# Space-based data centers

> Source: https://aiwiki.ai/wiki/space_based_data_centers
> Summary: Space-based data centers are data centers launched into Earth orbit (or, in some proposals, placed on the Moon), where satellites carrying AI accelerators draw power from solar arrays and reject heat by radiating it into space.
> Updated: 2026-11-01
> Fact-checked: 2026-09-25
> Categories: AI Compute, AI Hardware, AI Infrastructure, Data Centers
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Space-based data centers." aiwiki.ai, 1 Nov 2026. https://aiwiki.ai/wiki/space_based_data_centers
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

| Field | Value |
|---|---|
| Concept | [Data center](https://aiwiki.ai/wiki/data_center)s operating in Earth orbit or on the lunar surface |
| Also called | Orbital [data center](https://aiwiki.ai/wiki/data_center)s, space data centers, orbital compute |
| First [NVIDIA H100](https://aiwiki.ai/wiki/nvidia_h100) in orbit | Starcloud-1, launched November 2, 2025[1][2][68] |
| Largest filed constellation | [SpaceX Starmind](https://aiwiki.ai/wiki/spacex_starmind), up to 1,000,000 satellites (FCC application, January 2026)[11] |
| Other filed constellations | Orbital Compute (100,000), [Starcloud](https://aiwiki.ai/wiki/starcloud) (88,000), Blue Origin Project Sunrise (51,600), Cowboy Space Stampede (20,000)[55][7][17][51] |
| Preferred orbit | Sun-synchronous (dawn-dusk) low Earth orbit for near-continuous sunlight[9] |
| Key players | [Starcloud](https://aiwiki.ai/wiki/starcloud), SpaceX, [Google](https://aiwiki.ai/wiki/google), Blue Origin, Axiom Space, Cowboy Space, Orbital Compute, Aethero, ADA Space, Zhejiang Lab, Lonestar, Thales Alenia Space |
| Main obstacles | Heat rejection, launch cost, radiation, serviceability, orbital debris |

**Space-based data centers** are [data centers](https://aiwiki.ai/wiki/data_center) launched into Earth orbit (or, in some proposals, placed on the Moon), where satellites carrying [AI accelerators](https://aiwiki.ai/wiki/ai_accelerator) draw power from solar arrays and reject heat by radiating it into space. The idea moved from studies to flight hardware between 2024 and 2026: the startup [Starcloud](https://aiwiki.ai/wiki/starcloud) became the first to fly an [NVIDIA H100](https://aiwiki.ai/wiki/nvidia_h100), a data-center-class [GPU](https://aiwiki.ai/wiki/gpu), launching it into orbit in November 2025,[1][2][68] in May 2025 China launched the first 12 satellites shared by two overlapping computing-constellation programs, Zhejiang Lab's [Three-Body Computing Constellation](https://aiwiki.ai/wiki/three_body_computing_constellation) and ADA Space's 2,800-satellite Star-Compute plan,[92][40] and by June 2026 five U.S. companies had filed Federal Communications Commission applications: SpaceX for up to one million satellites, Orbital Compute for 100,000, Starcloud for 88,000, [Blue Origin](https://thespacewiki.com/wiki/blue_origin) for 51,600, and [Cowboy Space](https://aiwiki.ai/wiki/cowboy_space) for 20,000.[11][55][7][17][51] [Google](https://aiwiki.ai/wiki/google)'s [Project Suncatcher](https://aiwiki.ai/wiki/project_suncatcher) research program began its first in-orbit test on October 1, 2026, when a prototype satellite carrying Google TPUs, built with Planet, reached orbit on SpaceX's Transporter-18 rideshare; Google reported contact with the satellite the same day.[62][79][80] On October 30, 2026, Google said the prototype had returned its first AI response, a haiku about being in space written by Google's open [Gemma 4](https://aiwiki.ai/wiki/gemma_4) model running on board, which Google called "the first-ever AI response from a data center class TPU in orbit."[107] Proponents including [Elon Musk](https://aiwiki.ai/wiki/elon_musk), Jeff Bezos, and [NVIDIA](https://aiwiki.ai/wiki/nvidia) CEO [Jensen Huang](https://aiwiki.ai/wiki/jensen_huang) argue that near-constant orbital sunlight can escape the electricity, land, water, and permitting constraints closing in on terrestrial [AI infrastructure](https://aiwiki.ai/wiki/ai_infrastructure); critics, including OpenAI CEO [Sam Altman](https://aiwiki.ai/wiki/sam_altman) and several independent engineering analyses, counter that heat rejection, launch cost, radiation, and the impossibility of hardware maintenance make orbital compute uneconomical for at least the rest of the decade.[34][30][32]

## Rationale

The name covers two architectures with different purposes and maturity. A 2025 peer-reviewed *Nature Electronics* perspective calls systems that process sensor data at its source **orbital edge data centres**, while **orbital cloud data centres** are constellations of servers and broadband links intended for in-space or ground-outsourced computing. [48] The U.S. Government Accountability Office made a similar distinction in April 2026: individual power, cooling, and communications technologies are established, but operating them together at data-center scale remains unproven, and smaller systems processing data already generated in space are closer to maturity than large orbital clusters for training AI models. [49]

The case for computing in orbit rests on power. In a dawn-dusk sun-synchronous low Earth orbit, a satellite rides the day-night terminator and sees the Sun almost continuously; Google's Project Suncatcher research estimates that a solar panel in such an orbit can be up to 8 times more productive than the same panel on Earth and produce power nearly continuously, reducing the need for batteries.[9] Bezos made the same argument at Italian Tech Week in October 2025: "We're going to start building these giant gigawatt data centers in space. So, these giant training clusters, those will be better built in space, because we have solar power there, 24/7."[16]

This matters because terrestrial AI buildouts increasingly collide with grid limits. The International Energy Agency estimates global data-center electricity consumption reached about 485 TWh in 2025 and, in its central projection, roughly doubles to 950 TWh by 2030,[36] and interconnection queues, substation equipment, land, cooling water, and local opposition have all become gating items for new [gigawatt-scale campuses](https://aiwiki.ai/wiki/ai_energy_consumption). Starcloud CEO Philip Johnston framed the pitch in March 2026: "The AI revolution is colliding with the physical limits of our terrestrial energy grid... By moving AI compute to space, we unlock access to unlimited solar power and completely remove the energy bottleneck."[5] Former Google CEO Eric Schmidt gave the same rationale for taking control of the rocket company Relativity Space in 2025, citing projections that U.S. AI data centers could need tens of additional gigawatts by 2030; asked directly whether the acquisition was about launching data centers, he answered "Yes."[29]

The second pillar is the launch-cost curve. [SemiAnalysis](https://aiwiki.ai/wiki/semianalysis)'s June 2026 cost model uses $1,400 to $1,800 per kilogram for current [Falcon 9](https://thespacewiki.com/wiki/falcon_9) delivery, while SpaceX's published smallsat rideshare offer starts at $350,000 for 50 kg, with additional mass priced at $7,000 per kilogram.[32][58] Nearly every published feasibility case depends on fully reusable heavy lifters, chiefly SpaceX's [Starship](https://thespacewiki.com/wiki/starship), driving launch cost toward a few hundred dollars per kilogram or below.[32] Google's Suncatcher paper puts the threshold for rough parity with terrestrial data-center energy costs at less than $200/kg, a level it argues launch-price trends could reach by the mid-2030s.[9] Vendors also point to cooling: satellites reject heat to the vacuum by radiation alone, consuming no water and, per SpaceX, cutting cooling power overhead "by an order of magnitude" versus chillers and cooling towers.[12] Whether that trade is actually favorable is the central point of dispute with skeptics (see below).

