SpaceX Financials, Decoded:
A Profitable Engine Bolted to a Loss-Making Bet

SpaceX reported $18.7 billion in revenue for FY2025, up 43% from $13.1 billion in 2024 and more than double its 2023 revenue of $8.7 billion. But "make" is doing heavy lifting in that question. On a cash-profit basis the company generated $6.6 billion of adjusted EBITDA. On a GAAP basis — the one the SEC and auditors actually certify — it lost $4.9 billion, and it carries a $41.3 billion accumulated deficit, the scar tissue of two decades of reinvestment. The gap between those two numbers is almost entirely xAI, the consolidated artificial-intelligence segment, which posted a $6.355 billion operating loss on $3.2 billion of revenue in 2025. So the precise answer is two-part: the rocket-and-internet business makes a great deal of money, while the consolidated company, AI bet included, still loses money. Both facts live in the same S-1 (CIK 0001181412), and any headline that quotes only one of them is selling you half the income statement.

At the segment level, emphatically yes; at the consolidated level, no — and the distinction is the whole story. Starlink alone runs a 63% EBITDA margin on $11.4 billion of 2025 revenue, economics most software companies would envy, and the legacy space-plus-connectivity business is solidly cash-generative. Yet the consolidated entity posted a $4.9 billion GAAP net loss for 2025 against a $41.3 billion accumulated deficit. The swing factor is xAI, whose $6.355 billion operating loss is funded directly by the space business's cash flows. This is why contradictory headlines — "SpaceX is finally profitable" and "SpaceX loses billions" — run in the same week; they are describing different lines of the same statement. The honest framing is a profitable engine deliberately bolted to a loss-making, high-upside bet, both held inside one Musk-controlled share structure. Whether that is reckless or shrewd depends entirely on what you believe xAI becomes — which is the question the ~90x multiple on the valuation page is really asking.

Three places, in descending order. Starlink/Connectivity is the largest at $11.4 billion — 61% of 2025 revenue — spanning consumer, enterprise, maritime, aviation, and the Starshield government tier, and it carries a 63% EBITDA margin. Launch and space services account for most of the remaining legacy revenue, split between commercial satellite deployment and U.S. government missions for NASA, the Space Force, and the NRO. The newest leg is xAI at $3.2 billion, consolidated after the AI lab was folded into the company — though it contributes a $6.355 billion operating loss alongside that revenue. The trend that matters more than any single line: in 2021 SpaceX was a launch company with a satellite side project; by 2025, against $18.7 billion in total revenue, it is a connectivity company with a launch cost-center and an AI option attached. That inversion is the single most important fact in the filing, because it is what converts a cyclical contractor into recurring, unremovable infrastructure the rest of this site is built around.

Because the company chose to consolidate xAI, and xAI is mid-way through its capital-intensive build-out. Training frontier AI models and standing up compute burns billions before meaningful revenue arrives; xAI's $6.355 billion operating loss on $3.2 billion of revenue is exactly what the front half of that S-curve looks like. SpaceX is financing it with Starlink's cash flows — $11.4 billion of revenue at a 63% margin — and the IPO's roughly $75 billion of fresh capital. Strip xAI out and the legacy business is plainly profitable; leave it in, as the S-1 does, and the consolidated entity shows a $4.9 billion GAAP net loss against a $41.3 billion accumulated deficit. This is a deliberate decision to fund a moonshot off a cash cow inside a single share structure, not evidence the core is weak. The risk is concentration, not viability: one controller is routing a connectivity monopoly's profits into an AI bet, and public shareholders are along for both legs.

It has already lapped the sector on growth and is closing the gap on absolute size. SpaceX's $18.7 billion in 2025 revenue is approaching Lockheed Martin and Northrop Grumman territory, but it is compounding at 43% a year against the low-single-digit growth of the primes — and, unlike them, the majority of that revenue (Starlink, $11.4 billion) is recurring subscription income, not lumpy multi-year defense contracts. The sharper contrast is margin structure: defense primes earn high-single-digit to low-teens operating margins on cost-plus work, while Starlink runs a 63% EBITDA margin on a network SpaceX owns end to end. So SpaceX is not a better Lockheed; it is a different species — closer to a telecom-plus-cloud hybrid that happens to manufacture its own launch vehicles. That is precisely why the aerospace comp set, and its 1.5x–3x revenue multiples, is the wrong yardstick for a company the market is pricing as load-bearing infrastructure that defense, launch, and connectivity all route through.