SPCX: Everyone Is Telling the Float Story. The Filings Say It Is Wrong.

Everyone Is Telling the Float Story. The Float Story Is Wrong.

On Monday 13 July 2026, one month and a day after the largest initial public offering in history, SPCX closed at $139.14. [28] That is $4.14 above the $135.00 price at which SpaceX sold 638,888,888 shares of Class A stock, and it is more than ten dollars below the $150 at which the stock opened for trading on 12 June. [28] Measured from the IPO, a buyer who got an allocation is up 3.1%. A buyer who bought at the open is down 7.2%. A buyer who bought at the first-day close of $160.95 is down 13.6%.

The entire 19% first-day pop has been surrendered. Not one locked share has been released to cause it.

That single fact indicts the story almost everyone is telling about this listing. The story goes like this: SpaceX floated only about 4.9% of itself, starved the market of stock, and the resulting scarcity manufactured a 19% pop; the shares are therefore not really priced yet, and the true price discovery arrives when the lock-up expires and the supply finally shows up. It is a reasonable prior, it was written in terms by serious people on day one, and it is the version now circulating in every lock-up tracker and countdown clock on the internet. TechCrunch: demand for SpaceX "is also a function of its small float, with only about 4% of shares available for public trading, while early investors and employees hold the rest." [21] Morningstar: SpaceX "is seen as having structured its unconventional IPO mechanics to dampen volatility and keep the stock above $135 on Day One", with "its slim float and tiered lockup further constrain[ing] the number of shares available." [22]

We are going to dismantle that story, and we are going to do it without caricaturing it. It is the strongest available theory of this stock. It is also, on the evidence of the last month, wrong.

Our call, in four beats.

One: the 19% was ordinary underpricing, not scarcity. The book was more than 4x oversubscribed, retail alone placed orders for more than $100 billion of stock, and the price was fixed by hand at $135.00 weeks before the roadshow ended. [20] [26] Bankers left money on the table. That is what a 19% pop looks like, and no scarcity term is required to produce it.

Two: every supply mechanic the scarcity story depends on has already been tested in public, and every one of them failed. The greenshoe was exercised in full — 83,333,333 additional shares, straight into the float — and the stock went up. [3] Nasdaq-100 inclusion was fast-tracked at 15 trading days, aiming mandatory, price-insensitive buying at a ~4.9% float — and the stock fell. [22]

Three: what actually moves SPCX is the balance sheet and the denominator. A $25.0 billion debut bond deal across five tranches. [14] A $60.0 billion all-stock acquisition agreed four trading days after listing. [12] And a headline multiple of roughly 95x sales whose revenue line only exists because xAI and X were retrospectively combined into every prior period under common-control accounting. [10] SpaceX proper — rockets and Starlink — did $15,473m, not $18,674m. [9]

Four: the lock-up matters, but not as a cliff and not the way it is being described. It is a 17-row cascade across 15 distinct triggers, whose first release is tied to an earnings date SpaceX has not scheduled. [5] And its celebrated price trigger accelerates supply rather than adding any — the filing's own numbers settle it. [5] Meanwhile 7.8 billion shares, approximately 60% of post-offering shares outstanding, sit locked for over a year. [7] The structure that made all of this work was bespoke and accommodated, which makes it a kill switch, not a template — an unwelcome answer for the Anthropic and OpenAI listings that NBC has already queued up behind it. [23]

What Was Actually Sold: The Deal As Priced, and the Deal As Closed

Before anyone argues about this stock, they should be able to state what was sold. Most of the coverage cannot, and the confusion is not cosmetic — it runs straight into the float argument.

Start with the price, because it was not discovered. SpaceX did not book-build. The S-1/A filed on 3 June already told investors, flatly: "We expect the initial public offering price to be $135.00 per share." [16] A single fixed number, printed before the roadshow had run its course, and it never moved. [1] This is the first of several bespoke mechanics, and it matters enormously for the underpricing argument we make in a moment: a fixed price cannot be walked up into demand. Whatever excess demand showed up had nowhere to express itself except the aftermarket.

Next, who sold. Nobody did — nobody, that is, except the company. The offering was 100% primary. All 555,555,555 base-deal shares were newly issued by SpaceX; the over-allotment option was an option to buy additional shares "from us"; the closing 8-K says the underwriters bought "from the Company." [1] [3] The strings "selling shareholder" and "selling stockholder" appear zero times in the final prospectus. Not one existing holder sold a single share into the largest IPO in history.

Pause on that, because the float thesis has to reckon with it. The scarcity was not a constraint imposed on SpaceX by an illiquid cap table. It was a choice. The company could have brought insiders along; it declined to.

Then the greenshoe. The underwriters held a 30-day option on up to 83,333,333 additional Class A shares at the IPO price, and the 424B4 footnote states they receive no discount or commission on any over-allotment shares. [1] On 15 June they exercised it in full. The IPO closed with 638,888,888 shares sold at $135.00. [3] Because the underwriters waived their cut on those shares, the entire $11,249,999,955 — call it $11.25B — flowed to proceeds before expenses. The base-deal underwriting discount was $500,000,000, or $0.90 a share, and not a cent of it applied to the greenshoe. [1]

Which brings us to the number everyone has wrong.

The reconciliation is not complicated, but it has to be done in matched pairs. On a gross basis, the 424B4 cover page prints an aggregate initial public offering price of $74,999,999,925 — 555,555,555 shares at $135.00. [1] After the greenshoe was exercised in full, the gross figure is 638,888,888 shares at $135.00, or $86,249,999,880. On a net basis, the company's own Use of Proceeds section says approximately $74.4 billion on the base deal, or $85.7 billion if the option is exercised in full, after underwriting discounts and estimated offering expenses. [2]

Both pairs are correct. Neither is a lie. What is a lie is the version that has entered circulation — "to $85.7 billion from $75 billion" — which subtracts a gross number from a net one and thereby manufactures a greenshoe increment that does not exist. The real increment is $11.25B, gross to gross. There is no secondary component to explain any part of the gap. The gap is the greenshoe, and nothing else. [1] [3]

It is worth noting how this fog descended in the first place. SpaceX's own pricing press release of 11 June gives the share count and the price and contains no dollar proceeds figure at all. [19] Every "$75 billion" headline is a journalist's multiplication — correct when written, and stale from 15 June onward. TechCrunch, NBC, NPR and CBS all ran it. [20] [23] [24] [25] They were right on 12 June. They were simply never updated.

