Elon Musk’s company is no longer primarily a rocket builder. The latest quarterly report shows SpaceX earned over $15 billion from its AI-focused computing division alone, nearly ten times the revenue from its spaceflight operations. This shift highlights a dramatic pivot from the company’s origins. While SpaceX launched into the public spotlight with its ambitious spaceflight goals, the numbers now reveal a vastly different picture of where the company’s real financial momentum lies.
Shift in Revenue Sources
SpaceX’s space business, including rocket launches, generated less than $1 billion in the second quarter. That means under 10 percent of total revenue came from the sector that once defined the company. Meanwhile, the Starlink satellite internet service brought in $4.2 billion and turned a profit, but it is still outpaced by the computing arm. The contrast underscores how the company’s identity is evolving away from its space roots. The rocket business, which once served as the company’s core narrative and public spectacle, now plays only a minor role in SpaceX’s financial results.
The company’s neocloud division, which rents out data center space to AI developers, is now its primary growth engine. In Q2, it spent nearly $15.8 billion on AI infrastructure. Analysts have noted that this is a significant investment, and the money is being directed toward a rapidly growing field. According to Bloomberg, AI infrastructure spending is expected to hit $65 billion next year. The financial outlay is not just a sign of ambition but also an indicator of how seriously SpaceX is treating the AI business.
That’s a sharp turn for a company that once billed itself as building humanity’s escape plan. Now, the real action is in Memphis, where SpaceX’s Colossus 1 data center houses a mix of new and old chips—creating bottlenecks and latency problems for training AI models like Grok, Musk’s own AI. Colossus was originally intended to serve as the backbone for xAI’s ambitions, but early struggles led to the decision to lease the infrastructure to third parties. The mix of hardware has proven problematic, making training in-house models inefficient.
Neocloud as Growth Engine
In a twist, xAI ended up renting Colossus 1 out to other firms instead of using it for its own purposes. Musk now says just 10 percent of the compute the company builds will go toward Grok. This shift highlights the company’s financial pragmatism. With Grok failing to meet expectations, SpaceX found a more immediate use for the data center by leasing it out to other AI developers. The move not only generated revenue but also gave the company a foothold in a lucrative sector.
The company’s financial chief, Bret Johnsen, said partnerships with Google, Anthropic, and others—along with the Cursor acquisition—put SpaceX on track to hit $100 billion in annualized revenue run rate by year-end. Musk was even more confident, saying the $100 billion milestone is not in question. The optimism is based on the rapid pace of new contracts and the potential of the Cursor acquisition. These deals suggest that SpaceX is betting heavily on its ability to provide high-quality compute resources to a growing pool of AI clients.
Musk's Strategic Shift
Still, building data centers is expensive and competitive. SpaceX is now in the same race as CoreWeave and Nebius, where speed and cost matter most. As more companies build out compute capacity, the value of each unit drops. That means SpaceX might struggle to maintain its pricing power as more players enter. The bare-metal data center market is a tough one to navigate, and competition is fierce. As the market becomes more crowded, the pressure on profit margins will increase. SpaceX will need to scale quickly and efficiently to stay ahead.
And yet, Musk continues to think bigger. The company has proposed an orbital data center consisting of as many as 1 million satellites. But the FCC application for this project offers little technical detail—no schedule, no sizes. It reads more like a PR stunt than a serious blueprint. The lack of specifics makes it clear that the proposal is more about signaling ambition than actual execution. For now, the orbital data center remains in the realm of futuristic ideas, with no clear path to becoming a reality.