Space Exploration Technologies SPCX

> # ⚑ EARNINGS TODAY — 2026-08-04, AND A 10-Q FILED THE SAME DAY

>

> SpaceX reported second-quarter 2026 results this morning and filed its 10-Q for the quarter ended 2026-06-30 on the same day. Unlike several names in this batch, the quarterly financial statements ARE in the archive, so the numbers below are filing-verified rather than inferred.

>

> - Revenue $7,814M against a $6,826.7M consensus — a 14.5% BEAT, and +91.9% against $4,071M a year earlier.

> - Loss of $0.09 per share against an expected loss of $0.225 — a 60% narrower loss than consensus.

> - Operating loss narrowed to $143M from $970M. Gross margin 55.3% against 43.9%.

> - The stock rose 10.07% to $126.06 on 113,631,645 shares — roughly $14.3 billion of turnover.

> - And it is still 6.6% BELOW the $135.00 IPO price, fifty days after listing.

What this dive was commissioned to verify — and what it found

Three corporate actions were flagged for verification against the now-complete archive. Here is the result, before anything else.

1. The IPO — VERIFIED EXACTLY, and the payload's date is wrong

> "On June 15, 2026 the Company completed its IPO, in which the Company issued and sold 638,888,888 shares of its Class A common stock at a public offering price of $135.00 per share, including the full exercise of the underwriters' over-allotment option, which resulted in net proceeds of $85,675 million after deducting the underwriting commissions and offering costs of $575 million. The shares sold in the offering were registered under the Securities Act pursuant to our Registration Statement on Form S-1 (File No. 333-296070) which was declared effective by the SEC on June 11, 2026." — 10-Q, Item 2

Date confirmed: 2026-06-15. Price confirmed: $135.00. Net proceeds $85,675 million on $575 million of costs — a 0.67% cost ratio. The prospectus was filed 2026-06-12 under Rule 424(b)(4).

The vendor profile.ipoDate reads 2026-06-12, which is the prospectus date, not the completion date. The filing wins and the dive uses 2026-06-15.

The stated use of proceeds is worth quoting because it is the strategy in one sentence: "to fund its growth strategy, including the expansion of the Company's AI compute infrastructure, enhancements to the Company's launch infrastructure and launch vehicles, increases in the scale and capacity of the Company's satellite constellations, and any remaining amounts for general corporate purposes." Compute is named first.

2. The $25.0 billion notes offering — VERIFIED EXACTLY, with two new facts

The 2026-06-23 8-K: "On June 23, 2026, the Company priced its previously announced Offering of $7.0 billion of 5.350% Senior Notes due 2031, $6.0 billion of 5.650% Senior Notes due 2033, $6.0 billion of 5.875% Senior Notes due 2036, $2.5 billion of 6.600% Senior Notes due 2046, and $3.5 billion of 6.650% Senior Notes due 2056."

The 2026-06-26 8-K confirms issuance under an indenture with The Bank of New York Mellon Trust Company, N.A.

Two facts the earlier flag did not contain. First, the notes were issued under Rule 144A and Regulation S — they are unregistered, sold only to qualified institutional buyers and non-US persons. Second, the proceeds were used "to repay the outstanding borrowings under its bridge loan facility in full" — so this was a refinancing of bridge debt rather than net new money, which is consistent with the cash flow statement showing $51,812M of debt proceeds against $39,396M of repayments in the six months, plus a $1,153M debt-extinguishment premium paid and a $1,545M non-cash loss on debt extinguishment. The refinancing cost approximately $2.7 billion between the premium and the accounting loss, and it is the single largest component of the $1,962M "other expense" line that turned a $2,086M operating loss into a $4,788M pre-tax loss for the six months.

Interest is payable semi-annually on 15 January and 15 July, beginning 2027-01-15.

3. The 2026-08-06 lock-up step-up — NOT VERIFIED, AND THE SOURCE IS NOT A FILING

This is the finding that matters most and it corrects the earlier dive.

Nothing in the complete SEC archive states that a lock-up step-up occurs on 2026-08-06, or on any other date. The 10-Q refers to a lock-up period only by reference to a document not in this archive: "the lock-up period (as defined in the Prospectus)." The Prospectus is not in the archive.

What the filings DO say about lock-ups points the other way for the two officers named:

> "Pursuant to the extended lock-up arrangements described in the Prospectus, Bret Johnsen, Chief Financial Officer, agreed to subject the vast majority of his shares to the extended lock-up period. On June 16, 2026, Mr. Johnsen... adopted a Rule 10b5-1 trading arrangement that does not commence sales until 2027 and covers up to 919,497 shares of Class A common stock of the Company, which are subject to the lock-up period."

> "Pursuant to the extended lock-up arrangements described in the Prospectus, Gwynne Shotwell, President, Chief Operating Officer and a director, agreed to subject the vast majority of her shares to the extended lock-up period. On June 23, 2026, Ms. Shotwell adopted a Rule 10b5-1 trading arrangement for up to 585,605 shares of Class A common stock... The arrangement... expires on June 30, 2027."

The actual source of the 2026-08-06 date is a knowledge-base claim:

> 2026-07-21 · BEARISH · conviction 60 · fact · compound_and_friends · speaker Josh Brown · role independent

> "Staggered lockup unlock (float goes 3%→12% two days after Aug 4 earnings, ~40% by Dec 8) means you'll get a chance to buy SpaceX under $100/share"

Two days after 2026-08-04 is 2026-08-06. That is where the date came from.

And the knowledge base contradicts itself on this exact point:

> 2026-06-05 · neutral · conviction 65 · fact · empire · speaker Ryan Zur

> "Significant holders locked up 366 days (~72% incl. Elon's 42%); classic low-float/high-FDV buoys price longer — no true price discovery until 2027."

> 2026-07-31 · bearish · conviction 48 · fact · darius_dale · speaker Adam Turnquist · role independent

> "SpaceX down ~50% since IPO, in line with historical ~50% median 12-month post-IPO drawdown; the ~20% of shares unlocking after earnings will act as an overhang — too early to call a bottom."

> 2026-06-09 · bearish · conviction 74 · fact · arthur_hayes · skill 1.2

> "SpaceX IPO at ~$1.8T / ~100x sales is a low-float high-FDV trap; float ratchets to ~25% by September so insiders dump July-October."

Four independent sources, four different unlock schedules: 3%→12% on 2026-08-06 rising to 40% by 2026-12-08; roughly 20% "after earnings"; roughly 25% by September; and 366-day locks with nothing before 2027. They cannot all be right and no filing in this archive settles it.

The honest conclusion, stated as the standing rule requires. The earlier dive's flag was directionally reasonable and was sourced to a named, independent, credible speaker — but it was presented against a filing archive that does not support it, and this dive downgrades it from a corporate action to an unresolved knowledge-base assertion. A reconciliation exists and is plausible: an extended 366-day lock-up for a subset of significant holders can coexist with a staggered early-release schedule for the rest, which is a common IPO structure. But we cannot verify it, so we do not assert it. One further correction: the darius_dale claim's assertion that SpaceX is "down ~50% since IPO" is FALSE on our data — at last night's close of $114.53 the stock was 15.2% below the $135.00 IPO price, not 50%. The 50% figure is the drawdown from the post-IPO HIGH, not from the IPO price, and the claim conflates them.

And two things NEITHER dive could have anticipated

A $60 billion acquisition, signed and pending. From the 10-Q:

> "In April 2026, the Company entered into an option agreement... providing for the right, but not the obligation, to acquire Anysphere, Inc., a San Francisco-based private software company doing business as Cursor... The consideration for the acquisition of Cursor would consist of shares of Class A common stock based on an implied equity value of Cursor of $60 billion, and the price of Class A common stock that equals the volume-weighted average closing price thereof over the seven consecutive trading days immediately preceding the closing of the acquisition. In June 2026, the Company exercised the call option and entered into an Agreement and Plan of Merger... The Company currently expects the Cursor Merger to close during the third quarter of 2026."