## Technical challenges

The Space Wiki's [orbital data centers article](https://thespacewiki.com/wiki/orbital_data_centers) covers the same engineering limits from the spaceflight side: power, heat rejection, radiation and debris.

### Heat rejection

In vacuum there is no convection, so all waste heat must leave by radiation, governed by the Stefan-Boltzmann law. ABI Research analyst Andrew Cavalier calculated in IEEE Spectrum that a single 700 W H100 needs about 1.4 square meters of radiator surface to hold a 60 degree Celsius operating temperature (about 1 square meter if the chips are allowed to run at 85 degrees), so a 40 kW rack of 32 GPUs needs an 80-square-meter radiator, and a 100 MW facility needs at least 2,500 of them.[30] Radiator coatings also degrade under ultraviolet exposure and atomic oxygen; Cavalier estimates the required area grows about 40 percent over a five-year mission.[30] For scale, the [International Space Station](https://thespacewiki.com/wiki/international_space_station)'s external thermal control system rejects only about 70 kW, using radiator assemblies covering hundreds of square meters that SemiAnalysis prices at $340 million to $500 million.[32] Because a square meter of solar array in orbit yields around 400 W while a square meter of radiator sheds around 450 W, Cavalier notes that "every square meter of power generation now demands approximately another square meter of cooling."[30] Whether radiators sink the economics is contested: Forethought's May 2026 analysis finds radiator hardware is only 2 to 5 percent of total system cost across its median and conservative scenarios (assuming modern deployable radiators at 163 to 346 W/kg rather than ISS-era designs), but concedes that under worst-case radiator performance and $250/kg launch, cooling would add more than $8 billion per gigawatt, versus roughly $3 billion for terrestrial cooling.[33]

### Radiation

Commercial accelerators built on leading-edge process nodes were not designed for the space radiation environment, where energetic particles cause single-event upsets, latch-up, and cumulative total-ionizing-dose damage. Traditional radiation-hardened processors give up orders of magnitude in performance; a widely shared critique by a former NASA space-electronics engineer notes that hardened parts perform like a 20-year-old PowerPC.[31] The counter-evidence so far is encouraging for the boosters: Google put its Trillium ([TPU](https://aiwiki.ai/wiki/tpu) v6e) chips in a proton beam and found no hard failures attributable to total ionizing dose up to 15 krad(Si), with the high-bandwidth memory subsystem, the component most sensitive to total dose, showing irregularities only beyond a cumulative 2 krad(Si), nearly three times the 750 rad(Si) expected (behind shielding) over a five-year mission.[9] The peer-reviewed version of the paper, published in Joule in October 2026, gives the maximum tested dose as 10 krad(Si) rather than 15 krad(Si).[81] A revised version of the paper, posted in June 2026 after the team re-ran the radiation test with improved methods, identifies core logic and on-chip SRAM as the parts most sensitive to single-event effects, mainly as silent data corruption: about one event per 17 rad for typical transformer workloads, which the authors translate to roughly one failure per 3 million inferences at one inference per second (the first version had instead named HBM as the most sensitive component and, based on HBM memory errors, estimated about one failure per 10 million inferences). The authors call that rate likely acceptable for inference but say its effect on training needs further study.[50] In September 2026 Google said the proton-beam tests, run at UC Davis's Crocker Nuclear Laboratory, showed Trillium TPUs "can survive a radiation total ionizing dose greater than what they would receive during a five-year space mission," while noting that "some things can only be tested in space."[62] Starcloud-1's H100 has operated in orbit since November 2025, and TechCrunch described it as still operating in August 2026,[3][66] though a roughly 500 km orbit[82] below the inner Van Allen belt is a gentler environment than a multi-year mission at 600 km or above, and error rates at fleet scale remain undemonstrated. SemiAnalysis assumes orbital operators must carry about 20 percent spare GPU capacity, versus about 5 percent on the ground.[32]

### Bandwidth and latency

Training clusters need enormous internal bandwidth, and users need a fat, low-latency pipe to Earth. Inside a cluster, Google's answer is free-space optics: its bench demonstration reached 800 Gbps each way (1.6 Tbps aggregate) over a single transceiver pair, and its design flies satellites in tight formation, with next-nearest-neighbor satellites 100 to 200 meters apart within a cluster of about 1 km radius, to keep received optical power high; the published design models clusters of 81 satellites.[9][10] SpaceX's argument rests on its existing laser mesh: Musk wrote in October 2025 that "simply scaling up [Starlink](https://thespacewiki.com/wiki/starlink) V3 satellites, which have high speed laser links would work," with V3 designed for 1 Tbps-class throughput.[15] Skeptics point at the ground segment: a LEO satellite passes over a given ground station for only minutes at a time, and SemiAnalysis estimates traffic routed through intersatellite-link meshes accumulates 30 to 80 milliseconds of one-way delay, eroding any claimed latency advantage for interactive workloads.[32]

### Servicing, lifetime, and debris

There is no technician in orbit. A failed GPU, power supply, or pump stays failed, so operators must overprovision hardware and launch replacements; Forethought estimates that losing about 9 percent of compute per year with no repair forces roughly a 38 percent overbuy of launch and non-chip hardware across a data center's life.[33] Hardware also depreciates against a moving frontier: SemiAnalysis assumes a five-year useful life in orbit versus 15 years for a terrestrial facility shell, and orbital hardware cannot be upgraded rack by rack.[32] Finally, the filings themselves have raised debris alarms. NASA formally objected to Blue Origin's 51,600-satellite Project Sunrise application on April 30, 2026,[84] telling the FCC the constellation "presents significant safety and sustainability concerns" and requesting sustained technical engagement; DarkSky International objected over night-sky brightness, and the Center for Space Environmentalism called for a full environmental impact statement before approval.[18] Similar proceedings await SpaceX's million-satellite application, which asked the FCC to waive its standard six-year deployment milestones.[11] The FCC's Space Bureau wrote to all five orbital data center applicants (SpaceX, Starcloud, Blue Origin, Cowboy Space and Orbital Compute) on August 28, 2026, requesting more information, including orbital-debris details missing from their applications, before the filings could be processed further.[84] On September 30, the Commission adopted a Report and Order (FCC 26-64) finding that launching, deploying and operating space stations do not count as "major Federal action," the trigger for review under the National Environmental Policy Act.[85]

