Space Exploration Technologies — Sum of the parts

A business with a revenue line is valued at the multiple its peers trade at; a business with none is valued as an option: pick a year it might exist (2030), build what it would earn, apply a multiple, bring the number back at 10% a year, and take off what it costs to build. Everything is taxed at 20%. An option that owns its units pays for them, and an option is never worth less than zero. A business already inside a filed line counts only for what it adds to that line.

Net cash is added once for the whole company; every peer is on EV/sales, so none of their cash is counted with it. Each case divides by the shares that would exist in it: Mr. Musk's 1,302.1M restricted shares vest only beside a million people living on Mars or 100 terawatts of compute a year off Earth, which no case reaches by 2030, so no case counts them — though the price is quoted on a count that does.

Why these multiples. Space: bear 2× — Northrop Grumman's multiple: a government contractor's, for a line whose revenue is flat year on year and loses money on Starship's development. base 4× — Twice a prime's, for a 67% gross margin on launch. bull 10× — A quarter of Rocket Lab's 38×, the one launch company priced on growth. Connectivity: bear 2.4× — AT&T's multiple: a network operator's, as if Starlink's growth were over. base 5× — Above T-Mobile's 3.1×, for growth no carrier has: the line grew 49% in the first half. bull 8× — Two and a half times T-Mobile's: a subscription network growing near 50% a year, at FY2025's 39% operating margin, priced like a platform rather than a carrier. AI: bear 2× — Below every peer: the line lost $6.4B from operations on $3.2B of revenue in FY2025, and its advertising half is shrinking. base 5× — Below Oracle's 7.3×, the lowest peer, for a line that loses more from operations than it takes in revenue. bull 10.6× — CoreWeave's multiple: AI infrastructure sold by the gigawatt, the half of the line that grew more than fourfold in the first half.

Why these builds. Orbital compute: Bear: Nothing in operation by 2030: the first satellites are planned for 2028 at the earliest, and an option that has not flown expires. Base: Ten gigawatts — a tenth of the company's goal for one year, reached two years late, as its record runs — at $10,000 a kW a year, a little above the $8,800 a kW its AI infrastructure line earned in Q2 2026, annualised, on the 1.0 GW of compute it had running, and at the cost on the ground: $12.7B of AI capex in 2025 for 0.5 GW of new compute, about $25,000 a kW. Bull: The company at its word: satellites from 2028 and its goal of 100 GW launched a year reached by 2030, so a year of it in orbit, built for less than on the ground ($20,000 a kW) and earning a wider margin on free power.

Thesis v1, updated Oct 20, 2026: First version: the 10-Q for the quarter ended 2026-06-30, with FY2025 from the IPO prospectus; three filed segments at their peers' multiples, orbital compute as the one option. The same question in one number: the reverse DCF for SPCX. Everything else: the SPCX summary.

What is a sum-of-the-parts valuation?

Split a company into its separate businesses, value each one on the multiples its own industry peers trade at, add the pieces up, and add net cash. It suits a company whose businesses are too different to be captured by a single multiple. This page does it in three tiers — bear, base and bull — and holds the sum against the market value of the shares.

How are businesses with no revenue valued?

As options: pick a year when the business might exist, build what it would earn from a few stated inputs, apply a multiple, bring the number back to today at a required return, and take off what it costs to build. Each tier is a different set of inputs, all printed. A bear tier of zero means the option is assumed to expire, and an option is never worth less than zero.

What does working backwards mean here?

Every part is held at its base value except one, and the page solves for the size that one part would have to reach for the sum to equal the price. It turns the price into a single number about the future that can be checked against the company's own reports.

Can I change the inputs?

Yes. The controls edit the base tier — multiples on the filed lines, the target-year build of each option, the required return — and every figure on the page re-draws. Bear and bull stay as published, so you are always moving one bar against two fixed ones.