SpaceX has begun generating significant revenue from renting out data center capacity provided to major players in the artificial intelligence sector. Contracts with Anthropic and Google secure a monthly income for the company that exceeds SpaceX's annual turnover in 2025.
Anthropic has signed an agreement with SpaceX to pay $1.25 billion monthly (equivalent to 22.9 billion riyals) over three years. This deal, signed in May, grants Anthropic access to over 220,000 Nvidia GPUs and computing power exceeding 300 megawatts, comparable to the energy consumption of a small city. These capacities are located in the Colossus 1 data center in Memphis, Tennessee.
Several weeks later, the search giant Google also signed a lease agreement. Under the terms, Google will pay SpaceX $920 million (or 16.8 billion riyals) monthly, starting in October. This contract is estimated at 110,000 GPUs and will run until June 2029. In total, both contracts bring in $2.17 billion (or 39.7 billion riyals) per month, amounting to about $26 billion (or 476 billion riyals) annually. This sum surpasses SpaceX's entire annual revenue for 2025, which was $18.67 billion (or 341.7 billion riyals), according to the company's S-1 report.
Both companies are paying for infrastructure that was originally built for Musk's venture project—xAI. SpaceX acquired xAI in February through an all-stock deal valued at approximately $1.25 trillion (or 22.9 trillion riyals). After the merger, the combined entity went public in June at a price of $1.77 trillion (or 32.4 trillion riyals), and on its first day of trading on Nasdaq under the ticker SPCX, its value exceeded $2.1 trillion (or 38.4 trillion riyals), making it the largest initial public offering in history.
SpaceX is building data centers, and AI companies are paying to use them. The rental fee now surpasses all other sources of company income. The satellite internet business, Starlink, which was previously SpaceX's main source of revenue, generated $11.4 billion (or 208.6 billion riyals) in 2025. Just these two AI capacity rental contracts will provide an annual income more than double this amount.
The existence of these deals is due to the fact that the AI industry faces a problem that financial aspects alone cannot solve: there are not enough chips, electricity, or data centers on Earth to meet the growing demand. Anthropic needs capacity for its Claude models, while Google requires so-called 'bridge capacity' for its Gemini Enterprise agent AI platform, which is growing faster than Google's own infrastructure can handle.
In January, SpaceX applied to the U.S. Federal Communications Commission for permission to launch and operate up to one million satellites as part of its orbital data center system, internally named Starmind. The AI1 satellite design features a 70-meter wingspan and a peak computing load of 150 kilowatts, with the ability to swap equipment for various processors. Musk stated his intention to begin launching these satellites by 2028 using Starship, describing space as 'the only way to scale AI.'
The justification for this is continuous solar energy and reduced cooling costs, which are the two largest operational expenses for ground-based data centers. Nvidia has already developed a specialized chip for this purpose. The Space-1 Vera Rubin module, announced at GTC 2026, is designed to provide data center-level computing power in space, offering up to 25 times greater AI computing power per GPU compared to the H100. It is expected to be available in 2027.
The question of whether orbital data centers will be implemented on time remains open. However, terrestrial revenue is already real. The Anthropic contract alone could generate over $40 billion (or 732 billion riyals) over the three-year period. Additionally, if the Google contract is fully executed, it will add another $30 billion (or 549 billion riyals). The termination conditions of the agreements are quite flexible: Google can exit the deal with 90 days' notice after December, and the ramp-up period for Anthropic has already concluded. Thus, Musk has found a way to charge rent to companies creating the future, and regardless of whose model wins—Claude, Gemini, or its own Grok—the landlord gets paid.
Although Elon Musk may have presented the market with a vision of space-driven Artificial Intelligence, Wall Street analysts indicate that SpaceX's immediate and real value lies on Earth, where the company develops essential infrastructure for the growth of artificial intelligence, according to Reuters analysis.
The computing contracts signed by SpaceX with companies such as Anthropic, Google, and Reflection AI, intended for its Colossus supercomputer clusters, are projected to generate over US$ 28 billion in annual revenue. This amount exceeds the approximately US$ 3.2 billion in registered AI revenue for the company in 2025 and is higher than the individual revenues from rocket launch services and Starlink connectivity.
In 2025, SpaceX directed an investment of nearly US$ 18 billion towards AI infrastructure and development. This figure includes US$ 12.7 billion in capital expenditures related to AI and US$ 5.1 billion in research and development activities, significantly exceeding spending on the space and connectivity segments.
J.P. Morgan projects that SpaceX will increase its terrestrial processing capacity to approximately 9 gigawatts by 2029. This capacity would be equivalent to about four times the energy produced by the Hoover Dam in the United States. The Colossus and Colossus II complexes together offer about 1 gigawatt of computational capacity, positioning SpaceX among the leading global AI infrastructure providers.