The scale, for orientation. The previous record IPO was Saudi Aramco's 2019 listing, which TechCrunch reports raised $24.9 billion. [20] SpaceX's deal was several times that. Goldman Sachs led a syndicate of ten book-running managers alongside twelve co-managers — Morgan Stanley, BofA, Citigroup, J.P. Morgan, Barclays, Deutsche Bank, RBC, UBS and Wells Fargo among them. [19] At the $135 offering size, Musk was to hold approximately 82.4% of the voting power, on a dual-class structure in which Class B carries ten votes and Class B holders elect a majority of the board. [16] And the money, per the company, goes first to "the expansion of our AI compute infrastructure" — that phrase leads the list in both the prospectus and the closing 8-K, ahead of launch infrastructure, ahead of the satellite constellations. [2] [3] Remember that ordering. It is the whole company, disclosed in a sentence.

1. 3 June 2026

   The price is fixed before the roadshow ends[16]

   S-1/A Amendment No. 2: 'We expect the initial public offering price to be $135.00 per share.' A single number, not a range. It never moved.

2. 9 June 2026

   CFO tells buyers the hard part is ahead[17]

   In an SEC-filed roadshow transcript, Bret Johnsen concedes capital allocation 'really has been one of the bigger challenges of this job for 15 years' and 'the dollars are now bigger than they have been in the past'.

3. 11 June 2026

   Priced: 555,555,555 shares at $135.00[19][1]

   SpaceX's own pricing release gives the share count and the price — and no dollar proceeds figure at all. Every '$75 billion' headline is a journalist's multiplication.

4. 12 June 2026

   Debut: opens $150, highs $176.52, closes $160.95 (+19%)[21][24][26]

   An undershoot of pre-open expectations, which per Forbes had projected a $175 open. Ritter: 'This is not a moonshot.'

5. 15 June 2026

   Greenshoe exercised IN FULL — 638,888,888 shares sold[3][1][2]

   83,333,333 additional shares straight into the float, on which the underwriters took zero commission. Gross proceeds $86.25bn; net $85.7bn. Supply went up. The price did not break.

6. 16 June 2026

   Cursor: $60.0 billion, all in SPCX stock[12]

   Four trading days after listing, SpaceX agrees to acquire Anysphere at an implied equity value of $60.0bn, payable entirely in Class A stock at a seven-day VWAP.

7. 22 June 2026

   Notes launched; ~$100.8bn of cash disclosed[13]

   A Reg FD disclosure to prospective bond investors puts cash and equivalents at approximately $100.8 billion as of 19 June.

8. 23–26 June 2026

   $25.0 billion of senior notes priced and closed[14][15][9]

   Five tranches: $7.0bn at 5.350% (2031), $6.0bn at 5.650% (2033), $6.0bn at 5.875% (2036), $2.5bn at 6.600% (2046), $3.5bn at 6.650% (2056). Roughly $1.46bn of annual coupon, against FY2025 interest expense of $1,945m.

9. Early July 2026

   Nasdaq-100 fast-track — 15 trading days after listing[21][22]

   An unprecedented path into the index, after SpaceX lobbied to have inclusion rules changed. Mandatory, price-insensitive buying, pointed at a 4.9% float. The stock fell.

10. 13 July 2026

    Closes at $139.14 — the pop is gone[28]

    'Barely above its official IPO price -- and more than $10 below where the stock began trading on IPO day.' Attributed in part to China's recovery of a reusable Long March 10B. Zero locked shares have been released.

Three Natural Experiments the Scarcity Story Already Failed

Give the float thesis its full strength first, because it deserves it.

The float really was tiny. The 424B4's Dilution table shows new investors in the base offering buying 555,555,555 shares — 4.2% of the 13,075,865,175 shares outstanding after the offering, with Musk and existing investors retaining 95.8%. [4] With the over-allotment exercised in full, the prospectus's own footnote gives the split as approximately 95.1% / 4.9%. [1] Because the greenshoe was exercised, 4.9% is the actual free float. [3]

The mechanics really were engineered. Morningstar said so in terms, and Morningstar is not a wire service reaching for a phrase: SpaceX "is seen as having structured its unconventional IPO mechanics to dampen volatility and keep the stock above $135 on Day One." [22] A fixed price. A retail allocation set at up to 30% of the deal, an unusually large share, then cut closer to 20% at pricing. [22] A slim float. A tiered lock-up. If float mechanics set prices anywhere, they set them here.

Which is precisely why the theory is falsifiable — and why it has now been falsified three times, in public, with money.

Experiment 1: supply was added, and the price went up

On 15 June the underwriters exercised the over-allotment in full, adding 83,333,333 shares to a float of 555,555,555. [3] That is a roughly 15% increase in the number of freely tradable shares, delivered in a single settlement, with no offsetting demand event. A scarcity model makes an unambiguous prediction here: more supply, lower price.

The stock rose. It kept rising into its June high — a level we will not print, because the only source carrying it in our evidence base is a market-data aggregator we do not consider citable, and we would rather state the direction honestly than a number we cannot stand behind. The direction is not in dispute. New supply arrived and the price went up.

This is the natural experiment the scarcity story most needs to explain, and it cannot be laundered as a source conflict. It ran, and it ran against the thesis.

