Question: Pull the full SpaceX Form S-1 and go over it section by section, taking notes and flagging what's interesting.

Method: Downloaded the complete filing (11.8 MB HTML, ~29.7k lines of clean text), split it into 13 sections, and ran one sub-agent per section to extract facts + flags. This note is the synthesis. Raw notes preserved in the section-by-section digest below.

Source: SEC EDGAR — Space Exploration Technologies Corp., Form S-1, filed 2026-05-20, Registration No. 333-, accession 000162828026036936. https://www.sec.gov/Archives/edgar/data/1181412/000162828026036936/spaceexplorationtechnologi.htm

Which document this is. This is the original May-20 S-1 — the full registration statement, with the price range and most share counts left blank. The lane separately has events for the 2026-06-01 S-1/A (amendment, staggered lock-up) and the 2026-06-03 S-1/A pricing ($135/share, ~$75B offering). Every "blank" below is because we read the original; the priced terms are downstream. Read this as the complete narrative + financials + governance source, and the priced S-1/A as the deal-terms overlay.

---

The one-line reframe

The SpaceX IPO is no longer a pure space/Starlink IPO — it is a combined SpaceX + Starlink + xAI/Grok/X company. SpaceX acquired xAI on 2026-02-02 (and xAI had acquired X/Twitter on 2025-03-28). Both deals were done as reorganizations of entities under common control (Musk controlled all of them), so there is no purchase accounting, no goodwill step-up — and the financial "history" presented is a retrospectively-combined entity that never actually existed in 2023–2024. Ticker SPCX, dual-listed Nasdaq + Nasdaq Texas. This is a direct read-through to the workspace's existing SpaceX-vs-Anthropic track separation: the xAI merge is a real corporate merge into SpaceX, while Anthropic stays a separate counterparty (and is now disclosed as a compute customer, below).