$60 billion of stock consideration, struck on a seven-day volume-weighted average close. At $126.06 that is approximately 476 million Class A shares, 3.6% dilution. At $100 it is 600 million shares, 4.6%. The consideration is fixed in dollars and floating in shares, which means every point the stock falls before closing costs existing holders more dilution. Nothing about this transaction appears anywhere in the vendor payload.

And a spectrum purchase. The 10-Q describes a License Purchase Agreement dated 2025-09-07 with Spectrum Business Trust 2025-1 and EchoStar Corporation for EchoStar's AWS-4 and H-Block licences covering 50 MHz, amended and restated 2025-11-05 to add up to additional unpaired AWS-3 spectrum, transferring in two steps through the trust. Also absent from the payload.

Reference table

1. The three segments — from the filing, because the payload has none

seg_prod and seg_geo are BOTH EMPTY for SPCX. Everything in this section comes from Note 15 of the 10-Q filed 2026-08-04.

> "Space Exploration Technologies Corp. and its wholly owned subsidiaries... operate three segments – (i) the Space segment designs, manufactures, and launches reusable rockets to provide high cadence, reliable, and affordable access to space at unprecedented scale, (ii) the Connectivity segment operates a worldwide high-speed, low-latency broadband network powered by thousands of Starlink satellites in Low-Earth Orbit... and (iii) the AI segment operates a vertically integrated AI platform spanning a frontier [model name redacted per house style], AI solutions for consumer and enterprise customers, X — a real-time information, entertainment, and free speech platform — and AI computational infrastructure."

Second quarter 2026, all figures in $M, filing-verified:

Six-month capital expenditure by segment — the single most important table in this dive:

Read those two tables together and the company resolves into something specific: a highly profitable satellite broadband business, a loss-making rocket programme that is mostly an internal input to the broadband business, and a compute build that is consuming 83 cents of every invested dollar while losing money.

Revenue detail, one level down (Q2 2026 against Q2 2025, $M):

The advertising line is the only revenue line in the entire company that is shrinking, down 13.8% to $367M — the social platform inside the AI segment. The segment's 247.5% growth is entirely the compute and solutions line.

Operating metrics the company publishes, and they are the honest cross-check on the revenue:

Three observations that the revenue table alone would hide.

First, Connectivity's growth is subscribers, and price is going the other way. Subscribers doubled; revenue grew 66%; ARPU fell 22.4% from $85 to $66. That is the arithmetic of a network expanding into lower-price geographies and plans, and it is normal for this kind of business — but it means the 65.8% growth rate decays as the mix shifts unless subscriber growth stays near 100%.

Second, the Space segment is going backwards on volume while growing on revenue. Launches down 17.8%, mass to orbit down 25.6%, and internal launches down 25.0% — yet Launch Services revenue rose 32.2%, because customer launches rose from 9 to 10 and are the only ones that produce external revenue. The 10-Q is explicit: "Our Space segment revenue only reflects our customer launches and customer activities." A quarter with fewer launches and a lower internal deployment rate is a quarter in which the constellation grew more slowly.

Third, the compute build is real and it is enormous. Nameplate draw of 1.4 gigawatts, up 3.5x. $23,551M of capital expenditure in six months in one segment. For scale, that single half-year figure is larger than the entire fiscal 2025 capital expenditure of the whole company ($20,737M per the payload).

2. The income statement, filing-verified

From the 10-Q, in $M except per-share:

Four things to read from this.

The June quarter was close to operating breakeven. A $143M operating loss on $7,814M of revenue is −1.8%, against −23.8% a year earlier. Gross margin improved 11.4 points to 55.3%. This is the strongest single fact in the bull case and it is filing-verified.

Research and development is 45.4% of revenue and rising in absolute terms — $3,548M in one quarter against $1,958M. Of that, $2,178M is the AI segment and $1,076M is Space. This is not an expense that is going to fall; it is the strategy.

The six-month "other expense" of $1,962M is almost entirely two non-operating items: the $1,545M loss on debt extinguishment from the June refinancing and a $539M unrealised loss on digital assets (the balance sheet carries $1,098M of digital assets, down from $1,637M). Neither recurs, and neither is operational.

The six-month loss attributable to shareholders ($5,488M) exceeds the net loss ($4,817M) by $671M, which is the effect of the preferred-stock conversion at the IPO. That $671M is why the six-month loss per share is $1.12 rather than $0.99, and it is a one-time capital-structure item.

A payload reconciliation worth recording. The vendor's inc_q reports Q1 2026 diluted EPS of −$1.27, while earn_cal reports the Q1 actual as −$1.19. The filing supports the vendor: six-month loss attributable of $5,488M less the June quarter's $541M is $4,947M, over Q1's 3,884M weighted shares, is −$1.27. inc_q is right and earn_cal is wrong on that quarter.

3. Cash — the burn, and why it does not threaten solvency

From the 10-Q's consolidated statement of cash flows, six months ended 2026-06-30, in $M:

Free cash flow of MINUS $25,010M in six months, and the true figure is worse. The supplemental disclosures add $5,513M of property, plant and equipment sitting in accrued expenses and accounts payable and $3,921M of property, plant and equipment financed by other financings. Economic investment in the half was therefore well above $28.5 billion; the cash line understates it.

Interest paid, net of capitalised, was $1,667M in six months against $603M — a 176% increase reflecting the debt build.

Why this does not threaten solvency, stated plainly because the burn number invites the opposite conclusion. At 2026-06-30 the company held $93,522M of cash and equivalents plus $6,487M of marketable securities — $100,009M — against $39,364M of debt and finance leases, giving net cash of $60,645M. The 10-Q states: "As of June 30, 2026, we and our subsidiaries had outstanding $38,433 million in aggregate principal amount of indebtedness and no material debt principal payments are due until July 15, 2031" and "we believe we have sufficient sources of funding to meet our business requirements for at least the next twelve months." At the current half-year burn rate the net cash alone funds roughly twelve months and the gross cash roughly two years, before any operating improvement, and the debt does not come due for five years. The risk here is dilution and valuation, not insolvency.

One disclosure worth naming: $13,329M of the debt is related-party — $2,039M current and $11,290M non-current, against $455M and $4,052M at year-end. The 10-Q flags related-party amounts inside both the debt lines and the property, plant and equipment line ($329M of the six-month capital expenditure was related-party). A company with $13.3 billion of related-party borrowings deserves the disclosure to be read, and this archive contains only the balance-sheet parentheticals, not the Note 17 detail.

The capital-expenditure check — partially verified, and the limitation is stated. This archive contains NO 10-K for SPCX, because the company has not yet had a fiscal year end as a public registrant. The vendor's fiscal 2025 capital expenditure of $20,737M therefore CANNOT be filing-verified. What can be said: the 10-Q's filed six-month figures of $28,476M (2026) and $6,965M (2025) are consistent with a full-year 2025 figure in the $20 billion region and with a sharply accelerating 2026, and the implied capex-to-depreciation ratios (3.09x for FY2025 on the payload, 5.38x for H1 2026 on the filing) are high but entirely plausible for a company building a satellite constellation and gigawatts of datacentre. The datum is directionally corroborated and not year-verified, and this dive uses the FILED six-month figures for every cash-flow conclusion rather than the unverifiable annual one.

4. Balance sheet and share count — both corrected from the filing

From the 10-Q, in $M:

Property, plant and equipment rose 54.3% in six months, from $42,602M to $65,736M. Equity rose from $2,573M to $127,224M, of which $85,675M is the IPO and $37,475M is the conversion of preferred stock in the common-control merger.

The share count, and the vendor is wrong about it. The 10-Q cover page states: "As of July 28, 2026, the registrant had 7,696,293,669 shares of Class A common stock and 5,485,486,276 shares of Class B common stock outstanding." Total: 13,181,779,945. The vendor's market capitalisation of $1,648,344M at $126.06 implies 13,075,865,834 — 105.9 million shares, 0.8%, too few. True market capitalisation is approximately $1,661,691M, and every valuation figure in this dive uses the filed count.

