One rocket, borrowed money

2026-10-28

The company now listed as SpaceX is three businesses in one filing: launch, connectivity and compute. The ambitions are enormous and well advertised. What matters to anyone holding the paper, or lending against it, is narrower. Two things have to work. One is a rocket. The other is a balance sheet.

The rocket

Launch is the foundation. The company has flown roughly 670 orbital missions with a success rate above 99%, and has carried more than four fifths of all mass sent to orbit since 2023. That record was built on Falcon, whose booster lands and flies again. Starship is the larger vehicle on which every future plan depends, and its hard part is not the booster. It is the upper stage.

The upper stage comes back through the atmosphere at far higher speed than a booster and has to survive temperatures above 2,500 degrees Fahrenheit. The Space Shuttle faced the same problem and needed months and more than a hundred million dollars between flights.
The company's own filings give its rockets a useful life of five to twenty-five flights. Some Falcon boosters have flown more than thirty times. Starship's upper stage will start at the low end.
Nominal payload is about 100 tonnes. The most recent test carried 44. A later version is meant to lift 200.
One launch pad operates today. Three are expected soon and five within about a year. Current permissions allow something like 145 Starship launches a year across known sites. Management talks about tens of launches in 2027, hundreds in 2028 and thousands after that, which means permissions, pads, engines and fuel all have to scale by an order of magnitude.
The factory at Starbase is designed to build a thousand rockets a year by 2030. A methane pipeline and new liquid-oxygen capacity are being arranged because thousands of launches need fuel in volumes no launch site has ever handled.

Why does all this matter beyond the mission? Because the orbital compute plan, the one that produced the largest numbers in the prospectus, needs thousands of Starship flights by the early 2030s. The connectivity business needs far fewer. If the upper stage cannot be reflown quickly and cheaply, the compute story slips by years and the launch count that justifies the factory, the pads and the pipeline slips with it.

The balance sheet

The second thing to understand is how it is paid for. The space and connectivity businesses together produced about 7.8 billion dollars of adjusted segment earnings in 2025, up more than half on the year before. That is the cash engine. The company built those businesses on roughly 9 billion dollars of primary equity, which is remarkable by any standard.

The compute buildout is a different animal. It is front-loaded, largely pre-revenue, and street models do not have the company generating free cash before 2031. The gap is meant to be filled with debt. A 25 billion dollar debt raise followed the listing almost immediately, and street models carry several hundred billion dollars of further borrowing across the rest of the decade, with the heaviest years in 2028 and 2029.

Read that way, the listing was less an equity raise than the creation of a public credit profile. A listed company with a growing, cash-generative base can issue at scale, pursue an investment-grade rating and broaden its lender base. The connectivity earnings are the coverage; the compute plan is what the coverage is borrowed against.

Connectivity, the part that already works

The broadband business is the steadiest piece and the one with the clearest numbers.

The next satellite generation carries roughly eleven times the capacity of the current one, and a Starship lifts sixty of them. Street estimates put the cost of a terabit per second of capacity falling from around twelve million dollars to under two million once that combination is flying.
Terminals have fallen from 599 dollars to as low as 199, with a monthly rental option in some markets. Cheaper capacity and cheaper terminals are what let the service compete on price in the roughly 31 million American homes that will probably never see fibre, and in emerging markets with little competition at all.
Mobile is the open question. The three large American carriers have said they are not interested in a wholesale deal, so near-term mobile growth comes from international partnerships, direct-to-device service abroad and machine connections rather than American consumers. The spectrum auction in mid-2027 will show how seriously the company intends to build on the ground.

Our read

Strip away the Mars material and this is a company borrowing against a utility to build a factory in orbit. The utility is real and growing. The factory depends on one engineering milestone, the quick and cheap reuse of a rocket's upper stage, that has never been achieved by anyone.

For a holder of pre-listing paper, that framing sets the watch list: upper-stage reflight times, pad approvals, launch counts against the stated schedule, and each debt issuance's pricing. For a credit investor, the question is simpler. The connectivity cash flow is the thing being lent against, and it has to keep compounding while the compute plan absorbs capital for years.

We spend our time on access to companies like this before they list and on the secondary market in their shares afterwards. The lesson here is that a prospectus of this size rewards reading the engineering appendix and the financing section first, and the market-size pages last.