SpaceX is a launch monopoly, a 63%-margin telecom, and an unproven hyperscaler — stitched together for one IPO.

• Space — the moat. Falcon flew 165 missions in FY25 against ULA's 10 and Rocket Lab's 18, capturing more than 80% of global mass to orbit. Only 43 launches were paying customers — the cadence is consumed internally to deploy Starlink at cost. $4.1B revenue, $653M segment-adjusted EBITDA.
• Connectivity — the cash engine. Starlink ran $11.4B revenue at a 63% segment-EBITDA margin in FY25, already past Iridium's mature 42.5% in year six. Subscribers more than doubled to 10.3M year-over-year (Q1-25 to Q1-26); segment EBITDA grew 86% to $7.2B. Hides inside the consolidated −13.9% operating margin.
• AI (xAI) — the capital vortex. $3.2B FY25 revenue against $12.7B of capex (397% of segment revenue, ~940% in Q1-26 annualized). Segment-adjusted EBITDA loss of $1.2B. Consumed 60% of FY25 capex for 17% of revenue. Bolted on via stock-only common-control merger in February 2026.

At $1.75T the buyer pays full freight for an option whose cash flow looks like a vortex.

Every other lever in the long-term thesis — reinvestment runway, dilution, goodwill, governance leakage — routes through this number. The AI segment is funded by $18.8B of fresh equity and $16.1B of new debt in FY25, not by the business. Anthropic, contracted at $1.25B per month through May 2029, is both the dominant AI-revenue customer and a direct competitor in foundation models. If trailing-twelve-month AI capex/revenue is still above 300% at end-FY27, the optionality framing weakens and the $11.8B X-deal goodwill becomes an impairment candidate.

Musk holds 85.1% of votes on roughly 51% of equity; carve-outs and related-party flows do the rest.

• Voting and structure. 10:1 super-voting Class B gives the founder 85.1% of votes on roughly 51% of equity (per the proxy beneficial ownership table). Class B holders, voting separately, are entitled to elect 51% of directors. Texas re-domicile, mandatory arbitration, and a Nasdaq controlled-company exemption that strips the requirement for independent compensation and nominating committees. Press coverage of the structure has characterized it as 'only Elon Musk can fire Elon Musk.'
• $20.2B of related-party leases. Three xAI equipment leases with director Antonio Gracias's Valor Equity Partners total $20.2B of aggregate cash payment obligations ($6,986M + $6,633M + $6,587M per the proxy) — guaranteed by SpaceX. Gracias remains on the compensation and nominating committee. The xAI merger itself closed in February 2026 with no independent committee process at SpaceX.
• The pay metric switched weeks before the S-1. On January 4, 2026 the board re-cut the CFO's 4M performance options from a free-cash-flow trigger (about to be missed) to a $10B Adjusted EBITDA trigger that excludes SBC, interest, restructuring, and impairment. The metric the CFO defines is now the metric the CFO is paid on.

Reported operating cash is roughly half the headline once the working-capital lifeline reverses.

• The AP lifeline. Accounts payable jumped 167% in FY25 (+$7.4B; $4.4B → $11.8B) and total working-capital + other movements contributed about $2.3B to the reported $6.8B operating cash flow. AP started reversing in Q1-26 (−$1.8B). A sustainable operating-cash proxy lands closer to $3.9B against FY25 capex of $20.7B.
• The EBITDA bridge. About $11.5B of add-backs separate the $6.6B Adjusted EBITDA from the −$4.9B GAAP loss in FY25 — including $1.95B SBC (up 148% YoY), $1.95B of interest on $22.9B of debt, and restructuring charges that have now recurred in every year shown.
• Funded by issuance, not by the business. FY25 raised $18.8B of equity and $16.1B of debt to bridge a $14.0B free-cash-flow deficit and rebuild cash to $24.7B at year-end. The IPO is partly that next raise. FY24 'net income' of $791M flips to a $743M pre-tax loss without a ~$955M unrealized Bitcoin / digital-asset gain.

The window settles cost basis, capital cost, and the milestone the S-1 underwrites.

• Tonight — Starship V3 Flight 12. Inaugural V3 attempt scrubbed May 21 on a stuck pad hydraulic pin; second attempt at 6:30 pm EDT from Starbase. The S-1 explicitly anchors the 2026 case on 'commencement of payload delivery to orbit.' V2 lost 3 of 5 flights and management concedes V3 is 'almost total redesign' of primary structure, engines, electronics, and launch tower.
• June 11 — IPO pricing (Nasdaq: SPCX). Target $1.75T (~93× sales) against three external SOTP cuts at $700B–$1.35T. Polymarket pricing of a SpaceX IPO by June 30 has moved sharply higher (74% in late April, ~90%+ implied as the June 11 pricing date approaches). Lock-up expires around December 8, 2026 — the first chance for a borrow market to form against the equity.
• End of Q2 — Kuiper consumer service in 5 countries. Amazon's terminal price (sub-$400 target against Starlink's $349–$499) is the first real test of the residential ARPU floor since blended Starlink ARPU compressed from $86 to $66 in 12 months (Q1-25 to Q1-26). Management calls the compression mix, not pricing; Kuiper will tell.