Only Class A is listed. The cover page shows "Class A Common Stock, $0.001 par value | SPCX | The Nasdaq Stock Market LLC, Nasdaq Texas, LLC." Class B — 41.6% of shares outstanding — does not trade. The insider file shows the chief executive holding 3,788,654,145 shares after five C-Conversion transactions on 2026-06-15, which is 28.7% of the total count.

Capital-structure history, all disclosed in Note 1 and all absent from the payload:

• A five-for-one forward stock split effected in May 2026, applied retroactively to all periods.
• A common-control merger completed 2026-02-02, in which the company acquired an artificial-intelligence holding company (legal names redacted per house style — the transaction is described in Note 1 of the 10-Q as a common-control merger), which had itself acquired a social-media holding company on 2025-03-28.
• All redeemable convertible preferred stock — 2,046 million shares at year-end — converted at the IPO.

Because the mergers were between entities under common control, ALL prior-period figures are retroactively restated. That is why the payload's inc_a shows fiscal 2024 revenue of $14,015M and net income of +$791M for a company that most observers would not have described that way, and it is why the annual series cannot be compared to any pre-merger view of the business.

5. Valuation — computed from filings, because the ratio blocks are empty

km_ttm and ratios_ttm are both EMPTY arrays for SPCX. There is no vendor enterprise value, no vendor multiple, no vendor return metric. Every figure below is computed from the 10-Q and is labelled as such.

Enterprise value:

Multiples, all computed:

Price to book: $1,661,691M / $127,224M = 13.06x.

Price to FY2028E EPS of $3.771: 33.4x — on an estimate whose analyst range runs from MINUS $0.45 to $6.47.

The FY2026 consensus is arithmetically unreachable, and this invalidates the estimate array

Consensus FY2026 revenue is $38,930M, from 17 analysts. First-half actual revenue was $12,508M. That leaves $26,422M for the second half — 2.11 times the first half, or an average of $13,211M per quarter against the June quarter's $7,814M. Every quarter of the second half would need to be 69% above the June quarter, sequentially, twice. Nothing in the operating metrics supports that: launches fell, mass to orbit fell, subscriber growth is already at 100% and ARPU is falling.

We therefore treat the entire est array as unreliable on the level, while noting that the direction — very high growth — is corroborated by the filings. The consequence for this dive is that the FY2027 and FY2028 revenue figures ($74,200M and $141,593M) are used only as scenario markers, never as forecasts, and that is stated wherever they appear.

est.ebitdaAvg and est.ebitAvg are rejected outright for the fixed-ratio fabrication signature. In every year — FY2025, FY2026, FY2027 and FY2028 — ebitdaAvg is exactly 66.88% of revenueAvg and ebitAvg is exactly 25.48%. A 66.9% EBITDA margin for a company that builds and launches rockets is not a forecast; it is a formula. The check against actuals is decisive: the FY2025 row estimates EBITDA of $12,908.6M against the payload's own actual of $4,289M — three times too high. All valuation uses revenue and EPS only.

Estimate coverage: FY2026 rests on 17 analysts for revenue and 14 for EPS; FY2027 on 17 and 15; FY2028 on 20 and 18. The grades block, however, reports only FOUR analysts (4 buy, 0 hold, 0 sell). The two coverage blocks in the same file disagree by a factor of five, and neither can be reconciled to the other. The FY2028 EPS range of −$0.45 to $6.47 — a spread in which the low case is still a loss — is the most honest single number in the estimate array.

5a. What today's price assumes (the inversion)

At $126.06 — an enterprise value of roughly $1.60 trillion, 51.2x annualised second-quarter revenue, with free cash flow of minus $25.0 billion for the half — the price embeds:

• The AI segment is worth roughly $1.2 to $1.4 trillion on its own. (Our derivation; see the sum-of-parts below.) This is the most fragile assumption in the price and it is not close. That segment generated $2,561M of revenue and a $1,257M operating loss in the June quarter and consumed $23,551M of capital in six months. A knowledge-base claim from Jeremy Grantham asserts that "~90% of value rests on AI"; our own arithmetic, done independently, puts it between 78% and 90%. Two methods, same answer.
• Connectivity's 38.6% operating margin holds while ARPU keeps falling. (Our derivation from the 10-Q.) Subscribers +100%, revenue +65.8%, ARPU −22.4%. The margin has to absorb a continuing mix shift into cheaper plans and geographies.
• Revenue compounds at something close to the consensus path. (Consensus, labelled unreliable.) $74.2B in 2027 and $141.6B in 2028 against $31.3B annualised today — a 4.5x in thirty months. The FY2026 leg of that path is already arithmetically unreachable, which is the strongest available evidence that the array has not been re-based.
• The float schedule does not overwhelm the price. (Knowledge-base sourced; NOT filing-verified.) If the Josh Brown account is right, tradeable supply quadruples in two days and rises roughly thirteen-fold by December. At 3% float the current price is a marginal-buyer price; at 40% it is a market price, and the two need not be similar.
• The $60 billion Cursor consideration does not repricing-loop. (Filing-verified.) The consideration is fixed in dollars and floating in shares, struck on the seven-day volume-weighted average before closing. A lower share price means more shares, which means more dilution, which is a mildly self-reinforcing mechanism in a falling market.

5b. The return bridge (why the multiple moves)

There is no conventional return bridge available here and constructing one would be false precision. The company loses money, generates deeply negative free cash flow, pays no dividend, and its consensus earnings estimate for two years out spans a loss of $0.45 to a profit of $6.47.

What can be said honestly. Expected return over any horizon decomposes as revenue growth × terminal margin × terminal multiple − dilution, and three of those four terms are unknown. Revenue growth is the only one the filings pin down: +91.9% in the quarter, +53.7% for the half.

The multiple assumption in our base is explicit and it is COMPRESSION. Our base of $105 implies an enterprise value of roughly $1,323 billion, which is 42.3x annualised second-quarter revenue — down from 51.2x today — and 17.8x the FY2027 consensus revenue that we have separately labelled unreliable. We are assuming a lower multiple on a higher revenue base, and the net is a lower price. Anyone underwriting a higher number is either assuming the multiple holds at 51x or assuming a revenue path materially above the consensus.

A dilution note that belongs in the bridge. Weighted average shares rose from 2,929M (Q2 2025) to 5,864M (Q2 2026) — a doubling — and the count at 2026-07-28 is 13,182M. Some of that is the retroactive restatement and the preferred conversion rather than economic dilution, but the Cursor merger adds roughly 476 million more shares at today's price, and the company issued 8,510,615 unregistered Class A shares on option exercises and granted 25,917,613 restricted stock units in the ten weeks to 2026-06-12 alone. Share-based compensation was $1,470M in six months, 11.8% of revenue.