Bottom line — the dozen things that matter

1. Connectivity (Starlink) is the only thing making money, and it's carrying the whole company. Q1'26: Connectivity +$1,188M op income vs Space −$662M and AI −$2,469M → consolidated −$1,943M op loss on $4,694M revenue. FY25 Connectivity op income $4,423M (≈39% margin) funds everything else.
2. AI is the burn. AI capex was $7,723M in Q1'26 alone (76% of the company's $10,107M quarterly capex), on AI revenue of just $818M. AI segment was EBITDA-positive in 2023–24 (that was legacy X advertising) and went deeply negative once the xAI compute build hit the P&L.
3. Starlink ARPU is compressing hard and on purpose: $99/mo (FY23) → $91 (FY24) → $81 (FY25) → $66 (Q1'26). Subs are exploding (2.3M → 10.3M, +105% YoY) so revenue still grows, but it's a pure volume story; management guides ARPU lower.
4. IPO = mandatory debt paydown, not growth capital. There is a $20B SpaceX Bridge Loan (Goldman-led, drawn to refinance all the legacy X/xAI high-cost debt) maturing Sept 2, 2027, and the loan agreement requires net IPO proceeds be applied to repay it within 6 months. Total debt $29.1B vs cash $15.9B at 3/31/26.
5. Musk's lock-up is 366 days with ZERO early release — while everyone else is on 180 days with a staggered, partly price-triggered early-release ladder. This is the single cleanest fact for the liquidity-event thesis: the dominant overhang (Musk's 93.6% of Class B / ~85% pre-IPO voting power) is hard-locked for a full year.
6. Control is permanent and uncheckable. Class B = 10 votes; Class B elects 51% of the board as a separate class; Musk removable only by Class B vote (which he controls); no dual-class sunset; new Class B can only ever be issued to Musk/family; corporate-opportunity doctrine fully waived; controlled-company exemptions taken (no independent comp/nominating committee).
7. The Musk mega-grant is ~1.3 billion Class B shares. A Jan-2026 award of 1,000,000,000 performance Class B shares (15 market-cap tranches $500B→$7.5T AND a permanent ≥1M-person Mars colony) plus a 302.1M replacement xAI award (market-cap tranches AND "non-Earth data centers delivering 100 terawatts/yr"). Grant-date FV $90–96/share → on the order of **$90B+** of potential CEO comp, in super-voting stock. No expense booked yet (conditions deemed improbable).
8. Related-party financing is enormous and routed through a board member. Valor Equity Partners (founder/CEO Antonio Gracias sits on SpaceX's board) provides the AI-infrastructure financing via "failed sale-leaseback" deals: related-party debt $4.5B (12/31/25) → $9.0B (3/31/26), plus a further $6.6B Valor III lease signed April 2026; $186M interest to Valor in Q1'26 alone; finance-lease discount rate disclosed at 22.6%. PwC's audit opinion carries a standalone emphasis paragraph on related-party transactions.
9. Three big contingent overhangs are barely sized in the original S-1: (a) EchoStar spectrum purchase $19.6B (~$11.1B in ~262M Class A shares at a fixed $42.40 pre-split, + up to $8.5B cash), FCC-approved 5/12/26, closes ~Nov 2027; (b) Cursor call option at $60B implied equity value, exercisable ~30 days post-IPO, with a $10B break ($1.5B fee + $8.5B deferred services) if SpaceX walks; (c) Anthropic cloud-compute contract through May-2029 (MD&A elsewhere cites $1.25B/month), either side can terminate on 90 days' notice, and Anthropic is a direct Grok competitor being sold SpaceX's crown-jewel compute.
10. Everything depends on Starship, which hasn't delivered a payload yet. Falcon 9/Heavy cannot launch V3 Starlink or V2 Mobile sats — only Starship can. V3 broadband, direct-to-cell, the 2028 "orbital AI compute" satellites, and the cost-curve all route through Starship reaching commercial payload service (guided H2 2026; 11 test flights done, none operational). Single point of failure for multiple "moats."
11. Governance/litigation tail is unusually thick: controlled-company carve-outs, mandatory ICC arbitration + Texas Business Court forum + jury waiver + class-action ban + 3% derivative-suit threshold; no documented/tested internal controls pre-IPO (can't rule out a material weakness); live Grok CSAM/non-consensual-image class actions, a Clean Air Act injunction motion against the COLOSSUS II gas turbines (NAACP, May 2026), AI-training copyright suits, an EU DSA €120M fine under appeal, and a −17% effective tax rate (paying $718M tax while losing $4.2B pre-tax, from a $2.2B valuation-allowance build).
12. TAM is framed at $28.5T but explicitly excludes China and Russia — and the AI-segment "supported accounts" headline (1.3B) is soft: only 550M MAUs, of which only ~117M actually use Grok features.

---

Anchor numbers (issuer-disclosed; $ in millions)

Consolidated P&L

• Accumulated deficit $41,311M (3/31/26). FY25 EPS (basic & diluted) $(1.69). Q1'26 net loss embeds a $1,526M loss on debt extinguishment (incl. $1,163M prepayment penalty) from the bridge refi.

Segment (Q1'26: revenue / op income(loss) / Segment Adj EBITDA / capex)

• FY25 segment revenue: Space 4,086 / Connectivity 11,387 / AI 3,201. FY25 Connectivity op income 4,423 (the engine).
• Space went EBITDA-negative at the segment level in Q1'26 (Starship R&D $930M in the quarter) after being positive since 2018.
• AI Q1'26 revenue mix: Advertising $343M (down from $443M Q1'25), AI Solutions & Infra $475M (up from $284M). Grok/X paid subs ≈ 6.3M (4.4M X Premium + 1.9M SuperGrok).

Capex ramp (total): FY23 $4,415M → FY24 $11,163M → FY25 $20,737M → Q1'26 $10,107M. AI capex line: $463M → $5,633M → $12,727M → $7,723M.

Balance sheet / liquidity (3/31/26 unless noted)

• Cash & equivalents $15,852M (down from $24,747M at 12/31/25); + marketable securities $7,823M.
• Total assets $102,094M; PP&E net $53,879M (servers/data-center buildout).
• Total debt principal $29,132M; total debt + finance leases ~$30B. No principal due until the $20B bridge in Sept 2027.
• Deferred revenue $13,236M; backlog $27,621M (~36% recognized <1yr).
• Goodwill $11,681M — entirely legacy Twitter goodwill, none from the xAI/X mergers (common-control).
• Digital assets: 18,712 BTC, fair value ~$1,637M (cost $661M).
• Customer concentration: Customer A = 20.9% of FY25 revenue (24.2% FY24, 25.2% FY23) — almost certainly the U.S. government; ~one-fifth of revenue is federal.

