SpaceX’s rockets remain the machinery people see. Its 2026 securities filing shows that connectivity had become the business paying most of the bills.

In 2025, SpaceX’s Starlink-driven Connectivity segment generated $11.387 billion of the company’s $18.674 billion in revenue. That is 60.98 per cent, which rounds to the 61 per cent in the headline. It was also the only one of SpaceX’s three reportable segments to record an operating profit.

Those numbers come from SpaceX’s June 2026 amended Form S-1. This is one corporate filing, and it is a historical accounting snapshot rather than a timeless description of the company. Even so, the segment table makes a financial structure visible that rocket launches alone tend to hide.

What 61 per cent means

The 2025 revenue split was $4.086 billion for Space, $11.387 billion for Connectivity and $3.201 billion for AI. Connectivity therefore contributed nearly three times as much revenue as the rocket and spacecraft segment.

The operating results are more revealing. Connectivity reported $4.423 billion in operating income on $6.964 billion of costs and expenses, an operating margin of about 38.8 per cent. Space recorded a $657 million operating loss. AI recorded a $6.355 billion loss. Across the combined company, that left a consolidated operating loss of $2.589 billion before interest, tax and other items.

So “only profitable” has a precise meaning here: Connectivity was the only reportable segment with positive income from operations. It does not mean every Starlink contract was profitable, or that SpaceX as a whole made an operating profit.

Starlink grew by adding customers, not by charging each one more

Connectivity includes consumer Starlink subscriptions as well as enterprise, aviation, maritime, government and mobile services. Its revenue increased 49.8 per cent from 2024 to 2025. SpaceX attributed $2.377 billion of the increase to consumer subscribers and $1.411 billion to enterprise and government customers.

The filing says subscriber numbers grew 99.9 per cent while average revenue per subscriber fell 11.2 per cent, mainly because of international expansion and lower-priced plans. That is a useful detail. The segment’s growth was driven by a much larger network, not simply by extracting more money from each existing household.

When we wrote in May that Starlink was becoming SpaceX’s financial engine, the argument rested on the network’s scale. The S-1 put audited segment dollars behind it. That scale has continued to grow: our latest count found that Starlink had passed 11,000 satellites in orbit in August.

The rocket business is the infrastructure behind the service

It would be a mistake to read this as rockets becoming unimportant. SpaceX flew 165 Falcon missions in 2025, a cadence we examined in our earlier look at Falcon 9’s extraordinary launch year. The filing adds a crucial split: only 43 were customer launches, while 122 were internal.

That is the bridge between the two businesses. Much of the launch system was working for SpaceX itself, placing its own infrastructure into orbit. Satellites are recorded as assets that include manufacturing costs and launch costs incurred by the Space segment, including an allocation for flight hardware. Those capitalised costs then appear over time as depreciation in Connectivity.

The segment results are real, but they do not describe two unrelated companies trading at market prices. They describe a vertically integrated system in which the rocket operation builds the physical platform that the connectivity operation monetises.

Why Space still posted a loss

The Space segment’s revenue rose 7.6 per cent in 2025, while its launch-service revenue stayed roughly flat. Its cost of revenue actually fell by $189 million, partly because increased Falcon reuse reduced depreciation by $240 million.

The loss came from investment further ahead. Space research and development jumped 63.7 per cent to $3.004 billion. SpaceX said the increase was primarily caused by accelerated spending on Starship production, test launches, engineering and the facilities intended to support future flights.

Falcon was mature enough to deploy Starlink at industrial cadence. Starship was still consuming money as a development programme. Segment accounting places both inside “Space”, which helps explain how a launch operation capable of 165 Falcon missions could still finish the year in the red.

The AI column is retrospective

There is one more complication. SpaceX did not legally acquire xAI until 2 February 2026. Because Elon Musk controlled SpaceX, xAI and X, the transaction was treated as a reorganisation of entities under common control. The filing therefore recast prior periods as if the businesses had already been combined.

That makes the 2025 AI column useful for comparing the three divisions management now oversees, but it is not a literal history of SpaceX’s legal structure during that year. The $6.355 billion AI operating loss belongs to the retrospectively combined presentation.

A better description of SpaceX

“Rocket company” still describes SpaceX’s engineering identity. It no longer describes its financial centre of gravity. In 2025, the customer-facing network produced most of the revenue and more operating profit than the other two segments could preserve.

The filing suggests a more accurate formulation: SpaceX is a connectivity and infrastructure company with its own launch system. The rockets make the network unusually difficult to copy. The network gives the rockets a vast internal customer. And, at least in the 2025 accounts, Starlink was the part turning that circular advantage into operating profit.