5c. Variant perception (where we differ, what would surprise)

• We differ from the street enormously and in the unfashionable direction. Our base of $105 is 50.9% below the consensus target of $214. But note what "consensus" means here: FOUR analysts in the grades block, with a target range from $115 to $401 — a 3.5x spread. There is no consensus on this security; there is a distribution, and it is nearly uniform.
• Our sum-of-observable-parts independently reproduces the Grantham claim, and that is the most useful thing in this dive. Value Connectivity at 15x annualised revenue ($258B, which for a 66%-growth 38.6%-margin network is generous), Space at 20x annualised revenue ($77B, generous for a loss-making segment), the AI segment at 20x annualised revenue ($205B), and add $61B of net cash: approximately $601 billion, or $46 per share. The market pays $1,662 billion. The $1,061 billion difference is a claim about the AI segment's future, and it is 64% of the market capitalisation on those (generous) assumptions — and rises to roughly 78-90% if the parts are valued more conventionally. Watchable number: the AI segment's operating loss in the Q3 print. It was $1,257M this quarter and $1,524M a year ago.
• We think the Space segment's operating metrics are being under-read. Launches down 17.8%, mass to orbit down 25.6%, internal launches down 25.0%, and a single Starship launch in the quarter against three in the first half of 2025. The entire orbital-compute case — the most valuable idea in the bull thesis — requires Starship rapid reusability, and cadence went backwards.
• Where we have NO variant perception: the business quality of Connectivity. A 38.6% operating margin on a network doubling its subscriber base annually is exceptional and nothing in our analysis disputes it. Our disagreement is entirely about what the rest of the company is worth.
• Positive surprise that would force a re-rate: the AI segment reaching operating breakeven; a Starship cadence step-up with demonstrated rapid reusability; Starlink ARPU stabilising above $66 while subscribers keep doubling; or the float step-up occurring and being absorbed without a material decline, which would falsify the supply thesis outright.
• Negative surprise that would break the thesis: the float step-up occurring and the price breaking $108.37 (the low in the tech series, 14.0% below spot); the Cursor merger failing to close or closing at a materially lower share price and therefore higher dilution; a capital raise despite $60.6 billion of net cash, which would signal the burn is faster than disclosed; or Connectivity's margin falling below 30%.

Synthos fair values

Given that the earnings estimates span a loss and a profit in the same year, all three anchors are built from the FILED segment economics and cross-checked against the consensus revenue path.

• Bear ~$55 — the sum of observable parts. Connectivity at 15x annualised revenue ($258B), Space at 20x ($77B), AI at 20x ($205B), plus $61B of net cash, less roughly $60B of Cursor consideration = approximately $541B, or $41 per share, rounded up to $55 to allow for the deferred-revenue backlog and the option value in Starship. Cross-check: Josh Brown's "under $100"; 49.2% below the low tech reading of $108.37; 17.6x annualised Q2 revenue. The scenario: the float steps up, the AI premium compresses toward the operating economics, and the market prices what the segments currently earn. −56.4%.
• Base ~$105 — an enterprise value of roughly $1,323B, 42.3x annualised Q2 revenue and 17.8x the FY2027 consensus revenue. Cross-check: 27.8x the FY2028 consensus EPS of $3.771; 22.2% below the $135.00 IPO price; 10.9x book. The scenario: revenue keeps compounding fast but short of the consensus path, Connectivity's margin holds, the AI segment narrows its loss, and the multiple compresses modestly as float rises. −16.7%.
• Bull ~$210 — 32.5x the FY2028 consensus HIGH EPS of $6.47, and roughly the post-IPO high of $201.80 to $225.64. Cross-check: an enterprise value of roughly $2,708B, 86.6x annualised Q2 revenue and 19.1x FY2028 consensus revenue; 13.5% above empire's published sum-of-parts fair value of $185. The scenario: the AI segment reaches profitability at scale, Starship achieves rapid reusability and puts orbital compute in play, and the float absorbs without incident. +66.6%.

Base is 16.7% BELOW spot; asymmetry roughly 1.18:1 (56.4% down, 66.6% up), with no dividend and no buyback. A base below spot with a payoff ratio under 1.2:1 is not a Buy under any reading. Section 11 explains why it is nonetheless a Watch rather than an Avoid.

6. Knowledge base — 132 claims, two quarantine duplicates, and total dispersion

Raw hits: 210. After a case-sensitive entity re-run: 132. Text-only discards: 78. Further discarded as confirmed quarantine duplicates: 2. Claims used: 130.

The sweep ran SPCX, SpaceX, Space Exploration Technologies, Starlink, Starship and Falcon 9 across all 52,021 distilled claims. This is by a wide margin the largest lane in this programme.

The homograph test PASSED: zero case-insensitive-only matches.

THE DEFECT FOUND — the quarantine slug leaked duplicates again. Two claims dated 2026-05-23 sit under the channel slug quarantine_misattributed, with speaker: Dan and skill 2.0:

> "Assigns 80-85% odds that SpaceX and Tesla merge by 2027, converging their data to close the model gap versus [lab names redacted]." (bullish, 80)

> "SpaceX defines a new space sector—orbital data centers—as a 2nd/3rd-derivative AI play; 80% of its value is future vision, a 5-10-15 year theme." (bullish, 70)

Both are near-verbatim restatements of two Dan Ives claims filed under anthony_pompliano_show three days earlier on 2026-05-20:

> "80-85% odds SpaceX merges with Tesla by 2027; data convergence lets Musk narrow the model gap versus [lab names redacted]." (bullish, 82)

> "SpaceX defines a new space sector via orbital data centers; a 2nd/3rd/4th-derivative play off the AI revolution, a 5-10-15 year theme, 80% of value is future." (bullish, 70)

Both quarantined copies are EXCLUDED and the views counted once. At skill 2.0 — double the store's baseline — counting them twice would have materially distorted a lane where direction is the whole question. This is the SECOND name in this batch on which quarantine_misattributed returned a duplicate to an entity sweep, which suggests the quarantine is not being filtered out of retrieval.

The bulls, named and extreme

> 2026-07-24 · bullish · conviction 96 · thesis · compound_and_friends · speaker Ron Baron · role independent

> "SpaceX is Baron's largest position (~$25B of $70B); expects 20-30x from IPO price over next 10-15 years and a $20-30T (at least $40T) valuation — biggest company in the world."

> 2026-06-11 · bullish · conviction 85 · thesis · bill_gurley · speaker Brad Gerstner · role independent

> "SpaceX is a must-buy, must-own, set-it-and-forget-it for institutional investors — the best single bet on both the space and AI future."

> 2026-06-15 · bullish · conviction 60 · thesis · gavin_baker · speaker Gavin Baker · role independent

> "Once Starship is reusable, orbital compute math is compelling: a gigawatt in space ~$30B (no $25B power/cooling) versus ~$60B terrestrially."

> 2026-06-05 · bullish · conviction 90 · thesis · empire · speaker Ryan Zur

> "Biggest position, accumulating since 2018; probabilistic sum-of-parts across 12 business lines, any one hitting makes it multi-trillion."

> 2026-06-05 · bullish · conviction 60 · fact · empire · speaker Ryan Zur

> "His conservative sum-of-parts pegs fair value at $185 versus the ~$1.77T IPO price; thinks it likely reaches that this summer."

> 2026-06-11 · bullish · conviction 60 · thesis · bill_gurley · speaker Andrew Fox

> "Starlink broadband is under 1% of global households; it can scale to hundreds of millions of terminals and users."

Note that the most valuable bull claims are the specific, checkable ones rather than the trillion-dollar ones. The Starlink penetration claim, the orbital-compute cost arithmetic and the launch-reusability dependency are all statements that the filings will eventually test. Ron Baron's $20-30 trillion is not falsifiable on any horizon this dive can grade, and it is reported at that weight — as a disclosed position size (roughly 36% of a $70 billion book) rather than as an analytical input.

The bears, equally named and equally specific

> 2026-06-18 · BEARISH · conviction 72 · fact · odd_lots · speaker Jeremy Grantham

> "SpaceX at ~$2.7T on ~$20B of 2025 sales (price/sales >100) is a classic bubble stock whose ~90% of value rests on AI; check the numbers and avoid the hype."

> 2026-06-09 · BEARISH · conviction 74 · fact · arthur_hayes · skill 1.2

> "SpaceX IPO at ~$1.8T / ~100x sales is a low-float high-FDV trap; float ratchets to ~25% by September so insiders dump July-October."

> 2026-06-16 · BEARISH · conviction 60 · fact · compound_and_friends · speaker Josh Brown

> "SpaceX's ~$2.64T cap is 'phantom'—only 3% of float trades; once ~$1T of shares unlock in weeks it can't hold; expects it lower in a year, like the average IPO down 50%."

> 2026-05-22 · BEARISH · conviction 50 · thesis · empire · role independent

> "There is no fundamental way to price SpaceX anywhere near its ~$2T pre-IPO level; it's not profitable, so net-shorting it despite betting on Elon."