Experiment 2: the forced bid arrived, and nothing lifted

Now the strongest test a scarcity model could possibly ask for. SpaceX obtained an unprecedented path into the Nasdaq-100 just 15 trading days after the IPO, having lobbied indexes to change their inclusion rules. [22] TechCrunch flagged the manoeuvre on day one, noting the company "successfully lobbied a number of indexes (like the Nasdaq 100) to change their inclusion rules." [21]

Consider what index inclusion is, mechanically. It is a wall of mandatory, price-insensitive buying. Passive funds do not care what the stock costs; they are obliged to own it in proportion. Point that at a company whose free float is 4.9% and a scarcity model predicts a violent, near-vertical lift. Demand that cannot be refused, meeting supply that does not exist.

The stock fell. It fell into inclusion, it fell after it, and by 13 July it closed at $139.14, "barely above its official IPO price -- and more than $10 below where the stock began trading on IPO day." [28] If float mechanics set the price of this security, that outcome is not merely surprising. It is impossible.

Experiment 3: the round trip finished before the cascade even began

SPCX went from $135.00 to a June high and back to $139.14 — the full arc, up and down — with 100% of the locked shares still locked. [28] Approximately 60% of post-offering shares outstanding are restricted for over a year, and not one of them has moved. [7]

If the lock-up cascade were where "real price discovery" happens, this could not already be over. You cannot have discovered the price before the discovery mechanism has been switched on. Either the supply was pre-priced by a market perfectly capable of reading a lock-up table — in which case the cascade is not a discovery event — or price is being set by something else entirely. We think it is the latter, and we will name it in a moment.

Source: 424B4; 8-K of 15 June 2026; CBS, TechCrunch and Motley Fool day-one and 13 July coverage

And a 19% pop is small for a 4.9% float, not large

The cross-sectional intuition also runs the wrong way for the scarcity camp. Jay Ritter of the University of Florida — the leading academic authority on IPO underpricing, the man who has spent a career building the datasets everyone else cites — was distinctly underwhelmed. "Like most IPOs, the price jumped," he told CBS. "This is not a moonshot." [24] Like most IPOs. Not: unlike any IPO, because of the float.

Look at the intraday tape. SPCX printed a high of $176.52 — a gain of nearly 31% over the offer — and then closed at $160.95. [24] [21] Supply met demand inside the session, ten percent below the high, with a float supposedly too small to matter. A genuinely binding float prevents exactly that. It does not permit a stock to sell off 10% from its high on its first afternoon on the tape.

Nor is the short side trapped. Listed options trade, borrow exists, short interest is negligible. A true float squeeze requires the short side to be locked out. It is not.

One last piece of hygiene, because it is the number now circulating in the retail press. 24/7 Wall St. has claimed that "only 281 million shares of the 7.57 billion outstanding are currently in the public float," implying a float of about 3.7%. [31] Both halves are wrong. The numerator is wrong: 638,888,888 shares were sold, not 281 million. [3] The denominator is wrong: total shares outstanding are 7,380,196,910 Class A plus 5,695,668,265 Class B, before the greenshoe. [4] The 7.57 billion appears to be a Class A-only count. If you are working from that figure, you are working from a number that contradicts the filings twice. Discard it.

	Scarcity model predicts	What the tape did
Greenshoe exercised in full, 15 June (+83,333,333 shares, roughly +15% to the float)[3]	New supply with no offsetting demand event → the price falls	The stock rose into its June high. Supply arrived; the price went up.
Nasdaq-100 fast-track at 15 trading days — mandatory, price-insensitive buying against a 4.9% float[22][28]	A violent, near-vertical lift: demand that cannot be refused, meeting supply that does not exist	The stock fell into inclusion, fell after it, and closed at $139.14 on 13 July
Day-one price path with a supposedly binding float[21][24][26]	The open overshoots; the float binds all session and prevents any selloff	Opened $150 against a ~$175 expected print; then sold off from a $176.52 high to close at $160.95 — supply met demand inside the session
One month of trading with 100% of locked shares still locked[28][7]	Price held aloft by scarcity until the cascade releases the supply	The entire 19% pop surrendered before a single locked share moved; IPO buyer +3.1%, open buyer −7.2%
Explaining the 19% pop[16][20][26]	Scarcity manufactured it — a starved market bid up a tiny float	A 4x-covered book, >$100bn of retail orders and a hand-set $135.00 price. Textbook underpricing. No scarcity term required.
What the float actually does[21][23][26]	Sets the level of the price	Sets the precision of the price — 4.9% of the company established a ~$2.1 trillion mark, and a thin float makes that mark fragile, not robust

If Not Scarcity, Then What: An Underpriced Book, Priced by Hand

Here is the positive account of the 19%, and it needs no scarcity term at all.

SpaceX fixed its price at $135.00 and published that number in an S-1/A on 3 June, before the roadshow had finished running. [16] The book then came in more than 4x oversubscribed, per Bloomberg, carried by TechCrunch on IPO day. [20] [21] Retail investors placed orders for more than $100 billion of stock, per Bloomberg via Forbes; BlackRock alone ordered at least $5 billion, per the Wall Street Journal via Forbes. [26] The retail allocation was set at up to 30% of the offering and then reportedly cut closer to 20% at pricing. [22] And at the company's request, the underwriters reserved up to five percent of the offered shares for sale at the IPO price to employees and persons identified by executive officers. [1]

Set the mechanics aside and read that as an economist would. A price fixed weeks in advance, immovable. Demand arriving at multiples of the supply. Allocations rationed and then re-rationed downward. The unmet demand does not evaporate — it queues for the open. This is the textbook underpricing outcome, described in the literature for forty years, and it is entirely sufficient to produce a 19% first-day gain. The bankers left money on the table. That is the mechanism. Ritter, again: "Like most IPOs, the price jumped." [24]

And now the tell, the detail that ought to end the argument.

