Space-Based Data CentersSpace-Based Data CentersCoverage from Morgan Lewis, PostQuantum.com, and others
00/00/0000
DailyWeekly
Space-based data centers would place computing hardware on satellites or orbital platforms, with proposals often positioning them as a way to serve AI demand while drawing on solar energy and avoiding some terrestrial power, land, and water constraints.
Projects such as Google’s Project Suncatcher and concepts from SpaceX and Blue Origin illustrate different approaches to orbital computing. Making the systems practical depends on solving demanding engineering and economic problems: rejecting heat in vacuum, protecting and maintaining hardware, launching enough equipment affordably, and moving data through reliable communications links. Early uses may favor workloads that can tolerate delays or are generated in space.
This broad overview introduces the proposed benefits of orbital computing alongside the engineering and economic limits that keep it at an early stage.
It adds concrete launch-cost and heat-rejection hurdles, as well as an assessment that current orbital compute falls short of operational requirements.
Data Center Knowledge / Wylie Wong
Key Issues
01
Testing remains at prototype scale
Orbital compute has reached early hardware tests, but demonstrations and planned missions remain bounded experiments rather than operational data centers. Current tests target chip operation, thermal and radiation performance, and satellite links.
Drawn from 4 articles
02
Scale is gated by launch economics and engineering
Scaling requires transporting substantial power and heat-rejection infrastructure while managing launch, replacement, and maintenance costs. Available estimates place viable launch prices well below current costs, and radiator requirements rise sharply with compute capacity.
Drawn from 4 articles
03
Near-term fit favors latency-tolerant workloads
Constrained links and the locality needs of tightly coupled AI training make orbital systems a poor fit for hyperscale training at present. Onboard processing and workloads tolerant of latency or limited data transfer appear more feasible, though reported customer requirements for operational use remain unmet.
Drawn from 4 articles
04
Large constellations remain proposals awaiting authorization
Companies have proposed or filed plans for large orbital compute networks, but announced scales and authorization statuses differ and do not represent deployed capacity. FCC approval remains a practical step for some plans.
Drawn from 4 articles
Key Numbers
$2,500 to $3,000 per kilogram
current launch cost
orbital data centers
“Current launch costs of $2,500 to $3,000 per kilogram would need to fall to roughly $200 to $500 per kilogram for orbital data centers to make financial sense.”
Data Center Knowledge
40-kilowatt
AI rack power
“In a vacuum, heat can be removed only through radiation. A 700-watt Nvidia H100 GPU would require approximately 1.4 square meters of radiator surface at the cited operating temperature. A 40-kilowatt AI rack would therefore need roughly 80 square meters of radiator. Scaling the concept to a proposed 5-gigawatt facility would require approximately 10 square kilometers of radiators and 16 square kilometers of solar panels, before accounting for degradation, shielding, structural support, pumps, coolant, and other systems.”
PostQuantum.com
700-watt
GPU power
Nvidia H100 GPU
“In a vacuum, heat can be removed only through radiation. A 700-watt Nvidia H100 GPU would require approximately 1.4 square meters of radiator surface at the cited operating temperature. A 40-kilowatt AI rack would therefore need roughly 80 square meters of radiator. Scaling the concept to a proposed 5-gigawatt facility would require approximately 10 square kilometers of radiators and 16 square kilometers of solar panels, before accounting for degradation, shielding, structural support, pumps, coolant, and other systems.”
PostQuantum.com
hundreds of thousands chips
chips in large AI data centers being built on Earth
large AI data centers
“Large AI data centers being built on Earth can contain hundreds of thousands of chips and draw up to a gigawatt of power. Generating comparable power in orbit would require solar arrays kilometers across, and launching that much equipment would be a massive undertaking.”
Scientific American
thousands satellites
distributed architecture scale
proposed smaller-satellite architecture
“The proposed distributed architecture, involving thousands of smaller satellites, reduces some structural and pumping problems but does not eliminate the core scaling issues. Radiator and solar-panel area still scale with compute power. Thousands of satellites would increase collision-avoidance demands, replacement launches, debris exposure, and refresh costs.”
PostQuantum.com
Looking Back
155 Day Timeline
Articles published over time. Hover any bar for the period and its article count.