> 2026-06-19 · BEARISH · conviction 68 · thesis · empire · speaker Jordy Alexander

> "Traded long SpaceX, now flat and waiting to put on the inevitable short — a low-float 'shitcoin' facing month after month of share unlocks."

> 2026-06-28 · BEARISH · conviction 82 · thesis · real_vision · speaker Ron

> "SpaceX is a retail sentiment-driven name at 130x price-to-sales; Starlink and launch face competition (Bezos, Rocket Lab); when the share unlock comes into view it tops and won't end well for investors."

Grade the bears on what has happened since. The stock has fallen from a post-IPO high of $201.80 to $126.06 — 37.5% — and is 6.6% below its IPO price. Grantham's price-to-sales figure of "greater than 100" was struck on 2025 sales; on the June quarter annualised it is now 51.2x, so the multiple has halved partly through the price falling and partly through revenue doubling. Arthur Hayes's "insiders dump July-October" window is now open. Josh Brown's "under $100" is 20.7% below today's close. None is yet proven and all remain live.

The most useful claim in the lane, and it is neutral

> 2026-07-14 · neutral · conviction 40 · thesis · compound_and_friends · speaker Michael Batnick · role independent

> "After round-tripping from a 135 IPO to 217 and back to 136 in under a month, I have no interest in buying SpaceX today — no idea where it settles, not for me."

That is a fair description of the epistemic position, and it is closer to this dive's own than either the $20-trillion claim or the sub-$100 claim. A security that moved from $135 to $217 to $136 in four weeks, on 3% float, with the estimate array internally unreachable and the ratio blocks empty, is not being priced by analysis.

The one claim about governance

> 2026-07-29 · BEARISH · conviction 70 · thesis · we_study_billionaires · speaker Cory Doctorow · role independent

> "Musk stuffs money-losing [model and lab names redacted] into SpaceX and IPOs it straight into the NASDAQ 100, forcing index funds to buy at any price so insiders and backers offload onto pension holders."

We report it because it is the sharpest statement of a real structural fact, and we separate the fact from the motive. The fact is filing-verified: a common-control merger completed 2026-02-02 folded a loss-making artificial-intelligence business into the registrant, and that segment lost $1,257M in the June quarter while consuming $23,551M of six-month capital expenditure. The motive is an assertion about intent that no filing can test, and this dive does not adopt it.

Attribution across the lane. Unusually for this programme, most of the high-conviction claims carry a NAMED speaker — Ron Baron, Jeremy Grantham, Brad Gerstner, Gavin Baker, Arthur Hayes, Josh Brown, Michael Batnick, Cory Doctorow, Ryan Zur, Andrew Fox, Chamath, Thomas Laffont, Bill Ackman, Dan Ives, Adam Turnquist, Morgan Brennan. speaker_role reads independent on most and journalist on two Morgan Brennan claims; it is not relied on and content was checked in each case. NO claim in this lane is management voice. The 78 text-only discards are largely claims about the broader private-market and AI-listing environment that mention the company in passing.

Conclusion. Breadth 132, claim count 130, net conviction undefined. This is the first name in this programme where the correct summary of the knowledge base is that it does not have a view — it has two views, held with high conviction by named, credible, independent people, four weeks apart, that differ by a factor of more than twenty. Reporting a net conviction score would be a fabrication and none is given.

7. Data integrity — the least usable payload in this programme

Twelve findings. On no other name in this programme have five separate blocks come back empty.

TRUNCATION 1 — km_ttm is an EMPTY ARRAY. There is no enterprise value, no return on invested capital, no free cash flow yield, no working capital, no per-share metric of any kind. This is the km_ttm truncation defect this batch was directed to test for, and it is present in its most complete form. We say so rather than silently substituting: every enterprise-value and multiple figure in this dive is computed by hand from the 10-Q and is labelled.

TRUNCATION 2 — ratios_ttm is an EMPTY ARRAY. No margin, no liquidity ratio, no leverage ratio, no return metric, no price multiple. Same treatment: computed from the filing or not stated.

TRUNCATION 3 — seg_prod is an EMPTY ARRAY.

TRUNCATION 4 — seg_geo is an EMPTY ARRAY. For a company with three reportable segments whose economics diverge violently — one at a +38.6% operating margin and two at −49% and −56% — having no segment data at all is the single most consequential gap in the file. Everything in Section 1 comes from Note 15 of the 10-Q.

TRUNCATION 5 — peers is an EMPTY ARRAY. No comparison universe is supplied and none is constructed. No peer multiple appears anywhere in this dive.

DEFECT 1 — the share count is understated by 105.9 million shares. The vendor's market capitalisation of $1,648,344M at $126.06 implies 13,075,865,834 shares against the 10-Q cover page's 13,181,779,945 (7,696,293,669 Class A plus 5,485,486,276 Class B) at 2026-07-28. A 0.8% understatement, worth approximately $13.3 billion of market capitalisation. The filing wins and every figure in this dive uses 13,181.78M.

DEFECT 2 — profile.ipoDate reads 2026-06-12; the 10-Q says the IPO completed 2026-06-15. 2026-06-12 is the date the prospectus was filed under Rule 424(b)(4); the registration statement was declared effective 2026-06-11. A three-day error on the single most important date in this security's short history.

DEFECT 3 — inc_q contains only THREE quarters, and they are not consecutive. June 2026, March 2026 and March 2025. The June, September and December 2025 quarters are absent entirely, so no trailing-twelve-month figure can be built from the payload and none is attempted.

DEFECT 4 — inc_q reports Q2 2026 interest expense as NEGATIVE $629,000,000, against a positive $664,000,000 for Q1 2026. The filing shows the same quantity as "(629)" in a costs-and-expenses presentation. The vendor has carried the parenthetical through as a sign on one quarter and not the other, so the field is not internally consistent across the series.

DEFECT 5 — inc_q reports Q2 2026 depreciation and amortisation as ZERO. The 10-Q's cash flow statement shows $5,290M for the six months and inc_q itself shows $2,442M for Q1, so the June quarter was approximately $2,848M. A zero in the largest depreciation quarter in the company's history.

DEFECT 6 — inc_a fiscal 2024 EPS and share count are both wrong. The row reports epsDiluted of $0.0018 on 9,956,000,000 weighted diluted shares against net income of $791,000,000. $791M over 9,956M shares is $0.079, not $0.0018 — a factor of 44. And the 9,956M share count is inconsistent with 2,759M in fiscal 2023 and 2,926M in fiscal 2025, in a period containing no reverse split. The fiscal 2024 row is rejected.

DEFECT 7 — the entire annual series is retroactively restated for common-control mergers and cannot be compared to any pre-merger view. Note 1 discloses a merger completed 2026-02-02 that folded in an artificial-intelligence holding company which had itself acquired a social-media holding company on 2025-03-28. Because the transactions were between entities under common control, prior periods are restated as though the businesses had always been combined. The payload gives no indication of this anywhere, so a reader comparing fiscal 2024 revenue of $14,015M to any externally-published figure for the rocket company alone will not reconcile.

DEFECT 8 — a five-for-one forward stock split effected in May 2026 is nowhere in the payload. Note 1: "In May 2026, the Company effected a five-for-one forward stock split of its authorized, issued, and outstanding shares of Class A, Class B, and Class C common stock." The payload's series appear to be post-split, but nothing in the file says so.

DEFECT 9 — est.ebitdaAvg and est.ebitAvg carry the fixed-ratio fabrication signature, and the actual proves it. ebitdaAvg is exactly 66.88% of revenueAvg and ebitAvg exactly 25.48% in FY2025, FY2026, FY2027 and FY2028. The FY2025 row estimates EBITDA of $12,908.6M against the payload's own reported actual of $4,289M — a 201% overstatement. Rejected in full.