If a tiny float were genuinely binding against 4x demand, the open should have overshot. Instead it undershot. Pre-open expectations, per Forbes, had SPCX opening around $175 — a roughly 30% jump. [26] It opened at $150, about 11% above the offer. [21] Ritter called the opening print "disappointing relative to what betting markets had been predicting." [24]

Read that carefully. The most scarcity-constrained large IPO ever brought to market failed to reach its own expected opening price. A binding float does not do that. A book of ordinary excess demand, released into an aftermarket with an ordinary supply curve, does exactly that.

We want to be scrupulous about what we do and do not know about the size of that demand. If you take the 4x coverage ratio and apply it to the $74,999,999,925 base deal, you get on the order of $300bn of orders, of which roughly $225bn went unfilled. Apply the ~20% retail allocation to the same base and you get roughly $15bn of stock allocated to retail against more than $100 billion of retail orders — implying something like $85bn of retail demand with nowhere to go but the open market on Friday morning. These are our estimates, not reported figures. Bloomberg's underlying book data is paywalled and we did not fetch it. Treat our numbers as order-of-magnitude illustrations of a demand curve, never as facts. But note that even the conservative end of that range describes an enormous, unsatisfied, price-insensitive bid that had to clear somewhere. It cleared at $160.95.

So what is the float actually for, in the causal story? Not much — and what it does do cuts against the bulls, not for them. A 4.9% float is a lever arm. It means a very small number of shares set a very large mark. That is a statement about the precision of the price, not its level. When SPCX printed a market capitalisation of approximately $2.1 trillion at the day-one close — with intraday prints described as "nearly $2.3 trillion" and "more than $2.25 trillion" — that $2.1 trillion figure was established by trading in 4.9% of the company. [26] [21] [23] A thin float does not make a mark more robust. It makes it more fragile. And the last month has been the demonstration.

The Multiple Everyone Quotes Is Measuring a Company That Did Not Exist

Now the part of this report that we have not seen anyone else write.

Every headline valuation of SpaceX runs through a single number: FY2025 revenue of $18,674m. [9] At the $135.00 offering price and the full post-greenshoe share count, that is roughly 95x sales. It is the number in every note, every screener, every argument about whether the stock is expensive.

That revenue line is audited, and it is real. It is also, in a sense that matters enormously and that nobody is saying out loud, not SpaceX's.

On 2 February 2026, SpaceX completed its acquisition of X.AI Holdings Corp. Ten months earlier, xAI had acquired X. Both transactions were accounted for as reorganizations of entities under common control, because Musk held a controlling financial interest in each entity throughout. [10] The accounting consequence is stated plainly in the prospectus: "The Company's consolidated financial statements have been prepared to reflect the retrospective combination of the net assets of the entities at their historical carrying amounts for all periods presented. No new goodwill or other intangible assets have been recorded." And: "As the consolidated financial statements already reflect the reorganization of entities under common control for all periods presented, separate financial statements of xAI and X are not provided." [10]

Read that again. There is no purchase price. There is no goodwill. There are no separate xAI financials. Every prior period was restated backwards to include a company that was not part of SpaceX at the time. The prospectus itself concedes the point in the technical language of the accountants: "This presentation constitutes a change in reporting entity." [10]

So when the market says SpaceX trades at ~95x sales, it is applying a multiple to a reporting entity that did not exist until February 2026.

What did the actual rocket-and-satellite company do? The segment disclosure answers it. FY2025 revenue splits Space $4,086m, Connectivity $11,387m, AI $3,201m. [9] Strip the AI segment — strip xAI and X, the things that were retrospectively bolted onto every prior year — and the business the public thinks it bought did $15,473m. On the same market capitalisation, that is roughly 115x sales, not 95x.

Same price. Same day. Different company. The multiple moves twenty turns depending on which SpaceX you think you are buying, and the market has never once been asked to choose.

IPO price, $135.00Day-one close, $160.9513 July close, $139.14Morningstar fair value, $63Raymond James target, $800Reset

Shares outstanding (Class A + Class B, post-greenshoe)fact13.159bn

SPCX price per sharefact139.14$

FY2025 revenue AS REPORTED (xAI and X retrospectively combined)fact18,674$m

FY2025 AI-segment revenue (the part that was not SpaceX)fact3,201$m

Implied market capitalisation1,831$bn

Price / FY2025 sales — AS REPORTED98.0×

FY2025 revenue of SpaceX proper (Space + Connectivity)15,473$m

Price / FY2025 sales — ROCKETS AND STARLINK ONLY118.3×

Turns of sales the accounting hides20.3×

At $135.00 this prints 95.1x on the reported denominator and 114.8x on SpaceX proper. There is no P/E: the company lost $4,937m in FY2025. The reported revenue line is audited and real, but it exists only because the xAI and X mergers were accounted for as reorganizations of entities under common control — retrospective combination at historical carrying amounts, no purchase price, no goodwill, and no separate xAI or X financials ever published. Every headline multiple on this stock is applied to a reporting entity that did not exist until February 2026.

And the AI segment is not a rounding error you can wave through on the way to a clean number. It is 17% of revenue and it is the entire loss. FY2025 income from operations by segment: Space $(657)m, Connectivity a positive $4,423m, AI $(6,355)m. [9] Cumulatively, on our own arithmetic from the segment history, the AI business has produced roughly $14.4bn of operating losses on roughly $9.6bn of revenue. It has lost more money than it has ever booked in sales.

