An analysis has been raised that SpaceX's artificial intelligence business revenue will come, for the time being, from ground-based data centres rather than space. Elon Musk is promoting an era of orbital computing over the long term, but the view is that tangible AI sales currently come from operating large-scale supercomputer clusters.

On July 12 local time, blockchain media outlet Cryptopolitan reported that Wall Street analysts are looking for the core of SpaceX's AI business in ground-based AI infrastructure rather than "space computing". Orbital data centres are a long-term vision, while current AI revenue is being driven by data centre and computing service contracts, the analysis said.

SpaceX is already providing AI computing services to corporate customers. Customers include Anthropic, Alphabet's Google and Reflection AI. Expected annual revenue from its Colossus supercomputer cluster contracts with them was estimated at more than $28 billion. Some assessments also say it is hard to view all of that as stable recurring revenue because the contracts include termination clauses.

The scale of its existing AI business is also growing rapidly. SpaceX's AI-related revenue in 2025 was tallied at about $3.2 billion. The outlet reported that revenue from computing services was larger than its rocket launch business or its Starlink business when each was compared on a standalone basis.

It is also aggressively expanding AI investment. SpaceX spent about $18 billion last year on AI infrastructure and research and development. Of that, about $12.7 billion was used to build infrastructure such as data centres, and $5.1 billion was used for R&D. That exceeds the scale of investment in its space launch and satellite communications businesses.

The core of its AI infrastructure is the Colossus and Colossus II supercomputer clusters. The combined computing capacity of the two facilities is estimated at about 1 gigawatt, and SpaceX is being assessed as one of the world's largest AI computing operators through them. JP Morgan forecast that the company's ground-based AI computing capacity would expand to about 9 gigawatts by 2029.

There is also speculation that its business direction will not stop at simple server leasing. Brokerages cited by Reuters interpreted SpaceX's $60 billion acquisition of AI coding startup Cursor as a sign it is expanding its AI applications business. The fact that its AI research organisation, SpaceXAI, co-developed Grok 4.5 with Cursor is also being noted in the same context.

By contrast, assessments are prevailing that the space data centre business still needs more time. Anthony Milovanchev (앤서니 밀로반체프), a partner at Altman Solon, called expectations that orbital computing would replace ground data centres "somewhat exaggerated" and predicted it would take more than 10 years for real change to appear.

There are also significant technical challenges. For space computing to become reality, Starship's low-cost, high-frequency launch system and next-generation satellite hardware must first be secured. The point is that these core technologies have not yet been verified at scale. Bank of America also assessed that the orbital data centre business depends heavily on achieving unproven technological milestones.

Still, there is longer-term potential. If SpaceX uses Starship to place solar-powered computing satellites into orbit, it could cut power, cooling and land costs borne by ground-based data centres.

For now, market attention is focused less on the future of space data centres and more on how quickly SpaceX can expand its ground-based AI infrastructure business. The industry view is that, for the time being, the key variables determining the value of SpaceX's AI business will be the pace of data centre expansion and whether it can maintain contracts with large customers, not space computing.