---

Flag clusters (the interesting stuff)

A. The combined-entity / accounting reframe

• Common-control accounting on both mergers → no step-up, no new goodwill, history recast as if always combined. The 2023–24 P&L you're shown is not the history of the company that will trade. AI segment losses of $(3,973)M (2023) and $(1,561)M (2024) are baked into "history" even though SpaceX didn't own xAI then.
• xAI merger consideration (disclosed only in the notes): 321.7M Class A + 121.7M Class B shares (pre-split) + $2,947M cash; exchange ratio 0.1433, cash election $75.46/xAI share. This is dwarfed by xAI's ~$113B Series E implied valuation — the related-party deal entered SpaceX's books at historical cost, so the balance sheet massively understates the AI assets' market value.

B. Capital structure & Musk control (→ governance read for spacex-ipo-liquidity-event)

• Dual-class 10:1; Class B separately elects 51% of all directors; no sunset; Class B auto-converts to A only on transfer; new Class B issuable only to Musk/family. Pre-IPO Musk = 12.3% of Class A + 93.6% of Class B = ~85.1% combined voting power.
• Mega-grant ≈1.3B Class B shares with market-cap-to-$7.5T + Mars colony / 100-TW non-Earth datacenter conditions — functions as indefinite retention + a control-increasing overhang (it's super-voting stock).
• Controlled-company exemptions taken; comp committee not fully independent (Gracias sits on it while also on Neuralink/Boring boards); corporate-opportunity doctrine renounced (Musk/Tesla/Boring/Neuralink can compete with SpaceX with zero duty).
• Mandatory arbitration + Texas Business Court forum + jury waiver + class-action ban + 3% derivative threshold = securities-class-action enforcement effectively removed.

C. Lock-up & future-sale mechanics (→ direct payload for the liquidity-event lane)

• Company: 180-day lock-up. Founder (Musk) + significant investors: 366 days, and Musk has NO early-release.
• Staggered early-release ladder for the 180-day / partial-366 cohort ("Early Release Eligible Shares"): (a) up to 20% after the first post-IPO (Q2'26) earnings; (b) +10% if price ≥ 130% of IPO for ≥5 of 10 days into that earnings date (strong tape → more selling); (c) +7% each at 70/90/105/120/135 days (35% total); (d) +28% after Q3'26 earnings; (e) remainder at 180 days.
• Directed-share-program shares are NOT locked up (friends/family/business partners can sell day 1; size redacted). Registration-rights holders (Musk affiliates, Google, Valor, DFJ) can demand registration 6 months post-IPO for offerings >$250M.

D. The related-party web

• Valor (Gracias): the AI-infra lender of record — ~$9.0B related-party debt at 3/31/26 (+$6.6B Valor III lease April-26), 22.6% finance-lease rate, $186M Q1 interest. PwC emphasis paragraph.
• Tesla: Megapacks $506M (2025) + Cybertrucks $131M; bought $0.5B+ of goods/services from xAI's counterparties; $2B Series E investment (→ 3.8M SpaceX Class A pre-split); Terafab (chip fab, +Intel Apr-26) and Macrohard collaborations (terms "not yet finalized"). Three of eight SpaceX directors have current/prior Tesla board ties.
• Smaller Musk-affiliate flows: Boring Co. leases, Musk Industries LLC leases, Musk-owned security co, Craft Aviation. Musk pledged 237,530 Class A; Nosek pledged 2.38M Class A for personal debt.

E. Cash, debt & the bridge

• $20B bridge (Sept-2027) refinanced the legacy stack (X B-1 at ~12.4%, X B-3 ~9.8%, xAI 12.5% notes, xAI term loans ~12.5%) into SOFR+0.75–1.75% — only possible on the SpaceX parent balance sheet + IPO context. Covenant: leverage ≤3.75x. IPO proceeds must repay it within 6 months.
• Free cash flow deeply negative: FY25 op CF $6,785M vs capex $20,737M. Q1'26 investing outflow $(16,724)M. $9.1B of "Other financings" are failed sale-leasebacks on AI hardware (couldn't be derecognized).
• Adjusted EBITDA is flattering: FY25 D&A ($6,701M) alone exceeds Adj EBITDA ($6,584M); add-backs include $1,947M SBC, $487M restructuring, $1,945M interest. Satellite useful-life is a big earnings lever (±1yr ≈ $480M FY25 op income).