It is also where every dollar goes. Of $20,737m of FY2025 group capital expenditure, $12,727m — 61% — went to the AI segment. In Q1 2026 the ratio got worse: AI capex of $7,723m out of $10,107m of group capex. Three-quarters of every capital dollar, aimed at a segment that produced $818m of revenue and a $(2,469)m operating loss in the same quarter. [9] Management is candid about the payback: "we expect a multi-year investment horizon before these deployments translate into sustained positive AI Segment Adjusted EBITDA." [9]

Nor is the scale metric flattering. Roughly 550 million monthly actives across Grok and X at 31 March 2026, of whom about 117 million touched Grok's AI features — but only about 6.3 million active paid subscribers, up from 4.9m at the end of 2025. [9] Nameplate compute draw has reached 1.0 gigawatt, from 0.3GW a year earlier. Enormous compute, enormous audience, six million payers.

Which is not to say the company has no profit engine. It has exactly one, and it is not the rockets. Connectivity — Starlink — did $11,387m of revenue in FY2025, up 49.8%, with $4,423m of operating income, up 120.4%, and $7,168m of Segment Adjusted EBITDA, up 86.2%. [9] Approximately 9,600 broadband and mobile satellites in low-Earth orbit at 31 March 2026, roughly 75% of all active maneuverable satellites in orbit, with V3 satellites expected to begin deploying in the second half of 2026. [9] That is a real, dominant, cash-generating business.

It is also a business whose unit economics are deteriorating while it grows. Starlink had approximately 10.3 million subscribers at 31 March 2026, up from 5.0 million a year earlier — and monthly ARPU fell to $66 from $86 over the same period. [9] The prospectus does not hedge: "We generally expect Starlink Subscriber ARPU to continue to decline over the next few years." [9] Doubling subscribers while cutting revenue per subscriber by nearly a quarter is a land-grab, and land-grabs are fine. But the price being paid for it is not. Value the entire company on its only profitable segment and the IPO priced Starlink at roughly 156x that segment's revenue — around $172,000 of market capitalisation per subscriber, against annual service revenue per subscriber of about $792 at a $66 monthly ARPU.

Cadence is not revenue

Meanwhile the business the brand is built on is shrinking. Space segment revenue fell to $619m in Q1 2026 from $865m, a 28.4% decline, with Launch Services specifically at $330m against $566m, and the segment swinging to a $(662)m quarterly operating loss from $(70)m. [9]

How can that be, when SpaceX launched more than ever? Because launch cadence is not launch revenue, and the filing shows both facts at once. The Space segment key-metrics table reports 98 launches in 2023, 138 in 2024, 170 in 2025 and 40 in Q1 2026, with mass to orbit rising from 1,210 to 2,213 metric tons. [9] But the MD&A footnote shows that only 43 of 165 Falcon launches in 2025 were customer launches — and only 7 of 40 in Q1 2026, down from 12. Of 2,213 tons to orbit in 2025, just 312 tons were customer payloads. [9] SpaceX recognises no inter-segment revenue on its own Starlink launches.

The two launch counts — 170 and 165 — are both in the same document and measure different things; the first counts all vehicles including Starship flight tests, the second counts Falcon only. Neither is wrong. But the one that matters is 43 of 165. The most celebrated operational achievement in the history of the industry is, on the income statement, mostly the satellite division's cost line.

Segment	Revenue ($m)	% of revenue	Income (loss) from operations ($m)	FY2025 capex ($m)	What it is
Space (launch + Starship)	4,086	21.9%	(657)	not disclosed separately	170 launches in 2025 — but only 43 of 165 Falcon flights were for customers. Q1 2026 revenue fell 28.4% to $619m.
Connectivity (Starlink)	11,387	61.0%	4,423	not disclosed separately	The entire profit engine: +49.8% revenue, +120.4% operating income, $7,168m Segment Adj. EBITDA. 10.3m subscribers — at a falling $66 ARPU.
AI (xAI + X)	3,201	17.1%	(6,355)	12,727	61% of FY2025 group capex (76% in Q1 2026) for 17% of revenue and 100% of the losses. ~6.3m paid subscribers against 1.0 GW of nameplate compute.
TOTAL	18,674	100%	(2,589)	20,737	Net loss of $(4,937)m after $1,945m of interest expense. Operating cash flow $6,785m — against $20,737m of capex.

Two outside checks on the price, and one from the company

Morningstar's Nicolas Owens puts SpaceX's fair value at $780 billion, or $63 per share — against an IPO priced at $135 and a valuation of $1.77 trillion. [22] That is to say the deal came at roughly 2.3 times what one of the two major independent equity shops thinks the company is worth, and Morningstar's own number implies roughly 42x FY2025 sales, which is hardly a value investor's mark.

Raymond James's Brian Gesuale, meanwhile, maintains a strong buy, cites increased "operational cadence", and predicts the stock "will hit $800 within a year." [28]

$63 to $800. That is not a disagreement about a discount rate. That is two analysts looking at the same security and disagreeing by more than an order of magnitude about what it is. A dispersion like that is itself the evidence: price discovery has not happened here, and it is not going to be resolved by a lock-up table. It will be resolved when somebody decides what the denominator is.

And lest this seem like outside cynicism, note that the company printed the arithmetic itself. Adjusted pro forma net tangible book value after the offering is $7.85 per share. Buyers at $135.00 therefore took "immediate dilution... of $127.15 per share." [9] Ninety-four percent of what an IPO investor paid was premium over tangible assets. SpaceX told them so, on the page, before they bought.

What Actually Moves This Stock: The Balance Sheet, the Currency, and China

Here is the affirmative half of the call. In the 31 days between listing and 13 July — with the cap table one hundred percent frozen, not a single locked share released — SpaceX transacted more consequential business than the entire lock-up cascade will do all year.

16 June: a $60.0 billion acquisition, all in stock. Four trading days after listing, SpaceX agreed to acquire Anysphere — Cursor — at an implied equity value of $60.0 billion, payable entirely in SpaceX Class A stock priced at the seven-day volume-weighted average close before closing, expected to complete in Q3 2026. [12] Four trading days. The company reached for its newly-minted equity currency almost before the confetti was swept up.