## Programs and players

| Organization | Project | Status (as reported, September-October 2026) | Hardware | Key dates |
|---|---|---|---|---|
| ESA / IBM Research / KP Labs | Space-based Data Centres study | Closed feasibility study; no flight mission or assigned TRL | Generic distributed edge-compute architecture and Excel/VBA simulator | Study ran 2022-2024; executive summary May 6, 2024[45][46][47] |
| [Starcloud](https://aiwiki.ai/wiki/starcloud) (US) | Starcloud-1, Starcloud-2; 88,000-satellite constellation filed | Starcloud-1 operating; two Starcloud-2 satellites planned for rideshare launches in 2027 | [NVIDIA H100](https://aiwiki.ai/wiki/nvidia_h100) (flown); Blackwell, AWS hardware, bitcoin ASICs (planned for Starcloud-2); NVIDIA Space-1 Vera Rubin Module (planned for future satellites) | Launch Nov 2, 2025; first LLM trained in orbit Dec 2025; $170M Series A Mar 30, 2026; $250M extension at $2.3B valuation Aug 21, 2026[2][4][5][66][67] |
| SpaceX (US) | [Starmind](https://aiwiki.ai/wiki/spacex_starmind) | FCC filing pending; mass production and deployment planned to start as soon as late 2027 | NVIDIA Rubin GPUs and Vera CPUs per Aug 2026 partnership; first-generation satellite to be based on an optimized Vera Rubin NVL72 system | FCC filing Jan 30, 2026 (up to 1M satellites); AI1 design Jun 8, 2026; NVIDIA deal Aug 4, 2026; NVL72 plan Aug 24, 2026[11][12][13][14][69] |
| [Google](https://aiwiki.ai/wiki/google) (US) | [Project Suncatcher](https://aiwiki.ai/wiki/project_suncatcher) | Research; first prototype (built with Planet) launched on Transporter-18 on Oct 1, 2026, with contact established; first AI response (a Gemma 4 haiku) reported Oct 30, 2026; two-satellite laser-link test planned for 2027 | Google TPUs (Trillium radiation-tested on the ground; four TPUs on the first prototype, per Ars Technica and Google) | Announced Nov 4, 2025; first test launch announced Sep 24, 2026; prototype launched Oct 1, 2026; first AI response reported Oct 30, 2026[8][10][62][63][64][79][80][107][108] |
| Blue Origin (US) | Project Sunrise (with TeraWave relay network) | FCC filing pending; NASA objection April 30, 2026[84] | Undisclosed | TeraWave (5,408 satellites) announced Jan 2026; Sunrise filed Mar 19, 2026 (51,600 satellites)[17][18] |
| Axiom Space (US) | [Orbital Data Center (ODC) nodes](https://aiwiki.ai/wiki/axiom_space_orbital_data_centers), called ARC 1 and ARC 2 since October 2026 | Two nodes, hosted as computing payloads on Kepler Communications relay satellites, in orbit; Axiom said in October 2026 that it operates both; AxODC Node ISS planned for 2027 | Computing payloads on Kepler relay satellites, which Kepler says carry multi-GPU compute modules; AxDCU-1 prototype on the ISS | Announced Apr 7, 2025; AxDCU-1 deployed on the ISS fall 2025; ODC Nodes 1 and 2 launched Jan 11, 2026; ARC update Oct 6, 2026[20][21][75][103][104][106] |
| Cowboy Space (US; formerly Aetherflux) | Stampede Data Center System plus space solar power | FCC application under review; Reason-1 power-beaming satellite launched Oct 1, 2026, with no beam-test result published as of Oct 16; no orbital data-center hardware launched | Proposed 1 MW data center integrated with a company-built rocket upper stage; Reason-2, planned for the first half of 2027, to carry H200 GPUs | Rebranded May 11, 2026; 20,000-satellite application filed May 11 and accepted for filing Jun 12, 2026; FCC request Aug 28, 2026, answered Sep 18, 2026; Reason-1 launched on Transporter-18 Oct 1, 2026[51][52][53][84][96][97][98][99] |
| Orbital Compute (US) | Orbital Datacenter System | FCC application pending; hosted-GPU Pathfinder planned for 2027 | High-performance GPU hosted payload, followed by purpose-built multi-GPU nodes | 100,000-satellite application filed Jun 24, 2026 and accepted for filing Jul 31, 2026; Reflex Aerospace chosen as Orbital-1 bus partner Sep 10, 2026; Orbital-1 planned for 2028[54][55][73][86] |
| Aethero (US) | Deimos and Phobos edge-compute missions | Two edge-compute spacecraft launched; not a terrestrial-scale data-center system | NVIDIA Jetson Orin modules | Deimos launched Aug 16, 2024; Phobos launched Mar 30, 2026[56][57] |
| EnduroSat / NVIDIA | Progressive AI-compute pre-integration across FRAME satellite buses; Aethero Titan mission | Announced integration plan; Titan manifested and scheduled, not launched | Jetson Orin, Jetson AGX Thor, IGX Thor, then Space-1 when available | Collaboration announced Aug 28, 2026; Titan launch given as October 2026 in April 2026 and as 2026 in the August 28 release[59][60] |
| Zhejiang Lab (China) | [Three-Body Computing Constellation](https://aiwiki.ai/wiki/three_body_computing_constellation) | First 12 satellites (shared with Star-Compute) in orbit; two more joined on Jul 24, 2026; targets of about 100 satellites by 2027 and about 1,000 by around 2032 | Zhejiang Lab onboard computers and AI models; up to 744 TOPS on a single satellite (Xinhua's top figure); 5 POPS and 30 TB across the first 12; 1,000 POPS planned | First 12 launched May 14, 2025 (built by ADA Space); two satellites launched Jul 24, 2026[40][91][92][95] |
| ADA Space (China) | Star-Compute ("Xing Suan") | Group 01 (the same 12 satellites) in orbit; second batch of eight launched Oct 14, 2026; third group scheduled for 2026 (per April 2026 statement) | Laser links up to 100 Gbps (first 12); Group 02 to test a 10 kW-scale liquid-cooling system and deployable radiators | Group 01 launched May 14, 2025; Group 02 launched Oct 14, 2026; 2,800 satellites planned (2,400 inference, 400 training), full network by 2035[40][70][87][92][93][94] |
| Lonestar Data Holdings (US) | Lunar and LEO data storage | Freedom payload flew to the Moon, where the lander tipped over; first StarVault storage payload in low Earth orbit, on Sidus Space's LizzieSat-4, slipped to no earlier than April 2027 | Solid-state storage payload on Intuitive Machines lander; StarVault payload built by Sidus Space | IM-2 launch Feb 26, 2025; Athena landed sideways Mar 6, 2025 and lost power; StarVault announced Apr 15, 2026; launch moved to Bandwagon-6 Sep 28, 2026[23][24][100][101][102] |
| Thales Alenia Space / EU | ASCEND feasibility study | Study concluded; follow-on European work proposed | Modular in-orbit-assembled infrastructure concept | Results published Jun 27, 2024; targets 1 GW in orbit before 2050[27] |
| NTT / SKY Perfect JSAT (Japan) | Space Compass, "space integrated computing network" | Optical relay and HAPS first; compute network is the long-term vision | GEO optical data relay | Joint venture established 2022[28] |
| Relativity Space (US) | Launch capacity aimed at orbital [data center](https://aiwiki.ai/wiki/data_center)s | Development | Terran R rocket | Eric Schmidt took control and confirmed the data-center rationale in 2025[29] |