DEFECT 10 — the FY2026 revenue consensus of $38,930M is arithmetically unreachable, requiring $26,422M in the second half against $12,508M delivered in the first. This is not a vendor formatting defect; it is a stale or unrebased estimate array, and it means the FY2027 and FY2028 revenue figures cannot be treated as forecasts.

DEFECT 11 — the two analyst-coverage blocks disagree by a factor of five. grades reports 4 buy, 0 hold, 0 sell — four analysts. est reports 17 to 20 analysts on revenue and 14 to 18 on EPS. Both cannot be right and neither can be reconciled from the file. The pt block's range of $115 to $401 — a 3.5x spread on a consensus of $214 — is consistent with the smaller number.

DEFECT 12 — one insider record is entirely null. The 2026-06-16 entry for Roelof Botha (director) carries transactionType: "", securitiesTransacted: 0, price: 0, securitiesOwned: 0 and acquisitionOrDisposition: "". A malformed record that conveys nothing.

Also recorded: tech.sma50 and tech.sma200 are null and ret_3m, ret_6m and ret_12m are null, because the security has fifty days of trading history. This is CORRECT behaviour and it is reported as an absence rather than filled in. tech.last of $125.33 differs from quote.price of $126.06 by 0.6%, and quote.yearHigh/yearLow ($225.64/$107.01) differ from tech.hi52/lo52 ($201.80/$108.37) by 10.6% and 1.3% — the largest quote-versus-technical divergence in this batch; tech is used for the drawdown figures and both are stated. profile.beta reads 0, which is correct in the sense that there is insufficient history to compute one, and is used nowhere.

Vendor composite rating — noted and disregarded. D+ / 1 overall, with 1 out of 5 on discounted cash flow, return on equity, return on assets, price-to-earnings and price-to-book. Every one of those scores is computed on a loss-making company with fifty days of listed history and an empty ratio block. The rating is not wrong so much as vacuous, and it carries no weight.

Non-equity tripwire — checked and passed, with a share-class caveat that matters. SPCX is Class A Common Stock, $0.001 par value, listed on The Nasdaq Stock Market LLC and Nasdaq Texas, LLC. But only Class A is listed: 7,696,293,669 Class A shares trade and 5,485,486,276 Class B shares — 41.6% of the total — do not. Every market-capitalisation figure in this dive multiplies the price of the listed class by the total of both classes, which is the standard convention and is stated here because the unlisted class is unusually large. Price of $126.06 is not par-like; volume was 113.6M shares (roughly $14.3B of turnover); the range from $108.37 to $201.80 in fifty trading days is an 86% band. This is common equity.

8. Technicals — and what cannot be computed

• Price $126.06, up 10.07% on earnings day, on 113,631,645 shares — roughly $14.3 billion of turnover, by far the highest in this batch. Opened $117.33, ranged $115.72 to $126.71, closed within 0.5% of the high.
• −37.9% from the post-IPO high of $201.80 (tech) or −44.1% from the quote series' $225.64; +15.7% above the low of $108.37.
• 6.6% BELOW the $135.00 IPO price of 2026-06-15, fifty days after listing.
• NO 50-day or 200-day moving average exists. tech.sma50 and tech.sma200 are both null. This dive reports that absence rather than substituting a shorter window, because a 50-day average on fifty days of data would be meaningless.
• RSI 42.1 — below neutral even after a 10% day. MACD −10.42, deeply negative.
• NO 3-month, 6-month or 12-month relative return is computable and none is given. The comparison to SPY's +24.3% and QQQ's +30.8% over twelve months cannot be made and is not made.
• Sentiment: 4 buy, 0 hold, 0 sell — four analysts in the grades block; consensus target $214 (+69.8%), median $217.50, high $401 (+218%), low $115 (−8.8%). A 3.5x spread between the high and low targets is the widest in this batch by a large margin and is the correct statistical description of a security nobody can value.

Today's move and what it does to the entry

SPCX closed 2026-08-04 at $126.06, up $11.53, on the day it reported a quarter that beat consensus revenue by 14.5% and grew 91.9% year on year, and on the day it filed a 10-Q disclosing a $60 billion pending acquisition.

The honest read: today's 10% move is a response to genuinely good operating news, and it leaves the stock still below its offering price with a possible supply event two days away that we cannot verify. At $114.53 last night the base case was −8.3%; at $126.06 tonight it is −16.7%. Today's session moved the security further away from our number, not closer to it.

9. Insiders — one holder, and a null record

Seven of the eight records belong to one person, and they contain no purchase or sale of any kind. All five 2026-06-15 entries are C-Conversion transactions — the automatic conversion of preferred and other classes into common stock at the IPO, disclosed in Note 1 of the 10-Q as an automatic event. They are not decisions. The two 2026-02-02 A-Award entries of 532,689,090 and 511,289,725 shares coincide with the common-control merger date and are the share exchange, not compensation in the ordinary sense.

The single fact worth extracting: the chief executive holds 3,788,654,145 shares. Against 13,181,779,945 shares outstanding that is 28.7%, worth approximately $477.5 billion at $126.06. A knowledge-base claim (empire, 2026-06-05) puts his locked-up holding at 42%, which does not reconcile with 28.7% on the filed count and may include other vehicles; we report both and reconcile neither.

What the file does not contain: a single open-market purchase or sale by any officer or director. The 10-Q's Item 5 supplies what the insider block cannot: the Chief Financial Officer and the President and Chief Operating Officer have each adopted 10b5-1 plans, covering 919,497 and 585,605 Class A shares respectively, that do not commence sales until 2027. Those two facts — no open-market activity, and officer sale plans deferred to 2027 — are mildly favourable and are the only insider signal available.

The Roelof Botha record is malformed and is reported as a data defect rather than as a transaction.

10. Verdict, kill-criteria and flip conditions

Watch.

Say the uncomfortable thing first: our base fair value of $105 is 16.7% BELOW the current price of $126.06, and the payoff ratio is 1.18:1. On the arithmetic alone this is an Avoid.

Here is why it is not one, and the reasoning is about evidence rather than about optimism.

The two facts that drive the base below spot are each unverifiable from the archive. The first is the float schedule — the claim that tradeable supply quadruples on 2026-08-06 and reaches roughly 40% by December. No filing in this archive says that. Its source is a knowledge-base claim from a named commentator, it is directly contradicted by a second knowledge-base claim asserting 366-day locks and "no true price discovery until 2027", and the only lock-up evidence the filings do contain — two officers' 10b5-1 plans deferred to 2027 — points the other way. The second is the forward estimate array, whose FY2026 revenue consensus is arithmetically unreachable, which means the FY2027 and FY2028 figures on which any bull case rests cannot be trusted either. We are not willing to issue an Avoid on a security whose two load-bearing negatives are, respectively, a podcast assertion and a stale spreadsheet.

And the business underneath is not in question. Revenue grew 91.9% in the quarter and beat consensus by 14.5%. Gross margin rose 11.4 points to 55.3%. The operating loss narrowed from $970M to $143M. Starlink subscribers doubled to 12.0 million and the Connectivity segment earned $1,656M of operating income at a 38.6% margin. The company holds $100,009M of cash and securities against $39,364M of debt, and owes nothing material until 2031. These are extraordinary numbers and they are all filing-verified.

What we will not do is pay 51x annualised revenue for them while 83% of the capital goes into a segment that lost $1,257M in the quarter.

Three reservations stated as reservations.

First, the valuation rests on the AI segment and two independent methods agree on how much. Jeremy Grantham's claim says roughly 90%; our own sum-of-observable-parts, done without reference to it, puts it between 78% and 90%. That segment generated $2,561M of revenue, lost $1,257M, and consumed $23,551M of capital in six months. Its advertising sub-line shrank 13.8%.

Second, the Space segment's operating metrics went backwards. Launches −17.8%, mass to orbit −25.6%, internal launches −25.0%, one Starship launch in the quarter. The orbital-compute idea that carries the most valuable part of the bull thesis requires rapid Starship reusability, and cadence declined.

