require capital expenditure at or above the current elevated run rate for the foreseeable future, not a one-time step-up that moderates from here. Whether that spending path resolves toward the bull case — durable AI revenue justifying durable AI infrastructure — or the bear case Thum has articulated — cancelable contracts funding infrastructure with a multi-year payback horizon — is not a question this week’s disclosures answer. It is the question the entire remaining lockup schedule, running in staggered tranches through the fall, will keep testing.
China Landed a Rocket the Same Week
One further story, smaller in market impact but larger in what it represents, landed the same day as Tuesday’s selloff and deserves inclusion rather than omission. A Chinese launch company successfully landed a reusable rocket booster on August 19 — the first such achievement outside SpaceX’s own program. Coverage tied the news directly to Tuesday’s SPCX weakness, treating it as competitive pressure on the core launch business rather than the AI-compute story that has dominated this series’ recent chapters.
The near-term financial impact of a single successful landing by a foreign competitor is close to negligible — reusable-booster technology took SpaceX the better part of a decade and dozens of attempts to fully commercialize, and one successful landing does not constitute a competing launch-services business. What it does is end, at least partially, the uncontested-technology-leadership framing that has been part of SpaceX’s valuation story since long before its IPO. This series noted in July that Amazon’s Kuiper entering commercial service represented the first genuine competitive threat to Starlink’s consumer broadband business; China’s reusable-booster milestone is the analogous first crack in the launch-services side of the story. Neither threat is close to materially denting SpaceX’s near-term revenue. Both are reminders that the premium the market has assigned to SpaceX’s core aerospace business — separate from whatever premium the AI-compute segment now commands — was built partly on the assumption that no one else could do what SpaceX does. That assumption got measurably weaker this week, even if the practical consequences remain years away.
The timing coincidence between the China milestone and today’s lockup is, on its own, just that — a coincidence of two unrelated news cycles landing in the same trading week. But coincidental timing still shapes how a market interprets ambiguous events, and a stock already absorbing a specifically dated bear case about AI-revenue durability is more likely to treat a competitive threat to its legacy launch business as confirmation of a broader caution narrative than it would if the same China news had landed on an otherwise quiet week. Sentiment compounds this way even when the underlying facts are independent, which is itself a reminder that a single week’s price action often reflects which stories happened to cluster together rather than any single story’s standalone importance.
The Rest of the Schedule Is Still Ahead
Today’s tranche is not the end of the lockup story, and it is worth restating the full schedule this series has tracked since it was first disclosed, because the market’s attention tends to reset with each individual date even though the underlying supply overhang is cumulative. The base lockup structure released roughly a fifth of the total restricted shares around the August 6 earnings date, with today’s roughly 319-million-share tranche as the second scheduled release. Further staggered tranches remain on the calendar through September and October, tied to specific day-counts from the original prospectus date rather than to any additional corporate action. Each of those future dates will, in turn, present the same underlying question today’s does: does a specific population of holders, newly free to sell, choose to realize gains into whatever price the stock has reached by then, or does the market’s demand for the stock — AI-compute enthusiasm, launch-business fundamentals, or some combination — absorb the new supply without much disruption.
That repeated structure means this series will likely revisit this exact question at least twice more before the full lockup schedule concludes. What today’s event adds to the pattern, regardless of how the next few sessions resolve, is a second data point in what was previously a single-observation story. One squeeze is a data point; it does not, on its own, establish that SpaceX’s float structure reliably produces short covering rather than crashes every time a tranche unlocks. A second event with a meaningfully different setup — smaller tranche, larger embedded gains, a named and specific bear case, uncertain short positioning — is the first real opportunity to see whether the August 6 outcome was the beginning of a durable pattern in how this stock absorbs lockup supply, or whether it was a specific, non-repeating combination of circumstances that happened to align once.
What Would Actually Distinguish This Unlock From the Last One
Set against each other, the two lockup events share their most visible feature — a stock sliding into the date — and diverge on nearly everything beneath the surface. August 6 released a larger tranche onto a stock at its lowest price since listing, against a short base sized for exactly that scenario, into an earnings report that had just handed bulls a new, verifiable growth story. Today’s tranche is smaller, arrives on a stock that has already rallied more than a third off its low, against a short base whose current size cannot be confirmed from available data, alongside a specifically dated bear thesis targeting the durability of the very AI revenue that made the first squeeze possible.
Those are not small differences. A repeat of the August 6 outcome — a rally through the unlock day, driven by short covering into thin supply — requires that a meaningful short position actually exists to be squeezed, which is precisely the fact this piece could not verify. Its absence would not necessarily produce a crash; it would more likely produce something closer to an ordinary supply-and-demand test, in which the stock’s direction depends on whether the holders newly free to sell choose to realize a 35-percent-plus gain into a market already primed by Thum’s public skepticism, or whether the bull case’s roughly 60-percent-upside consensus target proves enough to keep the newly tradeable shares in strong hands.
The watch list into the coming days, in rough order of how quickly each resolves: today’s actual close, which this piece could not capture at time of writing and which will be the first real data point on how the market absorbed the tranche; whether Thum’s Sell call draws company or a second analyst to the same side, the way Piper Sandler’s July Neutral rating was, at the time, the first crack in an otherwise uniform buy wall; the next FINRA short-interest settlement report, which will finally answer whether traders rebuilt positions ahead of this event; whether the Cognition story resurfaces in any confirmed form; and Starship Flight 14’s actual date, which conflicting sources currently place anywhere between August 28 and September 30 — itself evidence that even the company’s own operational calendar has grown harder to pin down the further this series tracks it.
There is a broader pattern worth naming as this series moves into its second lockup event rather than its first. Every prior chapter — the pre-inclusion rally, the reversal, the IPO-floor breach, the trigger mechanics, the launch delays, the earnings beat, the first squeeze — resolved with a relatively clean, singular explanation once the dust settled. This one may not. A smaller tranche, an unverifiable short position, a bear case aimed at a different part of the business than the one that drove the last squeeze, and two unrelated news stories landing in the same week is not the setup for a clean natural experiment. It is closer to the ordinary condition markets are usually in — several partial explanations competing for the same price move, none of them individually sufficient, all of them worth tracking regardless of which one turns out to matter most once today’s session closes.