### Starcloud

Starcloud, a Y Combinator-backed startup founded in 2024 in Redmond, Washington, flew the first NVIDIA H100 GPU to orbit. Starcloud-1, a 60 kg satellite built on an Astro Digital bus, launched on a SpaceX Falcon 9 on November 2, 2025 into a roughly 500 km orbit[82] carrying a single NVIDIA H100, which NVIDIA described as delivering 100 times more GPU compute than anything previously flown.[1][2][3] In December 2025 the company ran Google's open [Gemma](https://aiwiki.ai/wiki/gemma) model in orbit (which Starcloud describes as running "a version of [Gemini](https://aiwiki.ai/wiki/gemini) in space") and trained [Andrej Karpathy's](https://aiwiki.ai/wiki/andrej_karpathy) nanoGPT model, the first [LLM](https://aiwiki.ai/wiki/large_language_model) training run in space.[4][68] On March 30, 2026 Starcloud announced a $170 million Series A at a $1.1 billion valuation, led by Benchmark and EQT Ventures, bringing total funding to about $200 million; its FCC application for up to 88,000 satellites in 600-850 km sun-synchronous orbits was accepted for filing on March 13, 2026.[5][6][7] Starcloud-2, reported in December 2025 as targeting an October 2026 launch,[4] is a far larger spacecraft with roughly 100 times Starcloud-1's power generation, carrying an NVIDIA [Blackwell](https://aiwiki.ai/wiki/nvidia_blackwell) GPU, [AWS](https://aiwiki.ai/wiki/amazon_web_services) server hardware, bitcoin-mining ASICs, and what the company calls the largest deployable radiator flown on a commercial satellite; [Crusoe](https://aiwiki.ai/wiki/crusoe) is an announced early customer.[6][7] Its schedule slipped into 2027: in August 2026 TechCrunch reported that the company was focused on launching two 8 kW Starcloud-2 satellites on rideshare flights in 2027, and was considering buying a dedicated Falcon 9 launch.[66] In May 2026 Starcloud ordered optical terminals from SpaceX to use Starlink as a data-relay network for its future orbital data centers.[39] On August 21, 2026 Starcloud announced a $250 million extension of its Series A at a $2.3 billion post-money valuation, led by Manhattan West, with NVIDIA and Cisco Investments among the new investors, bringing total funding to $450 million; the company said future satellites would carry NVIDIA's Space-1 Vera Rubin Module and that it was building production lines for Starcloud-3 at a 100,000-square-foot facility in Woodinville, Washington.[66][67] Johnston has been explicit that the roadmap depends on cheap heavy lift: "we do need Starship, or a much lower launch cost vehicle."[6] Starcloud's long-term concept, described on NVIDIA's blog, is a 5 GW orbital data center with solar and cooling panels roughly 4 kilometers on a side.[1]

### SpaceX Starmind

SpaceX runs the most aggressive program by filed scale. It applied to the FCC on January 30, 2026 for an "Orbital Data Center System" of up to one million solar-powered satellites at 500-2,000 km, days before announcing its acquisition of [xAI](https://aiwiki.ai/wiki/xai) with orbital data centers as a stated rationale.[11] Musk unveiled the AI1 satellite design in June 2026 (as presented then, a roughly 70-meter-wingspan spacecraft with a 150 kW peak and 120 kW average compute payload, up to 110 square meters of deployable liquid radiators, and laser links into the Starlink constellation),[38] named the constellation Starmind, and on August 4, 2026 announced a partnership under which Starmind satellites will carry NVIDIA Rubin GPUs and Vera CPUs from the [Vera Rubin](https://aiwiki.ai/wiki/nvidia_vera_rubin) platform.[12][13] On the same day's earnings call Musk said SpaceX had decided to build "exclusively on Nvidia" because it considered Vera Rubin the best architecture.[14] On August 24, 2026 NVIDIA said SpaceXAI planned to base its first-generation Starmind satellite on an optimized Vera Rubin NVL72 rack-scale system, and Musk wrote that the space-optimized NVL72 was designed "for launch to orbit in Q4 next year, with significant scale in 2028."[69][71] As of September 25, 2026, the official Starmind page lists a larger AI1 than the June presentation: a 75-meter wingspan, a 30-meter deployed height, a compute payload of "up to 250 kW peak" and 175 kW average, and a 210 kW solar array.[12] SpaceX's official Starmind page says rapid production and deployment of thousands of AI satellites is planned to start as soon as late 2027.[12] Mass production is planned at an 11-million-square-foot "Gigasat Factory" in Bastrop, Texas,[42] and chips are eventually planned to come from the [TeraFab](https://aiwiki.ai/wiki/terafab) joint project with Tesla.[43] On September 16, 2026, SpaceX filed a 12-page response to FCC questions that included re-entry and collision risk, giving the satellites' mass as under 4,000 kg;[83] SatNews reported that the application remained under review.[72] All Starmind capacity and schedule figures are company plans; nothing has launched. The full program history is covered at [SpaceX Starmind](https://aiwiki.ai/wiki/spacex_starmind).