22–26 June: the largest debut bond deal we can recall. On 22 June SpaceX launched an offering of senior unsecured notes and disclosed to prospective investors that as of 19 June it held approximately $100.8 billion in cash and cash equivalents. [13] On 23 June it priced $25.0 billion across five tranches: $7.0bn of 5.350% notes due 2031, $6.0bn of 5.650% due 2033, $6.0bn of 5.875% due 2036, $2.5bn of 6.600% due 2046 and $3.5bn of 6.650% due 2056. [14] The deal closed 26 June under an indenture with BNY Mellon, with proceeds repaying the bridge loan facility in full. [15] (The launch 8-K states no size at all, and the priced-and-closed figure in the company's own filings is $25.0 billion.)

Do the coupon arithmetic and the significance appears. Those five tranches carry roughly $1.46 billion of interest per year. FY2025 interest expense was $1,945 million. [9] SpaceX just grew its annual interest bill by something on the order of 75% — before a single dollar of the AI capex earns anything.

These are value inputs. Cash flow, dilution, discount rate, competitive position. They landed in public, in filings, with zero shares unlocked. That is price discovery. It is happening right now, it moved the stock, and it has nothing whatsoever to do with float.

The hole the bond deal is answering

Why does a company that just raised the largest IPO in history need $25.0 billion of debt eleven days later?

Because the burn is enormous. FY2025 net cash provided by operating activities was $6,785 million, against $20,737m of capital expenditure — free cash flow of $(13,952)m. [9] In Q1 2026 alone, operating cash flow of $1,047m against $10,107m of capex: $(9,060)m in a single quarter. [9] The balance sheet at 31 March 2026 showed total assets of $102,094m, total liabilities of $60,512m, shareholders' equity of $34,533m and cash of $15,852m, with contract backlog of $27,621m and deferred revenue of $13,236m. [9]

Stack the IPO and the notes against that run-rate and the picture is not of a company awash in capital. It is a company that has bought itself a defined number of quarters at the current burn — and is spending three-quarters of every capex dollar on a segment management has said outright will not pay for years. [9]

The CFO, on the record, in a transcript SpaceX itself filed with the SEC as roadshow marketing material, put it about as candidly as a CFO can: capital allocation "really has been one of the bigger challenges of this job for 15 years," and "the dollars are now bigger than they have been in the past." [17] He was talking about Terafab, Starship capacitisation and terrestrial data centres. He was, in effect, telling the buyers of the largest IPO in history that the hardest problem in his job is about to get harder.

Source: 424B4 — cash flow statement and segment capex disclosure

The equity is now a currency — and the currency has an owner

Watch what SpaceX has been buying with paper. The EchoStar Spectrum Transaction: total consideration of approximately $19.6 billion for 50 MHz of AWS-4/H-Block plus up to 15 MHz of AWS-3 spectrum, comprising approximately $11.1 billion in equity — around 261.8 million Class A shares issued at a fixed value of $42.40 per share — and up to $8.5 billion to pay off designated EchoStar debt. [11] Then Cursor: $60.0 billion, all stock, struck against a seven-day VWAP. [12]

SpaceX now pays for assets with a share price it has a direct and material interest in supporting. That is not an accusation; it is a structural observation, and it is the sort of thing that ought to be in the front of an investor's mind when reading a company's disclosure choices — including, as we are about to see, when it chooses to schedule its first earnings release.

And then there is the thing no supply model contains

The 13 July decline — the one that took SPCX back to $139.14 — was attributed in part to a fundamental, non-supply catalyst: over the preceding weekend, the China Aerospace Science and Technology Corporation launched and recovered a reusable Long March 10B, which "closes the technology gap between SpaceX and China." [28]

That is the moat, not the float. No lock-up table contains a row for it. No scarcity model has a coefficient for it. And it moved the stock.

The Lock-Up Is a 17-Row Cascade — and Its Price Trigger Accelerates Supply, It Doesn't Add Any

The lock-up still matters. It matters a great deal. It just does not do what the trackers say it does, and most of them have not even counted it correctly.

Get the structure right, because almost nobody has

Investing.com and others describe "16 different dates." [30] Counted directly from the 424B4, there are 17 release rows across 15 distinct trigger dates. Two rows share the First Earnings Release Date — the base release and the conditional price-triggered release. Two rows share 12 June 2027 — the final extended tranche and Musk's entire holding. [5] [6]

And there is a row that no published tracker we checked lists at all. On 10 September 2026 — the 91st day, the day after Rule 144's 90-day post-IPO seasoning is met — up to 59.1 million shares held specifically by Rule 144 "affiliates" are released: the shares carved out of the earlier tranches. [5] It is in the filing. It is in nobody's countdown clock.

There are three pools, not one.

The 180-day pool, which our arithmetic sizes at roughly 4.69 billion shares — derived two independent ways (summing the no-trigger release path, and grossing up a 7% tranche) that agree to within 0.4m shares. It runs as follows. Up to 911.5 million shares on the second full trading day after the First Earnings Release Date. [5] Then a ladder of five fixed tranches of 7% each, at 70, 90, 105, 120 and 135 days after the prospectus: 20 August 2026 (up to 319.0m), 9 September 2026 (up to 319.0m), 24 September 2026 (up to 328.4m), 9 October 2026 (up to 328.4m), 24 October 2026 (up to 328.4m). [5] Then the largest pre-Musk event of all: up to 1.3 billion shares — 28% of the pool — on the second trading day after Q3 results. [5] And finally the nominal "180-day expiry" on 8 December 2026, which by then is a residual.