May 11
Jun 7
Jul 1
Jul 28
Aug 21
Sep 17
Oct 11
The Story So Far
No material change
No new-member articles were supplied, so there is no evidence of a material real-world change since the prior state.
Previously
Companies are testing satellite-based AI computing and seeking approval for much larger orbital networks, positioning near-continuous solar power as an alternative to terrestrial grid, land, and cooling constraints. Google’s Suncatcher and earlier Starcloud demonstrations mark early technical steps, while SpaceX, Cowboy Space, Blue Origin, and others have described or filed plans for constellations. Launch economics, heat rejection, communications, radiation exposure, and limited repairability remain major barriers; current evidence supports niche and experimental uses more clearly than a near-term replacement for terrestrial data centers.
History
10/10/2026
Environmental groups have added a new challenge to orbital-computing proposals, urging the FCC to assess cumulative pollution, debris, and light impacts.
All Articles59 articles
Important5 articles · CI Score 60 and above
Morgan Lewis / Jared Wilkerson, Robert Bolgar-Smith, and Heidi Rasmussen70
A July 2026 petition asks the Federal Communications Commission to review proposed orbital data center satellites under NEPA before considering related licenses.
Orbital Compute and Reflex Aerospace agreed in 2020s development work across Los Angeles and Berlin to build Orbital-1, a planned 100,000-satellite orbital data center constellation.
Google launched its first Suncatcher datacenter satellite into space to test computing hardware and infrastructure requirements for potential orbital datacenters.
In 2026, the FCC accepted four applications for orbital data center constellations in the United States, bringing licensing and environmental review questions into active consideration.
Blue Origin, SpaceX and Google advanced orbital data-center concepts in the 2020s, proposing or exploring AI facilities in Earth orbit amid concerns over debris and emissions.
Industry executives and a US Air Force technology leader assessed orbital data centers at the AI Infra Summit in Santa Clara, identifying thermal, launch, bandwidth, power, and adoption barriers.
SpaceX proposed an FCC-backed low Earth orbit AI-1 satellite data center constellation, but analysts cite launch scale and in-space GPU cooling constraints as the limiting factors.
SpaceX seeks FCC approval for an integrated Starlink orbital computing constellation while JLL projects orbital AI will complement, not replace, Earth data centers.
Environmental groups petitioned the FCC in the United States to pause orbital data center satellite licenses while requiring National Environmental Policy Act environmental review.
Earthjustice petitioned the Federal Communications Commission in July 2026 in the United States to require programmatic NEPA review of proposed million-satellite orbital data centers.
Google plans an upcoming SpaceX launch to test four AI processors running Gemini models in orbit as an initial step toward orbital data-center computing.
SpaceX is targeting a late-2027 orbital data-center launch, while industry observers expect hyperscale deployment during the 2030s after major advances in cooling, connectivity, power, and radiation protection.
Google Research described Project Suncatcher in 2025 as an orbital, solar-powered AI infrastructure concept amid data-center electricity growth projections.
SpaceX asked the U.S. Federal Communications Commission in January to approve an orbital data center system of up to one million satellites, prompting environmental groups to seek expanded review.
Google is developing Project Suncatcher for orbital AI computing, while Starcloud-1 and partners test space-based processors in low Earth orbit from 2025 through 2027.
SpaceX recently announced AI1 orbital data center satellites and a Texas factory plan as orbital GPU demonstrations highlighted an alternative to terrestrial infrastructure constraints.
Starcloud, Axiom Space, Blue Origin, and SpaceX are advancing concepts for solar-powered data centers in low-Earth orbit to reduce terrestrial infrastructure constraints.
Bain & Company analysis projects orbital data centers could scale in the early 2030s as launch economics and satellite efficiency improve, with SpaceX seeking FCC approval in 2026.
SpaceX's proposed million-satellite AI data center constellation could generate substantial annual metal losses in low-Earth orbit through GPU replacement and satellite disposal.
A technical assessment highlights launch economics, radiative cooling scaling, optical downlink limits, and orbital reliability constraints as barriers to large-scale orbital data centers for terrestrial AI demand.