### Google Project Suncatcher

Google announced Project Suncatcher, "a research moonshot," on November 4, 2025, alongside a paper titled "Towards a future space-based, highly scalable AI infrastructure system design."[8][9] The paper first appeared as a Google-authored preprint;[50] a peer-reviewed version, titled "Toward a future space-based, highly scalable AI infrastructure system design," was published online by Joule on October 1, 2026.[81] The design flies TPU-equipped satellites in tight formations in a 650 km dawn-dusk sun-synchronous orbit, connected by free-space optical links, and treats the whole formation as one distributed machine-learning cluster. Unusually for the field, Google published its caveats up front: its Research blog post says the initial analysis shows the core concepts are "not precluded by fundamental physics or insurmountable economic barriers" but names thermal management, high-bandwidth ground communications, and on-orbit reliability as open engineering problems, and the paper conditions economics on launch prices below $200/kg by the mid-2030s.[9][50] A second [arXiv](https://aiwiki.ai/wiki/arxiv) version of the paper, posted June 17, 2026, re-ran the radiation test with improved methods.[50] In November 2025 Planet Labs announced that it would build and operate two prototype satellites carrying TPUs, targeted for launch by early 2027.[10] On September 24, 2026 Google said a first prototype satellite, developed with Planet, would fly on SpaceX's Transporter-18 rideshare "next week" to gather in-orbit data on how its TPUs handle launch stress, radiation, and thermal extremes, with a two-satellite test of the inter-satellite laser links to follow in 2027.[62] SpaceX's mission page listed the payload as "Project Suncatcher M1" and targeted launch from Vandenberg Space Force Base on October 1, 2026, with a backup opportunity on October 2.[63] Ars Technica reported that the satellite carries four TPUs, that Google chose to put its chips into a satellite Planet had already built rather than wait for the two custom 2027 satellites, that cooling limits the TPUs to runs of about 15 minutes, and that the satellite will operate for only a few months.[64] Space.com, citing The New York Times, gave a different figure, reporting that the prototype would operate in orbit for a year if all goes to plan.[78] Reuters reported Google as saying the mission is meant to gather data and identify failure points "rather than demonstrate an operational orbital data center."[65] The satellite launched on Transporter-18 from Vandenberg on October 1, 2026. Planet reported that it had made initial contact and begun commissioning, and Google said the satellite was performing as expected and that over the following weeks it would gather in-orbit data on how its TPUs hold up under the stresses of spaceflight, radiation and temperature extremes.[79][80] On October 30, 2026, Google said that after a few weeks of commissioning it had uplinked a prompt to the onboard Gemma 4 model during a ground-station pass and received a haiku in reply, the satellite's first AI response; it said its next tests would examine how well the TPUs tolerate thermal stress and the radiation of orbit.[107] Google's X account said the same day that the company had sent "four TPUs into orbit."[108] TechCrunch, reporting that day, wrote that the prototype generates up to 1 kW of electricity and that, because of how much power a TPU needs, the satellite can switch on just one of its four chips at once, for roughly 20 minutes per session.[109] The program is covered in detail at Project Suncatcher.

### Blue Origin

Bezos spent late 2025 talking up gigawatt orbital data centers on a 10-to-20-year horizon.[16] Blue Origin's concrete moves came in early 2026: in January it announced TeraWave, a 5,408-satellite laser-linked communications network aimed at data-center, enterprise, and government traffic, and on March 19 it filed the Project Sunrise FCC application for up to 51,600 compute satellites in 500-1,800 km sun-synchronous orbits, with TeraWave as the backbone.[17][74] In its April 30, 2026 objection, NASA told the FCC the application lacked an orbital debris mitigation plan and asked for "early and sustained technical engagement" with Blue Origin, according to SatNews.[18] Bezos himself has stayed conservative on timing, saying in May 2026 that the two-to-three-year timelines circulating in the industry are "probably a little ambitious" given energy, chip, and launch costs.[19]

### Axiom Space, Cowboy Space, Aethero, and Lonestar

Axiom Space's [orbital data center program](https://aiwiki.ai/wiki/axiom_space_orbital_data_centers) is building small, service-oriented nodes rather than gigawatt AI farms: processing capacity in orbit that other satellites can use for storage, AI/ML inference, and secure government workloads.[20][75] After deploying its AxDCU-1 prototype on the International Space Station in fall 2025, Axiom put its first two ODC nodes into orbit as two computing payloads it bought from Kepler Communications, hosted on Kepler's optical-relay satellites rather than flown as separate spacecraft.[75][106] Kepler's first tranche of ten relay satellites launched on January 11, 2026 on SpaceX's first "Twilight" rideshare, and Kepler and SatNews said each satellite carries multi-GPU compute modules with terabytes of storage.[21][105][76] Axiom specified 2.5 Gbps-capable optical links for the nodes and talks about scaling from kilowatts to megawatts.[20][75] In an October 6, 2026 release, Axiom called the two nodes ARC 1 and ARC 2, part of its Axiom Resilient Compute platform, and said it operates both and has confirmed ARC-to-ARC and ARC-to-ground communications. It had demonstrated post-quantum cryptography for the platform on the ground and named an on-orbit demonstration as its next milestone.[103] Its next announced node, AxODC Node ISS, is being developed with Spacebilt for launch to the ISS in 2027, with more than a petabyte of Phison solid-state storage and a Skyloom optical terminal providing up to 2.5 Gbps links to satellites in low Earth orbit.[104] This is closer to [edge computing](https://aiwiki.ai/wiki/edge_computing) for spacecraft than to a terrestrial data center replacement.

Cowboy Space, the company founded by Robinhood co-founder Baiju Bhatt as Aetherflux, rebranded on May 11, 2026 as it expanded from space solar power into rockets and orbital data centers.[22][52] Its Stampede Data Center System application, filed May 11 and accepted for filing by the FCC on June 12, requests authority for 20,000 satellites in 700 to 1,000 km dawn-dusk sun-synchronous orbits.[51] The company describes a vertically integrated design in which a rocket upper stage remains in orbit as a 1 MW data center; neither the rocket nor the data-center system has flown.[53] The FCC's Space Bureau wrote to Cowboy Space on August 28, 2026 with 11 numbered requests for information, including orbital-debris details missing from the application, "to render the Application suitable for further processing"; on September 11 it extended the response deadline to September 18.[84][96] In its September 18 response, the company said the satellite design was still in progress and had not been finalized, and again asked that any grant be conditioned on approval of a final orbital debris mitigation plan before the first Stampede launch. It said each satellite would have two solar arrays of roughly 57 m by 30 m, a maximum cross-sectional area of about 3,870 square meters and a deployed size of about 125 by 30 by 43 meters, and that reentries would be targeted. While still asking for flexibility across 700 to 1,000 km, it named shells at 700 to 745 km as its preferred configuration.[97] The company's first spacecraft, the power-beaming demonstrator Reason-1, flew on October 1, 2026, aboard the Falcon 9 carrying SpaceX's Transporter-18 rideshare.[98] Payload reported that the company hoped to attempt its first beam, an infrared laser aimed at a 10 to 20 meter spot at a ground test facility to deliver 30 to 100 W, as soon as late October; as of October 16, 2026, no result had been published.[99] Payload also reported that a second satellite, Reason-2, was slated to carry [NVIDIA H200](https://aiwiki.ai/wiki/nvidia_h200) GPUs in the first half of 2027 and that the first company-built rocket was targeted for December 2028; at the rebrand, Payload had reported that launch as no earlier than the end of 2028.[99][52]