Third, dilution is continuous and about to step up. Weighted average shares doubled year on year; share-based compensation was $1,470M in six months, 11.8% of revenue; and the Cursor merger will issue approximately 476 million shares at today's price for a business the company has valued at $60 billion, with the share count rising as the price falls.

Pre-registered KILL criteria — what would take this to Avoid:

• A float step-up occurring on or about 2026-08-06 and the price breaking $108.37 — the low of the tech series, 14.0% below spot — which would confirm both the supply thesis and its price consequence.
• The AI segment's operating loss WIDENING beyond the $1,257M recorded this quarter while capital expenditure stays above $10 billion a quarter.
• A capital raise despite $60,645M of net cash, which would mean the disclosed burn understates the requirement.
• Connectivity's operating margin below 30%, which would mean the ARPU decline from $85 to $66 is reaching the segment that funds everything else.
• The Cursor merger failing to close, or closing with materially more than 500 million shares of consideration.
• Starlink subscriber growth below 50% year on year, which would break the only metric currently doing the work.

Pre-registered UPGRADE conditions — what would take this to Buy — Tactical:

• A price below roughly $75, at which the base case is +40% and the payoff ratio moves through 2:1 with no change to any number in this dive. That is 40.5% below spot, and it is roughly where the sum-of-observable-parts sits plus a modest premium.
• The AI segment reaching operating breakeven at any revenue level, which would remove the single largest unpriced liability and validate the capital programme.
• A filed disclosure of the actual lock-up schedule — a prospectus supplement, an 8-K, or a Section 16 pattern — which would convert the largest uncertainty in this dive into a knowable fact in either direction.
• A Starship cadence step-up with demonstrated rapid reusability, which is the enabling condition for the orbital-compute case and would justify a materially higher terminal multiple.
• A re-based estimate array whose current-year revenue figure is achievable, which would restore any basis for forward valuation.

Where SPCX sits in the Synthos Framework Portfolio. No position, and no watch-list trigger above $75. Logged as a tracked Synthos call (Watch) as of 2026-08-04 at $126.06, with the fair-value anchors, kill criteria and upgrade conditions all gradeable. The specific thing this dive should be graded on is not the price target. It is the correction in Section "What this dive was commissioned to verify": the IPO and the notes offering verify exactly against the filings; the 2026-08-06 lock-up step-up does not, and its provenance is a commentator rather than a document. If the float does step up on Thursday, the earlier dive will look prescient and this one pedantic — and the discipline will still have been correct, because a research process that promotes assertions to facts without checking them will eventually promote a wrong one.

Single biggest risk: that roughly 78-90% of the enterprise value is a claim about a segment that currently loses money at a rate of $1,257 million a quarter. Connectivity is worth owning at some price. Space is a near-monopoly that presently loses $542 million a quarter and launched 17.8% less than a year ago. The AI segment consumed $23,551 million of capital in six months, grew revenue 247.5% to $2,561 million, and is where essentially the entire $1.6 trillion enterprise value lives. At 51x annualised revenue with free cash flow of minus $25.0 billion for the half, there is no valuation support of any kind beneath that position — and unlike every other name in this batch, there is no dividend, no buyback and no earnings floor to arrest a decline. The $60,645 million of net cash means the company survives any outcome; it says nothing about what the equity is worth.