The extended pool, roughly 1.76 billion shares, running all the way to Q2 2027 results on a 20/10/20/10/20/20 ladder: up to 351.9m after Q4 2026 results; up to 176.0m on 18 March 2027; up to 351.9m after Q1 2027 results; up to 176.0m on 17 May 2027; up to 351.9m on 12 June 2027; and the remaining 351.9m after Q2 2027 results. [6]

And Musk, who is not a cascade at all but a single cliff: up to 6.4 billion Class A shares — all of his holdings, including Class A issuable on conversion of Class B, which itself includes 350 million Class B shares underlying options — released in one go on 12 June 2027, the 366th day. The prospectus states flatly that his shares "will not be subject to any early release provisions." [6] [7]

Together: an aggregate of 7.8 billion shares, including 100% of the shares owned by Mr. Musk, restricted for over one year — approximately 60% of shares outstanding after giving effect to the offering. [7]

#	Earliest date available for sale	Pool	Shares (up to)	What it is
1	Second full trading day after the First Earnings Release Date — UNSCHEDULED	180-day	911.5m	20% of the 180-day pool. 1.43x the entire IPO; 1.64x the base deal. Morningstar's Owens: 'That's more than the IPO.'
2	Same day — ONLY IF the closing price is ≥ $175.50 (30% above the $135.00 offer) on 5 of the 10 trading days ending on the First Earnings Release Date	180-day (conditional)	455.8m	The 'Additional Release Shares'. This ACCELERATES supply; it does not add any — see the 8 December row. Test is on CLOSING prices, so the $176.52 day-one intraday print never counted.
3	20 August 2026 (70th day after the prospectus)	180-day	319.0m	7% of the 180-day pool
4	9 September 2026 (90th day)	180-day	319.0m	7% of the 180-day pool
5	10 September 2026 (91st day — the day after Rule 144's 90-day seasoning is met)	180-day (Rule 144 affiliates)	59.1m	THE ROW THE TRACKERS MISS. Shares held specifically by Rule 144 'affiliates', carved out of the earlier tranches. It is in the filing and in nobody's countdown clock.
6	24 September 2026 (105th day)	180-day	328.4m	7% of the 180-day pool
7	9 October 2026 (120th day)	180-day	328.4m	7% of the 180-day pool
8	24 October 2026 (135th day)	180-day	328.4m	7% of the 180-day pool
9	Second full trading day after Q3 2026 results — UNSCHEDULED	180-day	1.3bn	28% of the pool. The biggest pre-Musk event of all.
10	8 December 2026 (180th day — the nominal 'expiry')	180-day	328.4m if the Additional Release Shares were released; 797.6m if they were not	The proof that the price trigger accelerates rather than adds: the December residual absorbs the entire 469.2m difference. By now the ladder has let out 83–93% of the pool; the headline expiry is engineered to be a non-event.
11	Second full trading day after Q4 2026 results	Extended	351.9m	20% of the extended pool (~1.76bn shares), which excludes Musk
12	18 March 2027 (280th day)	Extended	176.0m	10% of the extended pool
13	Second full trading day after Q1 2027 results	Extended	351.9m	20% of the extended pool
14	17 May 2027 (340th day)	Extended	176.0m	10% of the extended pool
15	12 June 2027 (366th day)	Extended	351.9m	20% of the extended pool
16	12 June 2027 (366th day) — the same day	MUSK	6.4bn	A cliff, not a cascade: all shares held by Mr. Musk, including Class A issuable on conversion of Class B (which includes 350m Class B underlying options). 'Shares held by Mr. Musk will not be subject to any early release provisions.'
17	Second full trading day after Q2 2027 results	Extended	351.9m	The remaining shares subject to the extended lock-up

The correction, stated hard: the price trigger does not add supply

This is the error at the centre of the popular account, and the filing settles it in a single row.

The mechanism: up to 455.8 million additional shares — the "Additional Release Shares" — unlock early if the reported closing price on Nasdaq is at least 30% greater than the $135.00 offering price, i.e. at least $175.50, for at least five of the ten consecutive trading days ending on and including the First Earnings Release Date. [5]

The popular reading is that a higher price triggers more selling, so the pop is self-limiting — a clever piece of reflexive engineering. Motley Fool put it as ties "insider selling to strength rather than weakness." [27] It sounds sophisticated. It is wrong.

Read the 8 December row of the 424B4 table. It releases "up to (i) 328.4 million shares of Class A common stock if the Additional Release Shares were released on the First Earnings Release Date or (ii) up to 797.6 million shares if the Additional Release Shares were not released." [5]

There it is, in the company's own table. If the trigger fires, December shrinks. The delta between the two December outcomes — 797.6 million versus 328.4 million — is 469.2 million shares, which is the 455.8 million Additional Release Shares plus rounding. The same stock simply moves from December to July or August.

Run the percentages and the proof is airtight. Cumulative release from the 180-day pool by the Q3 print, without the trigger: 20% + (5 × 7%) + 28% = 83%, leaving 17% for 8 December. With the trigger: 20% + 10% + 35% + 28% = 93%, leaving 7%. The December residual absorbs the entire difference, exactly. Total supply is fixed. Only its timing is reflexive.

A higher price pulls roughly 456 million shares of supply forward by about four months. It does not conjure them from nowhere. Every model that treats the trigger as a supply-creation mechanism — and that is every model we have seen — is double-counting half a billion shares.

Source: 424B4 lock-up release table; percentages as printed in the filing

Nor is the trigger likely to fire on anything like current prices. From the 13 July close of $139.14, the stock must rally 26.1% simply to touch $175.50 — and then hold above it on five of the ten trading days ending on an as-yet-unscheduled earnings date. [28] [5] Note also that the test is on closing prices: the day-one intraday print of $176.52 would have cleared the bar and it never counted for anything. [24] On current prices the Additional Release Shares do not release early. They fall into the December bucket, which then swells to 797.6 million.

