Space Exploration Technologies just finished its first quarter as a public company and the print already answers the wrong question. The top line nearly doubled, yet the equity is not a launch-services story. The second-quarter result is a cash-engine test: whether Starlink and a new AI compute book can carry Starship development and a gigawatt-scale data-center build without consuming the IPO proceeds. Management folded xAI into the registrant in February, then listed Class A shares in June and raised $85,675 million of net proceeds.
Connectivity is the only segment that prints GAAP operating profit, and it is the reason the consolidated loss narrowed. Starlink ended the quarter at 12 million subscribers, double the year-ago base, even as monthly ARPU slipped to $66 from last year's richer mix. Enterprise and government connectivity grew faster than the consumer layer, which is the mix shift that keeps the cash engine from being a pure consumer-broadband story. The AI segment flipped to positive adjusted EBITDA only because new cloud-service contracts recognized $1,600 million of incremental infrastructure revenue in a single quarter. That is a real book, and it is also a concentration: one unnamed AI customer already accounts for almost a fifth of consolidated sales.
The Space segment still loses money at the operating line because Starship research is being pulled forward on purpose. Adjusted EBITDA of $3,538 million is the number the market is using to underwrite a multi-trillion equity value, and it is a non-GAAP construct that adds back depreciation, share-based pay, and interest. Capital spending in the quarter was $18,369 million, almost all of it in AI compute, against first-half operating cash flow of only $3,466 million. The next several prints decide whether the cloud contracts convert into durable cash or whether the IPO cash simply funds a longer build. Can Starlink's cash and the AI book fund Starship and Colossus without a second trip to the equity window?