Provenance & disclosures

• REDACTION DISCLOSURE. House style prohibits naming artificial-intelligence models or laboratories. This dive therefore redacts the name of the frontier model quoted from Note 1 of the 10-Q, the legal names of the artificial-intelligence entities involved in the 2026-02-02 common-control merger, and the laboratory names appearing inside three quoted knowledge-base claims. Every redaction is marked in place. No number, date or financial fact has been altered or omitted; the redactions affect proper nouns only, and the underlying transactions and segment economics are described in full.
• Traceability: 130 name-level knowledge-base claims name SpaceX out of 52,021 distilled claims — the largest lane in this programme. Raw hits 210; surviving a case-sensitive entity re-run 132; discarded as text-only 78; discarded as confirmed quarantine duplicates 2; breadth 132; net conviction UNDEFINED and no score is given, because the store holds a $20-30 trillion valuation target (Ron Baron via compound_and_friends, 2026-07-24, conviction 96) and a sub-$100 price target (Josh Brown via compound_and_friends, 2026-07-21, conviction 60) on the same security four weeks apart, from named independent speakers. The homograph test PASSED with zero case-insensitive-only matches. THE DEFECT FOUND: two claims dated 2026-05-23 under the channel slug quarantine_misattributed, speaker "Dan", skill 2.0, are near-verbatim restatements of two Dan Ives claims filed under anthony_pompliano_show on 2026-05-20 — same 80-85% merger odds, same orbital-data-centre framing, same "80% of value is future". Both quarantined copies are EXCLUDED. This is the SECOND name in this batch on which the quarantine slug leaked a duplicate into an entity sweep, which suggests the quarantine is not filtered out of retrieval. A further correction: the darius_dale claim of 2026-07-31 asserting SpaceX is "down ~50% since IPO" is FALSE on our data — at the prior close of $114.53 the stock was 15.2% below the $135.00 IPO price; the 50% figure is the drawdown from the post-IPO high and the claim conflates the two. Most claims in this lane carry a NAMED speaker, which is unusual for this store; speaker_role reads independent on most and journalist on two, and was not relied on. NO claim in this lane is management voice. All quotes are verbatim from the stored claim text except where redacted as disclosed above.
• Data as-of: fundamentals — income statement, balance sheet, cash flows, segment results and operating metrics through 2026-06-30, all filing-verified from the 10-Q filed 2026-08-04 · annual figures for FY2023, FY2024 and FY2025 from the vendor payload ONLY; this archive contains NO 10-K for SPCX because the company has not yet had a fiscal year end as a public registrant, so no annual figure in this dive is filing-verified · estimates 2026-08-04 · prices 2026-08-04, quote timestamp 1785873600 = 2026-08-04T20:00:00Z ($126.06, +10.07%; no 50-day or 200-day average exists; RSI 42.1; MACD −10.42) · knowledge-base claims 2026-08-04. SpaceX's fiscal year is the calendar year. All figures come from the Synthos vendor data file for SPCX or from the SEC filings in the SPCX archive; no figure comes from memory, recall or external retrieval.
• Filing archive contents — FOUR documents, and the gap matters: 10-Q filed 2026-08-04 (June 2026 quarter — the primary and almost exclusive source for this dive: the IPO terms, the five-for-one stock split, the common-control mergers, the Cursor merger agreement, the EchoStar spectrum purchase agreement, the three-segment results, the operating metrics, the debt disclosure, the officer lock-up arrangements and the complete financial statements); 8-K filed 2026-06-23 (Item 8.01 — the pricing of the five-tranche $25.0 billion notes offering and the stated use of proceeds to repay a bridge loan facility in full); 8-K filed 2026-06-26 (Item 8.01 — the indenture and issuance of the notes, with the par call dates and applicable spreads); 8-K filed 2026-08-04 (Item 2.02 — second-quarter results; cover page and exhibit pointer only, Exhibit 99.1 not in the extract, which does not matter because the 10-Q filed the same day carries the complete statements). There is NO 10-K and NO S-1 or prospectus in this archive, and the absence of the prospectus is the direct cause of the unresolved lock-up question.
• What the complete archive changed versus the earlier truncated one — this is the point of the rewrite. THREE flagged corporate actions were checked. (1) THE IPO — VERIFIED EXACTLY. The 10-Q states the offering completed 2026-06-15 at $135.00 per share for 638,888,888 Class A shares with net proceeds of $85,675 million after $575 million of costs, the registration statement having been declared effective 2026-06-11. The earlier flag was right on both the date and the price, and the payload's profile.ipoDate of 2026-06-12 is WRONG. (2) THE NOTES OFFERING — VERIFIED EXACTLY, with two new facts. $25.0 billion priced 2026-06-23 and issued 2026-06-26 in five tranches ($7.0B/5.350%/2031, $6.0B/5.650%/2033, $6.0B/5.875%/2036, $2.5B/6.600%/2046, $3.5B/6.650%/2056). NEW: the notes were sold under Rule 144A and Regulation S and are unregistered; and the proceeds repaid a BRIDGE LOAN FACILITY in full, a refinancing that cost approximately $2.7 billion between a $1,153 million extinguishment premium and a $1,545 million accounting loss. (3) THE 2026-08-06 LOCK-UP STEP-UP — NOT VERIFIED, AND DOWNGRADED. Nothing in the complete archive states it. The 10-Q refers to the lock-up only as "the lock-up period (as defined in the Prospectus)", and the Prospectus is not in the archive. The date's actual source is a knowledge-base claim from Josh Brown dated 2026-07-21 ("float goes 3%→12% two days after Aug 4 earnings, ~40% by Dec 8"); it is directly contradicted by an empire claim of 2026-06-05 asserting 366-day locks and "no true price discovery until 2027"; and the only lock-up evidence the filings contain is that the Chief Financial Officer and the President and Chief Operating Officer each subjected the vast majority of their shares to an "extended lock-up period" with 10b5-1 plans that do not commence sales until 2027. Two further sources give two more schedules (roughly 20% "after earnings"; roughly 25% by September). Four accounts, four schedules, no filing. AND TWO THINGS NEITHER DIVE COULD HAVE HAD: the company reported second-quarter results and filed its 10-Q on 2026-08-04, the day of this dive, beating consensus revenue by 14.5% with revenue up 91.9% and the stock rising 10.07%; and the 10-Q discloses a signed merger agreement to acquire a private software company at an implied equity value of $60 billion in Class A stock, expected to close in the third quarter, plus a two-step spectrum purchase agreement with EchoStar Corporation dated 2025-09-07 and amended 2025-11-05. Neither transaction appears anywhere in the vendor payload. The verdict is unchanged at Watch; the base fair value moves from $88 to $105, because the June quarter's operating economics — a 38.6% Connectivity margin, a $143 million consolidated operating loss against $970 million, and a 55.3% gross margin — are materially better than anything the earlier dive could see.
• Where the filings contradicted or corrected the vendor (detailed in Section 7): the share count — the 10-Q cover page states 7,696,293,669 Class A plus 5,485,486,276 Class B = 13,181,779,945 at 2026-07-28, against the 13,075,865,834 implied by the vendor's market capitalisation, a 0.8% and roughly $13.3 billion understatement; profile.ipoDate of 2026-06-12 against the filed completion date of 2026-06-15; inc_q, which holds only three non-consecutive quarters, reports Q2 2026 interest expense as negative $629 million and Q2 2026 depreciation as zero when the filed six-month figure is $5,290 million; inc_a fiscal 2024, whose $0.0018 diluted EPS on 9,956 million shares against $791 million of net income is wrong by a factor of 44 and whose share count is inconsistent with both adjacent years; the entire annual series, retroactively restated for common-control mergers the payload does not disclose; the five-for-one May 2026 stock split, absent from the payload; est.ebitdaAvg/ebitAvg, carrying a fixed-ratio signature of exactly 66.88% and 25.48% of revenue in every year, with the FY2025 EBITDA estimate of $12,908.6 million against a reported actual of $4,289 million — a 201% overstatement; the FY2026 revenue consensus of $38,930 million, arithmetically unreachable against a first half of $12,508 million; and the two analyst-coverage blocks, which disagree by a factor of five (4 analysts in grades, 14-20 in est). FIVE PAYLOAD BLOCKS ARE ENTIRELY EMPTY — km_ttm, ratios_ttm, seg_prod, seg_geo and peers — and this is reported as truncation rather than papered over: every ratio, multiple and enterprise-value figure in this dive is computed by hand from the 10-Q, no peer comparison is drawn anywhere, and the segment analysis comes entirely from Note 15. tech.sma50, tech.sma200 and all three relative-return fields are null and are reported as absent rather than substituted, which is correct behaviour for a security with fifty days of history. Where vendor and filing AGREED: the June-quarter income statement ties to the 10-Q on every line (revenue $7,814M, operating loss $143M, net loss $541M, EPS −$0.09, 5,864M weighted shares); the Q1 2026 diluted EPS of −$1.27 in inc_q is CORRECT and reconciles to the filing's six-month figures, while earn_cal's −$1.19 for the same quarter does not.
• Basis note: SpaceX reports on a GAAP basis and publishes no adjusted earnings measure in this archive. All loss-per-share figures are GAAP basic and diluted, which are identical because the company is loss-making. The six-month loss attributable to shareholders of $5,488 million exceeds the net loss of $4,817 million by $671 million, the effect of the preferred-stock conversion at the IPO, which is why six-month loss per share is $1.12 rather than $0.99. Free cash flow is recomputed throughout as filed operating cash flow minus filed purchases of property, plant and equipment, and the resulting minus $25,010 million for the half UNDERSTATES economic investment, because a further $3,921 million of equipment was financed by other financings and $5,513 million sits in accrued expenses and accounts payable.
• Estimate coverage: 17 analysts on FY2026 revenue and 14 on EPS; 17 and 15 on FY2027; 20 and 18 on FY2028. But the grades block reports only FOUR analysts, and the price-target range of $115 to $401 — a 3.5x spread — is consistent with the smaller figure. The FY2028 EPS range of MINUS $0.45 to $6.47 spans a loss and a profit in the same year and is the most honest number in the array. The FY2026 revenue consensus is arithmetically unreachable and the entire array is treated as scenario markers rather than forecasts.
• Peer note: the peers block is EMPTY. No comparison universe is supplied and none is constructed. No peer multiple appears anywhere in this dive. The filings name competitors only in a knowledge-base claim (real_vision, 2026-06-28, referring to competition in launch and broadband), not in the archive.
• Fair-value caveat: the $55 / $105 / $210 anchors are NOT built on the earnings estimate array, because the FY2028 consensus spans a loss of $0.45 and a profit of $6.47 and the FY2026 revenue consensus is unreachable. The bear is a sum of observable parts from the FILED June-quarter segment economics — Connectivity at 15x annualised revenue, Space at 20x, AI at 20x, plus $60,645 million of net cash, less roughly $60 billion of Cursor consideration — which produces approximately $41 per share before rounding up for the deferred-revenue backlog and Starship optionality. The base implies an enterprise value of roughly $1,323 billion, 42.3x annualised second-quarter revenue, i.e. MULTIPLE COMPRESSION from today's 51.2x on a higher revenue base. The bull is 32.5x the FY2028 consensus HIGH EPS and is roughly the post-IPO high. Stated arithmetic, not a discounted cash flow. The range is deliberately enormous — a 3.8x spread from bear to bull — because that is the honest width of the distribution for a fifty-day-old security with five empty payload blocks, an unreachable current-year consensus and an unverified float schedule.
• Timing: second-quarter 2026 results were released THIS MORNING, 2026-08-04, the day of this dive, with revenue of $7,814M against a $6,826.7M consensus (+14.5%) and a loss of $0.09 per share against an expected $0.225. The 10-Q was filed the same day and carries the complete financial statements. The next dated event is a possible float step-up on 2026-08-06, TWO DAYS AWAY, which this dive could not verify from any filing. The Cursor merger is expected by the company to close during the third quarter of 2026. All insider records are dated 2026-06-15 (five automatic C-Conversions at the IPO), 2026-02-02 (two share-exchange awards at the merger) and 2026-06-16 (one malformed null record). There is no open-market purchase or sale by any officer or director in the file.
• Accessibility note: no information in this dive is conveyed by colour. All emphasis is carried by bold text, table structure and explicit labelling.
• Not investment advice. Independent research, educational and informational only, never personalised. No recommendation to buy, sell or hold any security is made to any person.
• Version: 2026-08-04-full-rewrite.