Additionally, analysts point to the acquisition of the AI coding startup Cursor for US$ 60 billion as proof that SpaceX is diversifying its operations, moving from mere infrastructure to corporate software. This allows the company to monetize both AI applications and the computing capacity that supports them.
Musk's vision of using computing satellites in orbit remains a long-term possibility, but it depends on technological advancements that have not yet been fully achieved. Bank of America analysts, cited by Reuters, stated that the long-term viability of orbital data centers has not yet been proven and is strongly tied to pending technological milestones.
Anthony Milovantsev, a partner at the consulting firm Altman Solon, commented to Reuters that the idea that orbital AI will drastically revolutionize terrestrial data centers is an exaggeration. He estimated that any change in terrestrial data centers is more than ten years away.
J.P. Morgan predicts that SpaceX will keep its terrestrial clusters operational after 2029, gradually supplementing them with orbital capacity. Although solar-powered satellites could reduce costs related to energy, cooling, and physical space—major challenges for conventional data centers—this possibility depends on Starship achieving very low launch costs and a high reuse rate.
ispace, a Tokyo-based company, is expanding its lunar plans by incorporating SpaceX's Starship rocket. In an announcement made this Wednesday (8), the company secured 500 kg of cargo capacity on the Starship, the vehicle considered the largest and most powerful ever built, aiming for a lunar mission that could take place as early as 2030. This agreement was closed for a value of US$ 50 million (equivalent to R$ 258.2 million), as reported by Tokyo Brief.
Takeshi Hakamada, founder and CEO of ispace, expressed satisfaction with the possibility of using the Lunar Access Integration service through the partnership with SpaceX. He emphasized that the high-capacity and moderately priced lunar transport offered by Starship is fundamental to enabling the sustainable lunar economy that ispace aims to establish.
This statement suggests that ispace may become a recurring customer of Starship. The company intends to use the giant vehicle to transport its new Mobile Cargo System (MCS) to the lunar surface. The MCS is described as a flat platform rover capable of carrying up to 500 kg across the lunar terrain. The initial mission with the MCS aboard Starship is scheduled to launch only from 2030, depending on SpaceX's ability to make Starship a fully operational vehicle, given that it has already completed 12 test flights, all suborbital.
ispace already has flight experience with SpaceX. In 2022 and again in 2025, Falcon 9 rockets transported the Japanese company's robotic rover HAKUTO-R. Although HAKUTO-R successfully reached lunar orbit on these two occasions, it suffered landings during the landing attempts. Starship, on the other hand, is SpaceX's super-heavy vehicle, designed for full reusability and capable of placing up to 150 tons into low Earth orbit.
The development of the rocket began some time ago, first presented by Elon Musk during the International Astronautical Congress in Mexico in 2016. Predictions about its operational readiness have constantly changed. In 2021, SpaceX aimed to conduct the first lunar mission before 2024, but development delays postponed that date. In 2024, NASA was scheduled for the first crewed lunar landing of the Artemis program, but this plan changed. NASA contracted Starship as the lunar lander module for this event, which is now scheduled for Artemis 4 at the end of 2028, with agency officials responsible for citing Starship as a factor in the Artemis schedule delays.
NASA and ispace are not the only customers reserving seats on Starship destined for the Moon. Yusaku Maezawa, a Japanese businessman, announced the #dearMoon project in 2018, reserving Starship to take himself and a group of artists on a potential first crewed mission of the vehicle around the Moon. However, due to successive delays of Starship, Maezawa canceled the flight in 2024.
Nevertheless, the momentum for lunar missions with Starship is increasing. NASA already has two successes in its Artemis history: Artemis 1, uncrewed, which reached lunar orbit at the end of 2022, and the Artemis 2 flight, with four crew members, conducted around the Moon last month. The agency is preparing for Artemis 3, which should test rendezvous and docking operations with the NASA Orion capsule and two crewed lunar lander modules—Starship and Blue Moon from Blue Origin—in Earth orbit in mid-2027, if planning is met.
Thus, ispace positions itself as a crucial element in a probable lunar race. ispace stated that the advent of rockets capable of transporting large volumes of cargo to the Moon should accelerate the implementation of lunar infrastructure, covering energy, communication, construction, data, and mobility. The company added that establishing this infrastructure on the lunar surface will reduce obstacles for future projects, promoting rapid expansion in the transport of smaller loads for exploration, technological validation, and business development. As demand for missions increases, the cargo capacity of the Mobile Cargo System units will also grow. In addition to the new MCS design, ispace also plans three lunar landing missions with its ULTRA Lander, scheduled for 2028, 2029, and 2030.
