SpaceX's real business is data centers and telecom, not rockets

SpaceX's maiden earnings call as a public company tells a surprising story: the company bearing a space company's name is primarily operating as a telecommunications and data center rental business. Rockets, it turns out, are not the money-maker.

The real revenue drivers#

According to SpaceX's second-quarter results, space-based operations generated less than $1 billion in revenue and represented barely 10% of total company income. The actual engines of growth are Starlink, SpaceX's satellite internet venture, which pulled in $4.2 billion in revenue and was the only division to achieve operational profitability, and the company's emerging data center leasing business serving artificial intelligence firms.

Tellingly, SpaceX spent $15.8 billion on its AI and compute operations during the quarter—dwarfing the combined spending on space and connectivity sectors, each under $1 billion. The company now competes directly with specialized neocloud providers like CoreWeave by renting computing capacity to major players including Google, Anthropic, and Reflection AI.

How the data center business took over#

The trajectory began with Colossus 1, a Memphis facility Musk originally built for xAI's Grok chatbot. Technical complications—latency issues and incompatible chip mixes—made it difficult for xAI to use internally, so SpaceX opted to rent capacity to external clients instead. The strategy appears to be working: SpaceX's chief financial officer stated the company is "on a trajectory...to reach $100 billion of annualized revenue run rate," with Musk pushing the optimism even higher on the earnings call.

The uncomfortable math#

Yet revenue tells only part of the story. Data center expansion demands continuous heavy capital expenditure, and the compute market operates as a commoditized, cost-driven business where margins compress as supply increases. SpaceX faces stiff competition and the eternal problem of equipment obsolescence.

Looking forward, Musk has proposed an orbital data center constellation comprising up to 1 million satellites and discussed manufacturing facilities on the Moon powered by Terafab, his chip production company, and staffed by Optimus robots. These ventures remain largely speculative—regulatory filings contain minimal technical details, and no timeline exists for deployment.

Broader questions#