Orbital Compute filed on June 24, 2026 for up to 100,000 data-center satellites in 500 to 850 km low Earth orbits, each proposed at roughly 100 kW, for a notional 10 GW system.[55] The FCC accepted the application for filing on July 31, 2026.[86] In September 2026 the company said it was raising its per-satellite power target to about 250 kW, enough for a full NVIDIA Vera Rubin rack, and had picked Germany's Reflex Aerospace as the exclusive bus partner for Orbital-1.[73] Its company roadmap calls for a single-GPU hosted Pathfinder mission in 2027 and a purpose-built, multi-GPU Orbital-1 satellite in 2028; both are plans rather than deployed hardware.[54]

Aethero has flown smaller edge-compute demonstrators rather than terrestrial-scale infrastructure. Its Deimos satellite launched August 16, 2024 with a Jetson Orin-based module rated by Aethero at 100 TOPS; Phobos launched March 30, 2026 with a Jetson Orin NX in "Super Mode" rated at up to 157 TOPS and hosted containerized applications under a compute-as-a-service model.[56][57] NVIDIA's March 2026 space-computing announcement separately named Axiom Space, Cowboy Space under its former Aetherflux name, Kepler, Planet, Sophia Space, and Starcloud as platform users.[35]

[Lonestar Data Holdings](https://aiwiki.ai/wiki/lonestar_data_holdings) pursues data storage rather than AI compute, and its first storage hardware went to the Moon. Its Freedom payload, a solid-state storage device marketed as the first lunar data center and aimed at resilience and disaster-recovery use cases, launched aboard Intuitive Machines' IM-2 Athena lander on February 26, 2025; Lonestar said the payload completed its commercial operational tests during transit and remained healthy after entering lunar orbit.[23][44] Athena touched down on March 6, 2025 near the lunar south pole but came to rest on its side inside a crater and depleted its batteries within a day, ending surface operations prematurely.[24] Its next flight is planned for low Earth orbit. In April 2026 it announced StarVault, which it describes as a commercial space-based "sovereign data storage" platform, and said Sidus Space was building the first StarVault payload for launch that October on Sidus's LizzieSat-4 satellite; in August, Space.com reported that further StarVault missions would follow as Lonestar built a data-storage network in low Earth orbit.[101][102] The flight then slipped. On September 28, 2026 Sidus said LizzieSat-4, carrying hosted payloads including Lonestar's, was still in production and would launch with LizzieSat-5 on SpaceX's Bandwagon-6 rideshare, targeted for no earlier than April 2027.[100]

### EnduroSat and NVIDIA

On August 28, 2026, EnduroSat announced that it would progressively pre-integrate NVIDIA Jetson Orin, Jetson AGX Thor, and IGX Thor, followed by the Space-1 Vera Rubin Module when available, across its FRAME satellite buses. The companies described complementary roles: NVIDIA supplies the AI compute and software, while EnduroSat supplies the spacecraft platforms and mission operations.[59][61] This was a forward-looking integration plan, not an announcement that EnduroSat was operating an orbital data center. The first Blackwell-based [Jetson Thor](https://aiwiki.ai/wiki/jetson_thor) deployment was described as manifested on Aethero's Titan mission, flying on a FRAME-15 spacecraft. Aethero and EnduroSat's April 2026 announcement gave an October 2026 launch date, while EnduroSat's August 28 release said only that Titan was scheduled to launch in 2026; as of September 28, 2026, neither company had announced a launch.[59][60] NVIDIA separately said Space-1 would become available at a later date, so its use on FRAME buses also remained prospective.[35]

### China

China has two overlapping orbital computing programs that share their first satellites. On May 14, 2025 a [Long March 2D](https://thespacewiki.com/wiki/long_march) launched 12 computing satellites built by Chengdu-based ADA Space (Guoxing Aerospace). Xinhua described the launch as the first for the Three-Body Computing Constellation, which Zhejiang Lab leads, and also the first for ADA Space's own "Star-Compute" plan; Zhejiang Lab developed the onboard computers and space-based AI models, and ADA Space developed the satellite platform.[92][25] Xinhua gave 744 trillion operations per second (TOPS) as the highest computing figure for any single satellite in the batch, and a combined 5 POPS and 30 TB of storage for the 12; Computerworld and Live Science reported the 744 TOPS figure as each satellite's capacity, and Computerworld reported laser links of up to 100 Gbps.[92][70][41] The two programs have different leads and targets. Zhejiang Lab plans to grow the Three-Body constellation to about 100 satellites by 2027 and to a thousand-satellite scale by around 2032, with a combined capacity of 1,000 POPS once complete.[40][91][92][95] ADA Space's Star-Compute plan calls for 2,800 satellites (2,400 for inference, 400 for training), a thousand-satellite network in commercial operation by 2030, and the full network by 2035.[40] SpaceNews and Live Science gave the 2,800 figure as the size of the joint constellation at the first launch, while NASASpaceFlight later reported the two targets separately: up to 1,000 satellites for Zhejiang Lab and 2,800 for Star-Compute.[25][41][26] ADA Space executive vice president Zhao Hongjie said in April 2026 that Star-Compute's second and third satellite groups were in production and scheduled for orbital deployment in 2026.[40] In August 2026 NASASpaceFlight reported that ADA Space had maintained a continuous eight-day laser link between two of the 12 satellites, up to 1,000 km apart, and that Chenguang-1, the first satellite of a venture called Orbital Chenguang intended to demonstrate technologies for gigawatt-class orbital data centers, had launched from Jiuquan on a Kinetica-1 rideshare on July 24.[26] That Kinetica-1 (Lijian-1) launch also carried two satellites that Science and Technology Daily reported as joining the Three-Body constellation: the Jitianxing A-04 computing satellite and the Xiguang-2 03 remote-sensing satellite.[91] A Kinetica-1 rocket launched nine Chinese satellites on September 20, 2026, including two with on-orbit AI computing, according to China's Science and Technology Daily and CGTN: Chaozhisuan-1, a high-resolution remote-sensing satellite from Chaozhisuan (Beijing) Technology that carries an AI computer for in-orbit image processing and target recognition, and PEGA-SUS1, a roughly 150 kg demonstrator developed by GalaxySpace, Pengcheng Laboratory and the Southern University of Science and Technology that combines 5G non-terrestrial network communications, laser communications and an on-orbit computing payload.[89][90][77] Star-Compute's second batch, eight ADA Space satellites, launched on a [LandSpace](https://thespacewiki.com/wiki/landspace) Zhuque-2E rocket on October 14, 2026, together with two laser-communication test satellites from Lanta Intelligence.[87][88][93] CGTN, China's state-run international broadcaster, reported that all ten satellites reached orbit and that the eight would test a 10-kilowatt-scale liquid-cooling system, large deployable radiative-cooling wings, a laser communication network and foundation-model training in orbit.[87] The launch reports from CGTN and Science and Technology Daily did not describe the eight as part of the Three-Body constellation, and Cailian Press mentioned that constellation only as a separate Zhejiang Lab project, with its own target of a thousand-satellite scale and 1,000 POPS by 2032.[87][93][94]