F. Segment economics & the "moat" claims

• Connectivity ~63% Adj-EBITDA margin is partly an artifact: Starship/launch R&D is carried in the Space segment, and SpaceX recognizes no revenue when it launches its own satellites — so Space eats the cost and Connectivity/AI look cleaner.
• Starlink: 10.3M subs, 164 countries, ~9,600 sats (~75% of all active maneuverable sats), 23,000+ inter-sat lasers; enterprise retention flex ("no >$750K customer has voluntarily churned since 2023" — carefully scoped). Direct-to-cell (Starlink Mobile): ~650 V1 Mobile sats, ~7.4M monthly devices, ~30 countries; named MNOs incl. T-Mobile; competitors ASTS / Lynk / Globalstar / Skylo.
• AI/compute: COLOSSUS (~100k H100, 122 days) + COLOSSUS II (GB200/GB300 clusters, 91/64 days) ≈ 1.0 GW; next phase +220k GB300 / +400 MW; training Grok-5. NVDA read-through on GB200/GB300 scale; INTC via Terafab. Build-speed benchmark ("91 days vs 2 years") is real but apples-to-oranges (repurposed shells, not greenfield).

G. The speculative roadmap

• Orbital AI compute ("as early as 2028"): needs ~1M metric tons/yr to orbit for 100 GW — ~450× SpaceX's cumulative lifetime mass-to-orbit (~7,400 t). The filing even flags a disconfirmation condition: cheap terrestrial nuclear could kill the orbital-compute economics.
• Lunar mining, Mars city, in-orbit refueling ("not yet demonstrated") all sit in the long-dated optionality bucket.

H. Regulatory / legal tail

• No insurance on launch vehicles, payloads, or in-orbit satellites — uninsured by policy.
• Live: Grok CSAM/non-consensual-image class actions; NAACP Clean Air Act PI motion vs COLOSSUS II gas turbines (could curtail AI build); AI-training copyright suits; EU DSA €120M fine (appeal); Irish DPC GDPR inquiry into Grok/children; FTC chatbot-safety inquiry; Vidstream patent ($105M + $67M interest). Litigation accrual $399M (3/31/26).
• FAA / FCC / ITAR / CFIUS: Starship needs FAA waivers (return-to-launch-site reentry currently not permitted); international Starlink-Mobile spectrum granted in only "a limited number" of countries; CFIUS divestiture power is a structural constraint on foreign Class A holders.

---

Section-by-section digest (condensed raw notes)

1. Front matter + Prospectus Summary + The Offering (cover→summary). Ticker SPCX, Nasdaq + Nasdaq Texas; 23-underwriter syndicate (GS/MS/BofA/Citi/JPM leads); 5-for-1 split 5/4/26; price + share counts blank. Mission/segment narrative; Q1'26 headline financials; segment + capex tables; pending EchoStar/Cursor/Anthropic/Terafab. Flags: ARPU compression $99→$66; AI capex dominance; cash $24.7B→$15.9B; preferred $38.8B→$7.0B converting; dual-class board-control; Anthropic ~$45B-class contract w/ 90-day out; common-control history "never existed."

2. Risk Factors A. $29.1B debt; ~1/5 revenue federal; Starship-only for V3/V2 Mobile; GPUs bought purchase-order only; AI data centers on natural gas; orbital AI "no one has done it"; Brazil asset-freeze contagion precedent; no launch/sat insurance; Irish DPC + FTC inquiries; "Unhinged"/"Spicy" Grok modes named; foreign anti-satellite-weapon risk; orbital-AI nuclear-energy kill switch disclosed.

3. Risk Factors B. $41.3B accumulated deficit; no documented/tested internal controls pre-IPO; permanent no-sunset Musk control; controlled-company exemptions used; corporate-opportunity waiver; Texas Business Court + ICC arbitration + jury waiver + class-action ban + 3% derivative threshold; xAI integration "not complete"; copyright-training litigation active; "Money Product" (X payments) regulatory exposure; Anthropic compute named.

