SpaceX is expected to burn through $350 billion in cash by 2030 as part of a supercharged growth plan that would also see annual revenue from its AI division rise 100-fold from last year.
That's one forecast provided by investment bank Goldman Sachs, the lead underwriter of the SpaceX IPO. In it, it lays out the justification for the $1.75 trillion valuation the aerospace and AI company is reportedly hoping to achieve at IPO. Revenue from its AI division is projected to rise from $3.2 billion in 2025 to $322 billion by the end of the decade.
Total revenue for the entire business will supposedly reach $474 billion, meaning the AI side of the business is expected to be the main driver of revenue over the next half-decade. That would explain why the vast majority of the $350 billion it plans to spend will go to capital expenditures mainly tied to the AI business.
Goldman said it expects SpaceX to burn $120 billion of that this year and next, with a significant ramp-up over the next three years. That is a steep step-up from xAI’s earlier spending levels, as it has burned through approximately $23 billion since inception.
The AI research lab has already leased some of its available capacity to Anthropic, suggesting that its chatbot, Grok, is not seeing the same level of usage as rivals such as ChatGPT, Claude, and Google Gemini. It recently reshuffled a lot of its executive team at xAI to refocus on enterprise clients and coding.
With the Terafab plant also in the works, the compute buildout may ultimately lead SpaceX to transition to an infrastructure company if the software side fails to draw in enough interest. There is a huge amount of demand for neocloud resources at the moment, as shown by the surge in the values of many top neoclouds, such as CoreWeave, Nebius, Lambda, and Nscale.
Satellites and moon missions
However, the majority of SpaceX revenue currently does not come from xAI. Its Starlink satellite broadband division accounts for 61% of its total revenues in 2025, and launch services account for 22%.
According to Goldman's analysis, Starlink revenue is expected to rise to $144 billion by 2030. If we use figures from Omdia, that would make it roughly equivalent to 11% of the global telecom connectivity market.
Rocket launches, the thing that SpaceX is most known for, will only double in revenue from 2026 to 2030, reflecting the slow burn of government contracts and the nascent market. While NASA aims to get people back on the Moon by the end of the decade with a continued program of return missions, it's not clear whether there'll be a large enough industry outside this to support significant revenue growth.
Data centers in space
Part of the alignment between the two companies at the time of merger was the potential for SpaceX to deploy AI data centers in space.
These data centers, powered by solar panels, could be far more efficient than those on Earth, though this is still just a theory at the moment. Engineers still need to overcome radiation in low orbit, inadequate cooling mechanisms, and communication bandwidth constraints.
Google is reportedly working with SpaceX to launch a pilot of Project Suncatcher, its own orbital data center plan, within the next 12 months. It's not clear how far along SpaceX's own plan to launch orbital data centers is, or whether it was just a save-face move to merge the two companies.
Also read: A related question is whether Tesla can turn its humanoid robot ambitions into a viable business, with Elon Musk recently suggesting Optimus could reach commercial sales in 2027.