### Europe and Japan

Europe's activity has been studies rather than filings.

A separate ESA Discovery and Preparation study ran from 2022 to 2024, with IBM Research GmbH as prime contractor and KP Labs as subcontractor. ESA's project record lists it as closed and assigns no initial, target, or achieved technology-readiness level. [46] Unlike gigawatt orbital AI proposals, the study examined distributed edge computing for data that spacecraft already generate: a sensor satellite sending observations to a compute satellite in the same orbit, a low-Earth-orbit observer using a geostationary compute and relay node, and a lunar lander processing rover data before returning selected findings through an orbital relay. These were simulated scenarios, not approved missions. [45]

The team derived a generic architecture from use cases and stakeholder input, then implemented an adjustable Microsoft Excel and Visual Basic for Applications simulator combining technology roadmaps, user requirements, and scenario constraints. Its outputs included trends in cost per compute, mass per compute, and cost per power. The executive summary says the model showed, under its assumptions, an inflection point after which compressing data and extracting actionable information in orbit could be more economical than returning all raw data. It gives no universal crossover date. ESA's public story described a decade or two as a possible horizon, an attributed technology forecast rather than a deployment commitment. [45][47]

The ASCEND feasibility study (Advanced Space Cloud for European Net zero emission and Data sovereignty), coordinated by Thales Alenia Space for the European Commission under Horizon Europe, published results on June 27, 2024. It concluded that space data centers could reduce digital-sector emissions only if a launcher ten times less emissive over its lifecycle is developed, proposed modular infrastructure assembled in orbit by robotic systems, and sketched a target of one gigawatt of European capacity in orbit before 2050.[27] In Japan, NTT and SKY Perfect JSAT founded Space Compass in 2022 around a "space integrated computing network" concept that would process satellite data in orbit; its near-term business is geostationary optical data relay and high-altitude platform connectivity, and as of August 2026 the venture's public announcements center on relay services rather than an operating orbital compute satellite.[28]

## Economics

Every serious analysis reduces the question to dollars per kilogram and hardware lifetime. The numbers below come from named models and should be read as estimates with strong assumptions.

| Analysis | Date | Central finding |
|---|---|---|
| Google Suncatcher paper[9] | Nov 2025 | Rough parity with terrestrial data-center energy costs requires launch below ~$200/kg, plausible by the mid-2030s |
| Forethought (Parrack and Moorhouse)[33] | May 2026 | Space solar matches off-grid terrestrial power near $250/kg; full-system parity near $100/kg; possibly 3-5 years away if Starship stays on track |
| [SemiAnalysis](https://aiwiki.ai/wiki/semianalysis)[32] | Jun 2026 | 2026 levelized cost of compute ~$10.91/hr/GPU in space vs $2.49 terrestrial (>4x); base-case parity around 2040; early-2030s builds carry ~30% premium |
| Deutsche Bank ([SpaceX Starmind](https://aiwiki.ai/wiki/spacex_starmind) model)[37] | 2026 | Non-compute (infrastructure) cost premium ~6x today, falling to ~1.2x by 2029 with AI1 deployment and below parity (~0.6x) by 2032 |
| ABI Research (in IEEE Spectrum)[30] | Jun 2026 | Launching and running a GPU in space for a year costs at least an order of magnitude more than on the ground |

The disagreements trace to a few parameters. Launch: SemiAnalysis compares the roughly $250/kg future Starship cost envisioned by SpaceX with its $1,400 to $1,800/kg Falcon 9 assumption, and launch is the single most sensitive input in every model.[32] Lifetime and maintenance: a 5-year orbital life versus 15 years terrestrial, plus spare-capacity overbuy of around 20 percent, drives SemiAnalysis's estimate that monthly non-chip data center capex runs about 18 times higher in space.[32] Radiators: Forethought's scenarios make cooling hardware a minor cost line (about 5 percent in its conservative case and 2 percent in its median case), while worst-case radiator assumptions at $250/kg add over $8 billion per gigawatt.[33] Vendor claims sit well outside these ranges: Starcloud projects 10x lower energy costs and 10x lifecycle CO2 savings versus terrestrial facilities,[1] and the ASCEND study projected returns of several billion euros by 2050,[27] but both are advocacy positions from parties selling the concept.

A separate line of argument, made by SemiAnalysis, is that even a launch-cost miracle would not remove the binding constraint on AI buildout, which it identifies as semiconductor supply: in its projections AI demand consumes about 86 percent of TSMC's N3-class capacity and 70 percent of DRAM wafer capacity by 2027, so orbit adds power capacity but not chips.[32]

## Criticism

The concept has drawn unusually pointed criticism for something attracting billions in filings and funding. Altman said in February 2026: "I honestly think the idea with the current landscape of putting data centers in space is ridiculous," citing launch costs relative to terrestrial power costs and the impossibility of fixing broken GPUs, and adding that orbital data centers are "not something that is going to matter at scale this decade."[34] Musk's rebuttal to thermal critiques has been that it is "safe to say SpaceX knows how to do heat rejection in space," pointing at the more than 10,000 Starlink satellites it operates,[38] though a Starlink satellite dissipates orders of magnitude less heat than a single AI rack.