What does survive: the scale, and it is a dilution story

Strip out the reflexivity and one enormous fact remains. The first release alone — up to 911.5 million shares — is 1.43x the entire IPO, and 1.64x the 555,555,555-share base deal. [5] [3] Morningstar's Owens, on the same release, in four words: "That's more than the IPO." [22]

Free float goes from 4.9% to approximately 11.8% in two trading days — or approximately 15.2% if the trigger somehow fires. It roughly triples.

But notice which direction that cuts. Morningstar reached exactly our conclusion and said so: "That makes dilution, rather than scarcity, a potential headwind for SpaceX stock." [22] Owens adds the counterweight, fairly: index funds float-adjust their weightings and "will increase their allocation over time as the float increases." [22] Both things are true. Neither is a scarcity story. The float thesis and the dilution thesis are not two versions of the same argument — they are opposites, and the analyst most often cited in support of the first has actually written the second.

The date does not exist

And here is the finding that ought to be a scandal, and instead is a footnote.

The "First Earnings Release Date" — the trigger for the 911.5 million share release and for the $175.50 price test — is not a date. It is an event: the public release of results for the quarter ended 30 June 2026, expressly excluding "flash" numbers or preliminary, partial earnings. [5] As of 14 July 2026, SpaceX had not announced it. [3]

The trackers have filled the vacuum with guesses. Motley Fool frames it as "late July." [27] We will not print a date, because any date we printed would be a guess dressed as reporting.

The refusal is the point. The single most consequential date for price discovery in the largest IPO in history is unscheduled — and it is scheduled at the discretion of the issuer whose insiders are waiting on it, and whose stock is the currency it is using to buy Cursor.

The escape hatch nobody prices

One more sentence in the prospectus renders the entire table advisory: "Goldman Sachs & Co. LLC may release any of the securities described above at any time, subject to applicable requirements." [7]

Every date in the cascade is a ceiling on restriction, not a floor on supply. And behind the cascade the plumbing is already installed: approximately 12.2 billion Class A shares become generally available for Rule 144 resale starting 90 days after the offering, subject to the lock-ups; parties to the Investors' Rights Agreement can require SpaceX to register approximately 9.2 billion Class A shares; and an S-8 registering resales under historical equity-compensation plans was filed on 12 June, the first day of trading. [8] [18]

The company concedes the whole mechanism in a risk factor, and it is worth quoting in full because it is the most honest sentence in the document: "future resales of our Class A common stock by Mr. Musk or other existing shareholders, or the perception that such sales may occur, could cause the market price of our Class A common stock to decline significantly, regardless of our actual business performance." [8]

Regardless of our actual business performance. They wrote that. And note the same bank — Goldman — that led the syndicate holds the release valve. [19] [7]

Not a Template. A Kill Switch.

The title of this report promises an answer about the next two. NBC has already named them: SpaceX was "the first of what are expected to be several major IPOs related to AI in the coming months," with Anthropic and OpenAI both readying public offerings — while noting that "questions remain about just how profitable these companies will be." [23]

So: is SPCX the template?

Inventory what SpaceX had to be granted, none of which is available to the next issuer on demand. A single fixed price of $135.00 instead of a book-build. [16] Ten book-running managers and twelve co-managers — very nearly the entire street on one cover. [19] A retail allocation of up to 30% of the deal, an unusually large share, later trimmed to about 20%. [22] A directed-share reservation of up to five percent of the offering for employees and persons identified by executive officers. [1] A greenshoe on which the underwriters waived their commission entirely, against a $500,000,000 discount on the base deal. [1] A lock-up whose release valve is held by the lead bank. [7] And, decisively, index rules rewritten to fast-track the company into the Nasdaq-100 in 15 trading days. [22]

The structure did not survive because it was clever. It survived because it was accommodated. Every one of those accommodations is discretionary, revocable, and held by a party — Nasdaq, S&P Dow Jones Indices, Goldman Sachs — that owes the next issuer precisely nothing.

And here is the inversion. Suppose S&P DJI holds its line on twelve-month seasoning, profitability and a minimum float. Suppose SPCX finishes 2026 at or near $135. Then the lesson the next mega-cap draws from this listing is not "float less." It is "float more" — because the small float bought SpaceX nothing at all. It did not hold the price up: the stock round-tripped to $139.14 with the cap table frozen. [28] It did not create the pop: a 4x book and a hand-set price did that. [20] What the tiny float actually accomplished was to let 4.9% of a company set a $2.1 trillion mark [26] — and then to watch that mark evaporate inside a month on news, a bond deal, and a Chinese rocket that had nothing to do with shares.

What would prove us wrong

We will state the test before the fact, because a call that cannot be falsified is not a call.

Test one. Watch SPCX over the two sessions following the Q2 earnings release, when up to 911.5 million shares land and the free float roughly triples from 4.9% to approximately 11.8%. [5] If the stock moves less than roughly 5%, the supply was pre-priced, the market read the table long ago, and the entire "lock-up is where price gets discovered" thesis is dead on arrival — including any residue of it in ours.

Test two. Watch the earnings print itself, two trading days earlier. If SPCX moves more than roughly 10% on the numbers, that is value discovery arriving before the first share is unlocked — which is our call, not theirs.

And what would rescue the float thesis. If the first unlock produces a violent, sustained decline that dwarfs the reaction to the earnings themselves, we were wrong and the scarcity camp was right. We do not expect it. But it is the shape of the evidence that would move us, and we will not pretend otherwise if it arrives.

The last word belongs to the denominator, because that is what this report is actually about. When the market says SpaceX trades at ~95x sales, it is quoting a number built on $18,674m of revenue — a figure that exists only because two companies Elon Musk already controlled were retrospectively combined into every prior period at historical carrying amounts, with no purchase price, no goodwill, and no separate financials ever published. [10] The company the public thinks it bought did $15,473m. [9] The analysts covering it disagree about its value by a factor of more than twelve — $63 against $800. [22] [28]

Price discovery in SPCX is not waiting on a lock-up table. It is waiting on the market deciding which company it bought.