(Image credit: fotograzia/Getty Images)ShareShare by:
- Copy link
- X
Share this article 0Join the conversationFollow usAdd us as a favored source on GoogleNewsletterSubscribe to our newsletter
As artificial intelligence (AI) paradigms continue their expansion and demand increasingly more energy, researchers are initiating inquiries not about their ability to be trained — but rather their optimal location. This notion forms the underpinning of Google Research’s fresh proposition to examine AI framework positioned in space, an idea that hovers between sincere scientific investigation and cosmic excess.
The notion, termed “Project Suncatcher” and detailed within a document submitted on Nov. 22 to the pre-publication arXiv repository, considers if forthcoming AI tasks could be executed using satellite groups fitted with specific accelerators and energized predominantly through solar means.
You may like
-
‘Rainbow-on-a-chip’ could aid in controlling AI energy consumption — and it arose unexpectedly
-
Scientists claim they’ve taken out a key AI limiter — presently they are able to handle computations ‘at the velocity of light’
-
‘It will not likely be so much a ghost area as a zombie end of days’: How AI could irreversibly alter our use of the internet
The inclination to explore beyond Earth for AI framework isn&t arbitrary. Data hubs presently use up a noticeable portion of the globe’s power resources: current projections evaluate global data-center electricity use at around 415 terawatt-hours in 2024, or approximately 1.5% of total global power usage, with forecasts implying this figure might more than double by 2030 as AI workloads escalate.
Power companies in the U.S. are presently forecasting that data hubs, mainly motivated by AI activities, will make up around 6.7-12% of total electricity demands in some locations by 2028, leading certain executives to caution that there merely “isn’t sufficient electricity on the grid” to maintain unrestrained AI development without noteworthy new generation potential.
Within that context, proposals such as data facilities placed in space start to appear less like a sci-fi amusement and further as a sign of a sector confronting the tangible restrictions of terrestrial power and refrigeration. Conceptually, space-based data hubs appear to be an ideal fix. Yet, in practical terms, certain professionals remain dubious.
Reaching for the stars
Joe Morgan, acting as COO at data facility infrastructure entity Patmos, speaks directly on the immediate upcoming possibilities. “What won’t transpire in 2026 involves the whole ‘data facilities in space’ facet,” as he relayed to Live Science. “Potentially one of the technological magnates might come near to accomplishing it, though aside from boasting privileges, why go that route?”
Morgan notes that this industry has continually entertained various extreme cooling strategies, stretching from mineral-oil immersion to subaquatic installations, merely to drop them upon acknowledging their implementation-based actualities. “There persists ongoing publicity concerning constructing data facilities under the oceans, although any thermal positives are extensively surpassed by the complication of changing parts,” as he said, highlighting that rapid hardware turnover remains core to contemporary computing.
This turnover factor forms the center of the doubt encompassing AI located in orbit. GPUs and specialized accelerators suffer depreciation at a fast rate as innovative configurations provide game-altering advancements on a bi-annual tempo. On Earth, racks can be exchanged, cards swapped and systems modernized without pause. Within orbit, each maintenance action calls for launches, docking actions or automated servicing systems — none of which expand effortlessly or economically.
“Who would be eager to board a spaceship to upgrade the orbital framework every year or two?” inquires Morgan. “Imagine that a key element malfunctions? Actually, dismiss that, what can be said about latency?”
You may like
-
‘It will not likely be so much a ghost area as a zombie end of days’: How AI could irreversibly alter our use of the internet
-
Contentious startup’s intention to ‘vend daylight’ through gigantic reflectors across space might prove ‘catastrophic’ and ‘appalling,’ express stargazers
-
The AI-created image of a feline navigating on a fruit occurs because of offspring mining the soils for poisonous elements. Is that really justifiable?
Latency constitutes more than just a side issue. Majority AI duties lean strongly on interlinked setups possessing tremendously fast interconnects, inside data facilities and across diverse facilities. Google’s plan leverages laser-centered inter-satellite links designed to mirror these interconnections, even though the natural sciences continue to present inescapable conditions. At low Earth orbit, turnaround latency to ground basepoints remains unavoidable.
“Putting the servers within orbit represents an unwise decision, unless your patrons equally exist within orbit,” communicates Morgan. Nevertheless, opinions vary regarding immediate dismissal. Paul Kostek, as a distinguished IEEE affiliate and systems engineer functioning at Air Direct Solutions, shares that enthusiasm reveals authentic concrete stress across terrestrial infrastructure.
“The intrigue in positioning data centers across space has heightened with climbing construction costs for facilities here on earth,” affirms Kostek. “Several advantages exist regarding space-based or Moon-based facilities. Primarily, continuous access to solar power… and secondarily, the capability to chill facilities through diffusing extra heat across space compared to employing water.”
From a purely thermal perspective, these notions resonate. Heat dissipation represents one of the stringent constraints across computation, with earth-grounded data centers facing escalating limitations imposed through water supplies, grid capabilities, and community-level environmental conflicts.
