The Musk dependency graph is a structural map of how SpaceX (the IPO entity), Tesla, xAI, X, Starlink, and Starshield interlock through shared infrastructure, shared capital, shared technology, and shared customers. It matters for the IPO because traditional equity valuation models treat SpaceX as a single rocket company at roughly $1.75 trillion. We argue that is the wrong frame. SpaceX is the load-bearing infrastructure layer for four otherwise unrelated global systems — armed conflict response, energy transition, the AI arms race, and orbital access — and each system has independently decided it cannot route around SpaceX. The IPO is therefore a chance to buy infrastructure already awarded de facto monopoly status by counterparties who control trillions in non-SpaceX budgets: the DoD, the NRO, NATO members, NASA, and large enterprises. The S-1 confirms the financial fusion that makes the frame coherent — FY2025 revenue of $18.7B, Starlink at $11.4B and 61% of the total, and a consolidated −$4.9B GAAP net loss carrying the xAI build-out. The graph quantifies the unremovability claim that valuation alone hides.

Chain 1 (Conflict → Starlink): every active armed conflict since 2022 — Ukraine, the Levant, the Taiwan Strait standoff — has required civilian-grade resilient connectivity that no terrestrial network or other satellite constellation provides at scale. Chain 2 (Energy crisis → Tesla stack): residential and industrial customers exposed to grid instability or fuel volatility have a single vertically integrated path through Solar Roof + Powerwall + Tesla vehicles + Megapack + Optimus. Chain 3 (AI arms race → xAI compute corridor): xAI's compute, training, and distribution leg depends on SpaceX-class connectivity (Starlink), X-class distribution (the platform), and Tesla-class deployment surfaces (vehicles, Optimus). Chain 4 (Orbital access → SpaceX launch monopoly): more than 90% of mass to LEO and 100% of crewed US orbital launches go through SpaceX vehicles. The chains were identified by working backward from observed counterparty behavior — which budgets are committed, which alternatives were considered and rejected, and which procurement decisions have been made under stated contingency planning.

The four red edges marked headline=true are the load-bearing ones: Ukraine→Starlink (the war's continued operability), Solar Roof→Powerwall→Tesla→Optimus (the residential energy stack), xAI→Starlink-uplink (training data and inference distribution), and DoD→Starshield (classified national-security communications). Removing any single one of these edges does not break the system — each is independently fault-tolerant in isolation — but the graph as a whole has no alternative routing that does not pass through a SpaceX-controlled node. That distinction is the structural source of the unremovability claim. What makes these four edges heavier than the other forty is evidentiary: each marks a place where a counterparty with real budget actively attempted to substitute and failed. Ukraine evaluated alternative connectivity under fire and returned to Starlink. The Pentagon split the NSSL Phase 3 launch award and still routed its irreplaceable lanes through SpaceX vehicles. xAI's compute corridor was built on Starlink uplink rather than a terrestrial alternative. An edge backed by a failed substitution attempt is worth more, analytically, than ten edges backed by a roadmap, and the graph weights them accordingly.

The fixed IPO price of $135.00/share values SpaceX at roughly $1.75T, about 90x FY2025 revenue ($18.7B) and roughly 90x against $6.6B adjusted EBITDA only if you ignore the −$4.9B GAAP net loss. The multiple is rich relative to aerospace peers (Boeing near 2x revenue, Lockheed Martin near 1.5x) and rich even against AI hardware peers an order of magnitude lower. The dependency graph argues the right comp is neither aerospace nor AI — it is global infrastructure that cannot be substituted at any price, closer to a pre-divestiture AT&T than a current telecom. On that frame a 90x multiple is defensible because the relevant denominator is not current revenue but the embedded option value of every counterparty contract awarded under no-alternative procurement, with Starlink ($11.4B, 61% of revenue, 63% margin, 10M-plus subscribers) already funding the launch cadence that keeps rivals off the manifest. The −$4.9B consolidated loss is the xAI build-out being financed by that margin. Whether the market agrees on day one is a separate, tactical question; the thesis here is structural.

Three things would weaken it materially. First, a peer launch provider — Blue Origin's New Glenn, Rocket Lab's Neutron, or a Chinese sovereign provider — achieving Falcon 9-class reusability and cadence; that would erode Chain 4. Second, a competing LEO constellation — Amazon's Leo (formerly Kuiper), the Chinese state networks, or a European sovereign mesh — reaching 5,000-plus active satellites with cross-link laser parity; that would erode Chain 1. For scale, Amazon's Leo had roughly 300 satellites in orbit by mid-2026 against a 3,236-satellite first-generation target, an order of magnitude short of parity and the gap widening with each SpaceX flight. Third, a regulatory event forcing cap-table separation of SpaceX from Tesla, xAI, and X; that would break the cross-collateralization that makes the chains mutually reinforcing. None of these has happened as of the filing. The IPO crystallizes the bet that none will happen before the lock-up ladder loosens (the first major release is the Q3-2026 earnings report) — and because the score moves only on realized events, not forecasts, each falsifier is a concrete, watchable trigger rather than a matter of opinion.