By Akash Sriram
July 10 (Reuters) – Elon Musk may have pitched a future where space powers AI, but Wall Street analysts say SpaceX’s near-term value remains firmly tethered to Earth, where it is building out the infrastructure underpinning the AI boom.
Infrastructure providers, particularly data centers, are poised to be among the biggest beneficiaries of the AI boom as businesses and consumers rapidly adopt the technology for applications ranging from software coding and robotics to everyday tasks such as shopping and planning.
SpaceX, a major AI player, has already started monetizing compute by striking deals with enterprise customers, including Anthropic, for its Colossus supercomputer clusters. In contrast, its orbital AI plan remains a longer-term opportunity that depends on the introduction of Starship rockets, lower launch costs and technological advances.
Company filings analyzed by Reuters and commentary from research houses show that revenue from its latest compute contracts is set to far exceed sales from other segments this year.
“The narrative that (orbital) will fundamentally disrupt terrestrial data centers is a little bit overblown. Any kind of displacement of terrestrial data centers is 10 years plus out,” said Anthony Milovantsev, a partner at consultancy firm Altman Solon.
J.P. Morgan expects SpaceX to expand terrestrial AI compute capacity to about 9 gigawatts by 2029 – roughly equivalent to four times the power generated by the Hoover Dam.
“Beyond 2029, we expect SpaceX to pivot to orbital compute for incremental capacity additions, while continuing to operate and maintain its terrestrial compute clusters,” the brokerage said.
TERRESTRIAL COMPUTE BEGINS TO PAY OFF
SpaceX’s deals with Anthropic, Alphabet’s Google and Reflection AI for its Colossus compute facilities are expected to generate more than $28 billion in annual revenue.
That figure far exceeds SpaceX’s 2025 AI revenue of about $3.2 billion and surpasses revenue from its launch and Starlink connectivity businesses individually, according to Reuters calculations.
Analysts cautioned that the contracts contain termination provisions and should not be viewed as guaranteed long-term recurring revenue.
SpaceX invested nearly $18 billion in AI infrastructure and development in 2025, including about $12.7 billion in AI-related capital expenditure and $5.1 billion in AI research and development, dwarfing spending on its space and connectivity businesses, according to company filings analyzed by Reuters.
Colossus and Colossus II together provide roughly one gigawatt of AI compute capacity, placing SpaceX among the world’s largest AI compute operators. Analysts expect that footprint to expand several-fold over the next few years.
Several brokerages also cite SpaceX’s $60 billion acquisition of AI coding startup Cursor as evidence the company is expanding beyond infrastructure into enterprise software, allowing it to monetize both AI applications and the computing capacity behind them.
ORBITAL AI REMAINS A LONG-TERM BET
Most brokerages that initiated coverage of SpaceX following its IPO are treating orbital AI as a longer-term opportunity, while basing their near-term financial forecasts around terrestrial AI infrastructure.
“We view the long-term viability of orbital data centers as unproven and highly dependent on key technological milestones that have yet to be fully realized,” BofA analysts said in a note.
Analysts broadly expect terrestrial compute to drive SpaceX’s growth and earnings for the rest of the decade, with orbital AI dependent on Starship achieving rapid reusability, lower launch costs and advances in satellite engineering.
Most say that Starship may eventually help deploy solar-powered computing satellites that could reduce costs associated with energy, cooling and land – some of the bugbears of terrestrial data centers.
The debate, analysts say, is no longer about SpaceX’s ability to build and monetize AI infrastructure. Rather, it is more about how quickly it turns that opportunity into a viable business beyond Earth’s atmosphere.
(Reporting by Akash Sriram in Bengaluru; Editing by Anil D’Silva)