4. Use of Proceeds / Dividend / Capitalization / Dilution. Proceeds → AI compute / launch / constellation / GP (no debt-repay line, despite the bridge); no dividends (covenant-restricted). Cap table: Class A 2.88B→6.82B pro-forma, Class B 2.42B→5.70B, Class C eliminated; total debt $29.1B; APIC $74.1B; deficit $(41.3)B; equity $34.5B→$41.6B pro-forma. $20B bridge is the dominant debt item; $9.1B failed-sale-leaseback "Other financings."

5. MD&A A (overview / segments / results). Full segment unit economics (tables above); launch mix shifting to internal (33 of 40 Q1'26 Falcon launches were internal); Connectivity cost-of-rev growing faster than revenue (satellite depreciation); AI EBITDA reversal = compute build, not business decay; ad revenue −$100M (platform "overhaul"); EchoStar cited at $19.6B.

6. MD&A B (non-GAAP / liquidity / cash flow / critical estimates). Adj-EBITDA reconciliation (D&A > Adj EBITDA); bridge-loan terms + 6-month mandatory IPO repay; covenant ≤3.75x; cash-flow detail; satellite-life sensitivity $480M; −17% effective tax from $2.2B VA build; digital-asset MTM losses in "Other expense."

7. Business A (the three segments + TAM). Launch cost-curve ($18.5k/kg → ~$1,400/kg Falcon Heavy, Starship target −99%); >$15B invested in Starship; Starship payload H2'26; Starlink metrics; COLOSSUS/COLOSSUS II detail; Grok-1→4.3, training Grok-5; Cursor $60B option, Anthropic $1.25B/mo, Terafab (Tesla+Intel); orbital AI 2028; TAM $28.5T ex-China/Russia. Flags: "supported accounts" soft (1.3B → 550M MAU → 117M Grok); Connectivity margin artifact; benchmark cherry-picks.

8. Business B (manufacturing / IP / competition / government & regulation). Starbase/Hawthorne/McGregor/Redmond/Memphis-Southaven facilities; >22,000 employees, zero unionization; competitors named per segment (ULA/Blue Origin/RKLB/Firefly/Relativity; AST/Lynk/Globalstar; OpenAI/Anthropic/Google/Meta/MSFT; Coreweave/Nebius as compute comps); FAA/FCC/ITAR/CFIUS/NDAA detail; "Department of War" terminology; gov revenue % not disclosed here; Starshield secure constellation; COLOSSUS II behind-the-meter gas plant; Grok "minimally censored" = DSA/OSA exposure.

9. Management + Executive Compensation. 8-person board; 5 Class B directors / 3 common; controlled-company carve-outs; Musk 2025 cash comp $54,080; Shotwell $85.8M (one $76.8M retention grant); Musk 1B-share Mars/market-cap mega-grant + 302.1M xAI-replacement (100-TW non-Earth datacenter); CFO option metric flipped from $2B FCF to $10B-EBITDA tranches; Harrison (Google) on board while Google is a customer; Pampena v. Musk 10b-5 partial judgment (Apr-26, post-trial motions pending June-26).

10. Related-Party + Capital Stock + Shares Eligible + Tax + Underwriting. (Highest-value bundle.) Valor $20.2B aggregate GPU/compute leases guaranteed by SpaceX; Tesla flows; Musk 366-day lock-up, zero early release; the staggered early-release ladder (incl. 130%-price trigger); DSP shares unlocked; dual-class voting mechanics + ~85% pre-IPO Musk voting power; greenshoe 30-day; Morgan Stanley = xAI advisor + lead underwriter + stabilization agent + DSP administrator; all top bookrunners are also bridge/credit-facility lenders (dense conflict web); PwC related-party emphasis; retail selling group (Schwab/Fidelity/Robinhood/SoFi/E*TRADE).

11. Audited Financial Statements (primary). PwC clean opinion (auditor since 2012) with two emphasis paragraphs: digital-asset accounting change + related-party transactions; balance sheet/P&L/cash-flow tables; FY25 R&D +149% YoY is the swing item; redeemable preferred $38.8B stack; Customer A 20.9%; 2023 $3,775M Twitter/X brand impairment; satellites depreciated 3–5 yrs; BTC on balance sheet.

