Contents
Key Takeaways
- Launch expenses need to decrease to under $300/kg for orbital computing facilities to become economically feasible — today’s rates range from $1,500–$3,600/kg
- Building a 1-GW space-based data center would exceed $100B at present rates compared to $35B–$50B for terrestrial equivalents
- BNP Paribas projects Google, Amazon, and Meta as frontrunners for initial experimental deployments once costs decline
- According to Elon Musk, space will become the most cost-effective location for AI infrastructure in the next 30–36 months
- Starcloud, supported by Nvidia, successfully deployed the first Nvidia H100 GPU beyond Earth’s atmosphere in November 2025
The concept of operating data centers in orbit is transitioning from speculative fiction to serious strategic planning. BNP Paribas recently released a comprehensive assessment of this emerging possibility, acknowledging its future promise while emphasizing current economic obstacles.
With today’s transportation costs to orbit — ranging from $1,500 to $3,600 per kilogram — constructing a 1-gigawatt data center in space would exceed $100 billion. By comparison, building an equivalent facility on Earth costs approximately $35 billion to $50 billion.
According to BNP analyst Nick Jones, the financial institution does not consider orbital computing facilities as a “viable near- to medium-term solution.” He pointed to elevated launch expenses, costly space-certified equipment, and complications involving thermal management and power distribution systems in space environments.
To achieve economic viability, BNP Paribas calculates that launch costs must decrease below $300 per kilogram. This represents a dramatic reduction from current market rates.
Should transportation costs reach that threshold, BNP Paribas anticipates that Google, Amazon, and Meta will be among the pioneering organizations conducting preliminary proof-of-concept trials with orbital computing infrastructure. The financial institution did not provide a precise timeframe for when such conditions might materialize.
The Energy Crisis Driving the Space Push
The momentum toward extraterrestrial computing stems largely from artificial intelligence’s enormous power requirements. Terrestrial data facilities are drawing unprecedented electrical loads. According to Department of Energy statistics, US data centers consumed approximately 4.4% of the nation’s total electricity in 2023.
McKinsey projects that satisfying global data center demand through 2030 will necessitate $6.7 trillion in infrastructure investment. Technology sector capital expenditures are forecast to reach $600 billion in 2026, with Amazon independently planning $200 billion in spending.
Elon Musk has positioned orbital data centers as a fundamental component of SpaceX’s long-term vision. He has projected that within 30 to 36 months, space will emerge as the “most economically compelling place” for deploying AI computing resources. SpaceX aims to establish a constellation of one million satellites functioning as orbital computing nodes, each producing approximately 100 kilowatts of processing capacity per ton.
Musk’s rationale centers not primarily on energy cost reduction but on energy accessibility. He has highlighted that electricity generation outside China has remained essentially stagnant, creating uncertainty about power sourcing for expanding terrestrial data center networks.
SpaceX Advances Beyond Planning Stage
SpaceX has progressed from conceptual discussions to active implementation. Michael Nicolls, who serves as vice president of Starlink Engineering, announced via X that the organization is recruiting for “many critical engineering roles” supporting space-based computing infrastructure, including a Space Lasers Engineer position located in Redmond, Washington.
The company revealed plans to acquire Musk’s artificial intelligence venture xAI, emphasizing orbital AI infrastructure as a strategic long-term objective.
Experimental Deployments Underway
In November 2025, Nvidia-supported startup Starcloud successfully placed the first Nvidia H100 GPU in orbit aboard a SpaceX Falcon 9 launch vehicle. The Starcloud-1 satellite massed approximately 60 kilograms — comparable to a compact refrigerator.
Starcloud’s ambitious objective envisions a 5-gigawatt orbital computing facility measuring roughly 4 kilometers across, equipped with extensive solar collection and thermal management arrays.
BNP Paribas observed that over extended timeframes, technological progress in ground-link communications, cooling methodologies, and solar energy systems could narrow the operational cost gap between orbital and terrestrial data center facilities.