Cowboy Space Corporation plans 20,000 space-based data centers, following a $275 million raise and an FCC-filed timeline targeting launches by late 2028.
SpaceX is expanding AI compute capacity through ground-based facilities and proposed solar-powered orbital data centers, with initial satellite launches planned for 2028.
Pentagon and intelligence officials said orbital data centers remain early as investors increase funding for space based AI GPU compute, at GEOINT Symposium on May 5.
NASA Administrator Jared Isaacman is promoting orbital AI data centers as Google, SpaceX, and Starcloud develop satellite computing systems in Earth orbit during the current data-center expansion.
SpaceX and Nvidia plan to launch an optimized Vera Rubin NVL72 datacenter rack into low Earth orbit next year, with larger deployments planned for 2028.
IEEE Spectrum assesses SpaceX’s proposed million-satellite orbital data-center approach, citing launch scaling and in-space cooling limits as major feasibility barriers.
SpaceX and other companies explore orbital AI compute satellites to reduce terrestrial land, power, and water constraints, but in-orbit power, heat rejection, and maintenance challenges remain limiting.
SpaceX filed an FCC application for an AI1 satellite constellation and discussed Terafab in Austin as an orbital AI data-center path amid rising terrestrial constraints.
SpaceX and Blue Origin pursue FCC and permitting steps for orbital data center concepts, while JLL and thermal-management experts emphasize cooling and launch-cost hurdles.
AI infrastructure developers and SpaceX are examining U.S., orbital, and lunar locations for future computing facilities as construction demand and launch-related opposition evolve.
8/28/2026 • Infrastructure & Site Development • General
SpaceX plans 2027 launches of AI1 orbital data center satellites as astronomers forecast increased interference with optical astronomy beginning in 2027.
SpaceX revealed AI1 late Monday, positioning an orbital compute satellite with deployable solar power and liquid radiator cooling ahead of a Friday market debut.
SpaceX is developing the StarMind orbital-computing system for possible deployment as early as 2027, with satellite manufacturing planned in Bastrop, Texas.
Nvidia said SpaceXAI is receiving early Vera CPU shipments and plans orbital Vera Rubin deployments from 2027 as SpaceX expands high-density AI capacity.
An assessment of orbital AI data centers links energy advantages to launch-cost and in-orbit maintenance limits as grid interconnection queues push off-grid generation.
Blue Origin submitted Federal Communications Commission plans for 51,600 low Earth orbit data center satellites under Project Sunrise as Bezos cited energy limits.
Starcloud raised $250 million in 2026 to expand its Woodinville, Washington, manufacturing facility and develop orbital data-center spacecraft for future launches.
Sophia Space plans orbital data-center tests next year using passive cooling and solar power, while launch cost targets and capacity bottlenecks shape competitiveness with terrestrial facilities.
Space scholars criticized Musk’s plan for million-satellite orbital AI compute as far more expensive than terrestrial data centers, based on Starlink cost estimates.
SpaceX and Google propose orbital AI data centers using solar-powered satellites, with sun-synchronous deployment timelines referenced as early as 2028.
SpaceX is planning a Gigasat Factory for AI1 orbital data-center satellites at its Bastrop County, Texas, campus before submitting a county development permit application.
8/7/2026 • Infrastructure & Site Development • General
Marsh and underwriters discussed insurance coverage for orbital AI data centers on June 18 in New York and London, with risk-modeling gaps limiting near-term financing.
Marsh and underwriters began preliminary discussions on insurance coverage for orbital AI data centers on June 18 in New York and London, addressing risk modeling for AI hardware.
Jeff Bezos said orbital data centers face cost-driven timing limits, citing higher energy impact and a 10x launch cost reduction target ahead of Blue Origin Project Sunrise.
Cowboy Space raised 275 million in a Series B led by Index Ventures on May 11 to build an orbital AI compute and power infrastructure with an early Galactic Brain node deployment.
Cowboy Space Corporation raised $275 million in a Series B to fund rocket development for orbital data centers, with a first launch targeted before 2028.
Marsh and Lloyd's of London insurers engaged with space startups in preliminary talks about orbital AI data center coverage, with risk-modeling uncertainty slowing broader underwriting and financing.
6/19/2026 • Project Finance & Incentives • General