12. Notes A (policies / mergers / goodwill / debt / leases / tax / commitments). Common-control merger mechanics + xAI consideration; EchoStar spectrum $19.6B (~262M shares @ $42.40 fixed + $8.5B cash + ~$3B loan to the Trust); legacy debt rates (X B-1 12.4%, etc.); leases (finance-lease rate 22.6%); VA $8,286M, federal NOL $9,728M, R&D credits $3,586M; non-cancelable commitments $25.5B (2027 alone $21.5B = EchoStar); litigation incl. EU DSA €120M, Vidstream, GDPR class actions, Grok CSAM suits.

13. Notes B (Q1'26 interim / SBC / related-party / segments / subsequent events). Q1'26 SBC $639M; Musk awards detail (1,000M + 302.1M Class B, no expense yet); Valor related-party debt $9.0B at 3/31/26, $186M Q1 interest; segment tables; backlog $27.6B; NAACP Clean Air Act PI motion vs COLOSSUS II turbines; subsequent events — Cursor $60B option / $10B break, Turbine $2B asset buy, Anthropic compute (through May-2029, 90-day out), $20B bridge, 5-for-1 split.

---

S-1/A update (priced, 2026-06-03 — acc 000162828026040364) — pulled & confirmed

The priced amendment fills the original S-1's blanks:

• 555,555,555 Class A shares @ $135.00 = ~$75.0B, ALL PRIMARY (company takes the cash; no selling-stockholder secondary on the cover). Greenshoe = standard 30-day over-allotment.
• Musk = 82.4% of post-IPO voting power (82.3% if greenshoe exercised), ~81.1% of it from Class B — he elects a majority of the board through Class B alone. Confirmed "controlled company."
• Lock-up ladder + 366-day founder hard-lock unchanged from the body below. (Exact lock-up share counts / DSP % still partly redacted.)
• Read-through: a ~$75B primary raise = a huge new float to absorb, on top of the staggered post-lock-up supply — reinforces the "not a clean day-one buy" call for the liquidity-event lane. See the companion picks-and-shovels supply-chain map for the supplier side.

Open questions / still to verify

• Primary vs any later secondary, exact lock-up share counts, DSP %, and how $42.40-struck EchoStar shares compare to the $135 print (the cleanest SATS read).
• Anthropic contract value is stated as $1.25B/month in MD&A but left unquantified in the subsequent-events note — confirm the ACV and whether it's in backlog. (Read-through: this is the compute-landlord relationship, not a SpaceX→Anthropic dependency — consistent with keeping the Anthropic track separate.)
• Cursor option economics if exercised post-IPO — a potentially large near-term dilutive acquisition + $10B downside if abandoned.
• EchoStar $42.40 fixed share price vs the eventual IPO price decides whether the spectrum deal is dilutive or a bargain for SpaceX (and is the cleanest SATS read-through).
• Government revenue % by agency (only "Customer A 20.9%" + "~one-fifth federal" disclosed) — matters for DoD/NASA budget sensitivity.

Routing

• Feeds the spacex-ipo-liquidity-event perspective (lock-up ladder + Musk 366-day hard lock + control mechanics + the staggered-release supply schedule) and ai-power-bottleneck (1 GW compute, gas turbines, GPU/capex scale, Terafab escape-from-Nvidia). Sits under the ai-mega-ipo-fast-entry-liquidity-regime thread alongside the xAI-S1 segment crosswalk.
• Producer-side only: no trade, no paper trade, no publish/promote. Price-truth untouched (all figures here are issuer-disclosed S-1 financials, not market prices).

Sources

• Primary: SEC EDGAR SpaceX Form S-1 filed 2026-05-20 (accession 000162828026036936).
• Cross-refs: 2026-06-02-xai-s1-segment-crosswalk-read; events 2026-06-01-spacex-s-1a-amendment-offering-blank-staggered-lockup, 2026-06-03-spacex-s-1a-priced-135-per-share-75b-offering; thread 2026-06-02-ai-mega-ipo-fast-entry-liquidity-regime.