SpaceX AI infrastructure bets on Earth-based compute while orbital plans remain years away

SpaceX AI Infrastructure Anchored to Terrestrial Compute

Elon Musk may envision a future where space powers AI, but Wall Street analysts maintain that SpaceX AI value remains firmly grounded on Earth. Major banks initiating coverage following the company's June IPO are bullish on SpaceX's terrestrial AI infrastructure, even as they treat orbital compute as a longer-term bet dependent on technological breakthroughs that remain years away1

1

.

The company's Colossus supercomputer clusters have already begun generating substantial revenue through deals with enterprise customers including Anthropic, Google, and Reflection AI. These terrestrial compute contracts are expected to produce more than $28 billion in annual revenue, a figure that far exceeds SpaceX's 2025 AI revenue of approximately $3.2 billion and surpasses revenue from its launch and Starlink connectivity businesses individually, according to Reuters calculations1

1

. Analysts caution, however, that these contracts contain termination provisions and should not be viewed as guaranteed long-term recurring revenue.

Massive Investment in AI Infrastructure and Vertical Integration

SpaceX invested nearly $18 billion in AI infrastructure and development in 2025, including approximately $12.7 billion in AI-related capital expenditure and $5.1 billion in AI research and development. This spending dwarfs investments in its space and connectivity businesses1

1

. Colossus and Colossus II together provide roughly one gigawatt of AI compute capacity, placing SpaceX among the world's largest AI compute operators.

Morgan Stanley initiated coverage with an Overweight rating and $300 price target, arguing that SpaceX's vertically integrated terrestrial and orbital compute stack could transform it into one of the most powerful AI infrastructure platforms. The bank's base-case model projects SpaceX revenue rising from about $45 billion in 2026 to $319 billion in 2030 and $3.3 trillion by 20403

3

. The analysts believe SpaceX can achieve industry-leading compute cost and time-to-power through vertical integration, targeting infrastructure capex of roughly $4-$5 per watt versus industry averages closer to $9 per watt.

Wall Street's Bullish Yet Cautious Stance

Major banks that served as underwriters for SpaceX's IPO released overwhelmingly positive research coverage, with Deutsche Bank calling the company "the apex of civilizational ambition" and Raymond James describing Starship as "the defining industrial innovation of our generation"2

2

. JPMorgan 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 Dam1

1

.

Cantor Fitzgerald emphasized that "a planetary infrastructure company does not compete within existing markets—it defines the cost structure of new ones," assigning an Overweight rating with a $246 price target2

2

. The $60 billion acquisition of AI coding startup Cursor signals SpaceX's expansion beyond infrastructure into enterprise software, allowing it to monetize both AI applications and the computing capacity behind them.

Orbital AI Compute Remains a Decade Away

Despite Musk's vocal support for orbital AI compute—recently claiming "SpaceX will be worth more than the rest of Earth if we accomplish our goals"4

4

—analysts view space-based AI infrastructure as unproven and highly dependent on key technological milestones. "Any kind of displacement of terrestrial data centers is 10 years plus out," said Anthony Milovantsev, a partner at consultancy firm Altman Solon1

1

.

Bank of America analysts noted that "the long-term viability of orbital data centers as unproven and highly dependent on key technological milestones that have yet to be fully realized"1

1

. The orbital AI plan depends on Starship rockets achieving rapid reusability, dramatically lower launch costs, and advances in satellite engineering.

Starmind Constellation and AI1 Satellite Unveiled

SpaceX revealed renderings of its AI1 satellite, part of the Starmind constellation dedicated to orbital AI compute. Each AI satellite will support chips from all providers and possess a 150 kW peak compute payload with an average payload compute of 120 kW. The satellites will have a wingspan of 70 meters and utilize a 150 kW solar array4

4

.

Morgan Stanley models orbital compute deployments beginning in 2028 at 160 MW, reaching 2.7 GW in 2030, 21 GW in 2032, and 364 GW by 2040, with orbital becoming the majority of total AI compute capacity by 2032. The bank estimates orbital compute could reach cost parity with current terrestrial data centers by 2031, with capex per watt falling from about $166/W in 2028 to $9/W by 20403

3

.

Anthropic Deal Fuels Infrastructure Buildout

The Anthropic deal, which commits $45 billion through May 2029 to use SpaceX's Colossus and Colossus II data centers, has sparked debate about whether the AI company is inadvertently funding a future competitor5

5

. Critics argue that Anthropic's money directly funds SpaceX's infrastructure buildout, potentially the same infrastructure that will eventually compete with Claude through xAI.

For investors watching the AI space race, the debate is no longer about SpaceX's ability to build and monetize AI infrastructure. Rather, it centers on how quickly the company can transform orbital AI compute from a visionary concept into a viable business beyond Earth's atmosphere. JPMorgan expects SpaceX to pivot to orbital compute for incremental capacity additions beyond 2029, while continuing to operate terrestrial compute clusters1

1

. The coming years will determine whether Musk's bold claims about space-based AI infrastructure can match the proven success of his Earth-bound data centers.