The growth of AI technology has created substantial demand for new facilities to support computing needs. AI systems rely on multiple GPU clusters operating in hyperscale data centers, where high power consumption is accompanied by significant heat generation.
AI data centers require more electricity than thousands of domestic households. Some upcoming major facilities could even reach an electricity demand that equals the consumption of two million residential units.
Tech companies are therefore considering new approaches to their computing facilities, as current power requirements pose challenges to operational efficiency.
Electricity will be the primary limitation. National electricity demand receives a major share from data centers. For example, in the United States, data centers accounted for more than 4% of total electricity consumption in 2023, a figure that has since increased due to the growth of AI computing.
The water and cooling problem
Additionally, data centers require substantial freshwater supplies for server cooling. AI hardware generates significant heat, prompting many facilities to use advanced cooling systems.
A 100-megawatt data center requires up to two million liters of water each day for cooling — equivalent to the daily water use of 6,500 typical households.
Data centers worldwide have experienced a rapid increase in water usage. In the U.S., they consumed 66 billion liters of water in 2023 — a major increase from 21.2 billion liters in 2014. This creates new challenges for the sustainability of local freshwater supplies and groundwater resources.
The implementation of data centers in space
Here comes the initiative of space-based data centers. This is supposed to be a solution to several existing obstacles. The concept is based on the idea that orbital infrastructure systems can leverage continuous solar power together with the extreme cold of the space environment.
The initiative is gaining increasing attention. At the latest GTC conference, Nvidia unveiled a new chip system designed for orbital AI data centers. Though the announcement had limited impact on Nvidia stock performance, it reflects growing interest in such projects. Elon Musk has also repeatedly emphasized the potential of space-based data centers and recently merged SpaceX and xAI to accelerate their development. If the strategy proves successful, it could boost future SpaceX stock.
But before we get there, several critical technical challenges must be addressed. First, orbiting solar panels can generate power continuously, giving them a theoretical advantage over Earth-based systems, which are limited by nighttime interruptions and atmospheric obstruction.
Yet powering a large-scale AI data center would still require vast solar infrastructure. High-performance computing clusters rely on continuous power supply at gigawatt capacities, necessitating extensive solar panel installations throughout space.
For both engineers and investors, the process of building and launching such an infrastructure remains extremely difficult. Additionally, even though the area is cold, the GPUs may not be cooled by the void’s cold. It simply doesn’t work that way.
Space’s cold can’t just “cool the GPUs”
The extreme cold of outer space doesn’t facilitate hardware cooling, as such systems must operate in a vacuum. Earth-based servers use either air-based or liquid-based cooling systems to release their generated heat.
The space environment lacks an atmospheric presence, making convection heat transfer impossible. The systems rely on thermal radiation, which requires large radiators to remove heat effectively. Thermal management, therefore, becomes a complex engineering challenge for all computational systems.
Once the power generation and the cooling are resolved, data transmission back to Earth presents another major obstacle.
Modern data centers on Earth depend on fiber-optic networks that deliver exceptional bandwidth with minimal transmission delays. A space-based data center would need to transfer processed data wirelessly using radio or laser communication systems.
The signals would face potential interference from weather conditions, cloud cover, and other atmospheric factors. This will cause signal delays, reduced quality, and connection interruptions. As a result, the user experience will differ from that of fiber-optic connections.
Therefore, maintaining hardware systems in space requires engineers to address three major challenges: energy requirements, cooling systems, and communication operational needs.
Data centers that use AI technology depend on thousands of GPUs and specialized chips that also require ongoing maintenance and replacement throughout their operational lifetime.
The hardware in space is further exposed to intense cosmic radiation that can damage the electronic component. On Earth, most terrestrial infrastructure is protected by the magnetic field, a barrier that does not extend to spacecraft, satellites, and space-based datacenters.
The concept of space-based data centers demonstrates how AI has evolved into a power-consuming commercial system. Theoretical orbital computing enables access to unlimited solar energy, which would reduce demand for earthbound energy resources. However, it creates major technical difficulties, as we’ve seen.
For now, the concept is solely a research project, though the increasing need for AI computational resources could bring it to life. And we hope that such an outcome won’t be a disaster for the investors.