The counteraction directed towards terrestrial AI infrastructure is not limited to energy and water matters; fears concerning health increasingly feature in the discourse. Within Memphis, locals close to xAI’s extensive Colossus data center voices apprehensions regarding air purity and lasting respiratory impacts, with individuals reporting exasperated signs including anxiety about pollution-related ailments given the facility’s operation commenced. Across various states, opposers of proposed hyperscale data hub undertakings frame their opposition surrounding probable health and environmental consequences, suggesting large sites might degrade localized air and water grades coupled with escalating prevailing health burdens.
Placing data centers across orbit would dismiss certain limits, although simultaneously replacing them with others.
Staying grounded
“The technological queries that demand resolutions involve: Can the current processors employed across data centers present on Earth survive within space?” observes Kostek. “Would those processors withstand solar events or elevated radiation levels upon the Moon?”
Google analysts initiated earlier investigations surrounding several of these inquiries through earlier work around Project Suncatcher. That team identifies radiation assessments of its Tensor Processing Units (TPUs) together with modelling relating to how tightly arranged satellite formations can underpin extensive-bandwidth inter-satellite links requisite for distributed processing. Even so, Kostek emphasizes how these efforts remain investigational.
“Preliminary evaluations get performed to determine the operational validity of data centers residing within space,” he shares. “Despite remaining technical obstacles and a several-years gap for implementation, that methodology could eventually present one effective pathway for accomplishing expansion.”
That single word — expansion — could act as the factual clue. According to selected investigators, this persuasive logic concerning off-planet processing hardly has to do with servicing terrestrial individuals whatsoever. Christophe Bosquillon, the co-chair associated with Moon Village Association’s working collection focused on Disruptive Technology & Lunar Governance, posits that space-based data facilities seem to present improved sense given framework intended for space per se.
“With humanity scheduled to soon create one stable lunar presence, some infrastructural core geared toward future data-driven lunar industry alongside the cis-lunar economy seems justified,” he spoke to Live Science.
Based on these understandings, data hubs located within space should not substitute those on Earth, but instead exist as mechanisms enabling cosmic activities, handling everything ranging from lunar sensor data up to automated systems and navigation.
“Affordable power forms one pivotal problem across every activity, plus that factor would incorporate one nuclear element positioned alongside solar power including matrices comprising fuel cells and batteries,” Bosquillon clarified, furthermore identifying how the obstacles expand beyond the engineering realm up to governance, jurisdiction plus global collaboration.
RELATED STORIES
—Why do AI chatbots consume substantial power?
—’Rainbow-on-a-chip’ could aid in controlling AI energy consumption — and it arose unexpectedly
—Scientists claim they’ve taken out a key AI limiter — presently they are able to handle computations ‘at the velocity of light’
Significantly, space-based processing can detach responsibilities insusceptible to latency stemming completely from Earth. “Resolving the energy problem across space coupled with detaching responsibilities stemming completely from Earth, towards processing Earth-related details susceptible to less latency… indicates merit,” expressed Bosquillon, furthermore extending toward considering space in conjunction with the Moon acting as one safeguarded repository geared toward “civilizational” details.
Recognizing it under those terms, Google’s idea sounds less like one answer directed toward prevailing data hub shortages yet rather like one investigation directed toward prolonged computation’s physics. Since AI edges closer to planetary-scale power consumption, the concern would not indicate whether Earth maintains sufficient capabilities, although whether investigators could afford dismissing environments whereby energy proves plentiful with everything difficult.
Presently, AI situated across space holds strictly in experimental realms. No matter if this idea ever leaves Earth’s magnetic drag may rest less within solar elements together with lasers compared to how desperate power competitiveness occurs.
TOPICSnews analyses
Carly Page
Carly Page serves as a technology journalist and copywriter possessing more than a decade in spanning cybersecurity, forthcoming technological solutions, including digital protocols. Formerly, she functioned in senior cybersecurity reporter at TechCrunch.
Presently functioning as a freelancer, she composes news content, analyses, interviews, plus comprehensive features targeting publications encompassing Forbes, IT Pro, LeadDev, Resilience Media, The Register, TechCrunch, TechFinitive, TechRadar, TES, The Telegraph, TIME, Uswitch, WIRED, and various ones. Carly equally develops copywriting plus editorial content targeting technological firms in conjunction with happenings.
Show More Comments
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.
LogoutRead more
‘It will not likely be so much a ghost area as a zombie end of days’: How AI could irreversibly alter our use of the internet
Contentious startup’s intention to ‘vend daylight’ through gigantic reflectors across space might prove ‘catastrophic’ and ‘appalling,’ express stargazers
The AI-created image of a feline navigating on a fruit occurs because of offspring mining the soils for poisonous elements. Is that really justifiable?
Experts divided over claim that Chinese hackers launched world-first AI-powered cyber attack — but that’s not what they’re really worried about
Some people love AI, others hate it. Here’s why.