The engineering critiques cluster around thermodynamics. Cavalier's IEEE Spectrum analysis argues the industry's "free cooling" framing inverts reality, because radiative-only heat rejection is the binding constraint: "the constraint is never the silicon. It's the thermodynamics."[30] A widely circulated November 2025 essay on the Taranis blog, written by a self-described former NASA space-electronics engineer who later spent a decade at Google, works the same numbers from the ISS baseline (a station-sized 2,500-square-meter solar array yields power for only a couple hundred GPUs) and calls the concept a "catastrophically bad idea," while conceding it is "just about possible if you really want to do it."[31] SemiAnalysis adds that several standard talking points are simply wrong as stated: generic LEO satellites see sunlight only about 60 percent of the time (dawn-dusk sun-synchronous orbits fix this but are a narrow, crowded orbital slot), and even those orbits can suffer up to about 35 minutes of eclipse per day.[32]

Institutional objections center on the space environment. NASA's April 30, 2026 filing against Project Sunrise, and parallel objections from DarkSky International and the Center for Space Environmentalism, mark the start of a regulatory fight over whether tens of thousands to a million additional large satellites can be operated safely, with collision-cascade risk and interference with ground-based astronomy the recurring themes; SpaceX's application meanwhile asks the FCC to waive the standard milestone that would require half the constellation to be in orbit within six years.[18][11] Even sympathetic observers note the gap between filings and demonstrated hardware: prominent deployments as of August 2026 consist of single-satellite or small-node demonstrations, including Starcloud-1's one H100, Axiom's two nodes hosted on Kepler relay satellites and its ISS prototype, Aethero's Deimos and Phobos edge computers, the first 12 Chinese computing satellites, and Lonestar's lunar storage payload on a tipped-over lander.[1][20][21][25][44][56][57] Bezos, whose company filed one of the largest applications, put the sober version on the record himself: timelines of two to three years are "probably a little ambitious."[19]

## See also

- [SpaceX Starmind](https://aiwiki.ai/wiki/spacex_starmind)
- [Project Suncatcher](https://aiwiki.ai/wiki/project_suncatcher)
- [Starcloud](https://aiwiki.ai/wiki/starcloud)
- [Data center](https://aiwiki.ai/wiki/data_center)
- [AI infrastructure](https://aiwiki.ai/wiki/ai_infrastructure)
- [AI energy consumption](https://aiwiki.ai/wiki/ai_energy_consumption)
- [NVIDIA Vera Rubin](https://aiwiki.ai/wiki/nvidia_vera_rubin)
- [Edge computing](https://aiwiki.ai/wiki/edge_computing)
- [TeraFab](https://aiwiki.ai/wiki/terafab)

## References

1. NVIDIA Blog. "How Starcloud Is Bringing Data Centers to Outer Space." October 15, 2025. https://blogs.nvidia.com/blog/starcloud/
2. Gunter's Space Page. "Starcloud 1 (Lumen 1)." Accessed August 12, 2026. https://space.skyrocket.de/doc_sdat/starcloud-1.htm
3. Data Center Frontier. "Starcloud Launches Orbital AI Data Center With NVIDIA H100 GPU." November 2025. https://www.datacenterfrontier.com/site-selection/article/55337494/starcloud-launches-orbital-ai-data-center-with-nvidia-h100-gpu
4. CNBC. "Nvidia-backed Starcloud trains first AI model in space." December 10, 2025. https://www.cnbc.com/2025/12/10/nvidia-backed-starcloud-trains-first-ai-model-in-space-orbital-data-centers.html
5. Via Satellite. "Starcloud Raises $170M to Fund Orbital Data Center Plans." March 30, 2026. https://www.satellitetoday.com/finance/2026/03/30/starcloud-raises-170m-to-fund-orbital-data-center-plans/
6. Payload. "Starcloud Raises $170M Series A at $1.1B Valuation." March 30, 2026. https://payloadspace.com/starcloud-raises-170m-series-a-at-1-1b-valuation/
7. SpaceNews. "Starcloud files plans for 88,000-satellite constellation." March 2026. https://spacenews.com/starcloud-files-plans-for-88000-satellite-constellation/
8. Google. "Meet Project Suncatcher, a research moonshot to scale machine learning compute in space." The Keyword, November 4, 2025. https://blog.google/innovation-and-ai/technology/research/google-project-suncatcher/
9. Google Research. "Exploring a space-based, scalable AI infrastructure system design." November 4, 2025. https://research.google/blog/exploring-a-space-based-scalable-ai-infrastructure-system-design/
10. Planet Labs PBC. "Planet to Build and Operate Advanced Space Platform for Google's Project Suncatcher Moonshot." November 2025. https://www.planet.com/pulse/planet-to-build-and-operate-advanced-space-platform-for-project-suncatcher-moonshot/
11. SpaceNews. "SpaceX files plans for million-satellite orbital data center constellation." January 2026. https://spacenews.com/spacex-files-plans-for-million-satellite-orbital-data-center-constellation/
12. SpaceX. "Starmind." Official SpaceX website, accessed October 16, 2026. https://www.spacex.com/spacexsi/starmind
13. SpaceX (@SpaceX) on X. "SpaceX is partnering with @Nvidia to design the Starmind AI1 satellite compute payload." August 4, 2026. https://x.com/SpaceX/status/2084723854534951218
14. Axios. "SpaceX tops revenue expectations in first earnings report after IPO." August 4, 2026. https://www.axios.com/2026/08/04/spacex-earnings-elon-musk
15. Teslarati. "Starlink V3 satellites could enable SpaceX's orbital computing plans: Musk." October 31, 2025. https://www.teslarati.com/starlink-v3-satellites-enable-spacex-orbital-computing-plans-musk/
16. Alan Boyle. "Jeff Bezos says orbital data centers will be the 'next step' for space ventures that make Earth better." GeekWire, October 6, 2025. https://www.geekwire.com/2025/jeff-bezos-orbital-data-centers-next-step/
17. GeekWire. "51,600 more satellites? Blue Origin adds another twist to the data center space race with Project Sunrise." March 2026. https://www.geekwire.com/2026/blue-origin-data-center-space-race-project-sunrise/
18. SatNews. "NASA Objects to Blue Origin's 'Project Sunrise'." May 5, 2026. https://satnews.com/2026/05/05/nasa-objects-to-blue-origins-project-sunrise/
19. CNBC. "Bezos says 2-3 year timeline for space data centers is a 'little ambitious'." May 20, 2026. https://www.cnbc.com/2026/05/20/bezos-blue-origin-space-data-centers.html
20. Axiom Space. "Axiom Space to Launch Orbital Data Center Nodes to Support National Security, Commercial, International Customers." April 7, 2025. https://www.axiomspace.com/release/axiom-space-to-launch-orbital-data-center-nodes-to-support-national-security-commercial-international-customers
21. Kepler Communications. "Kepler Successfully Launches First Tranche of Optical Relay Satellites." January 11, 2026. https://kepler.space/kepler-successfully-launches-first-tranche-of-optical-relay-satellites/
22. SpaceNews. "Space-based solar power startup Aetherflux enters orbital data center race." December 2025. https://spacenews.com/space-based-solar-power-startup-aetherflux-enters-orbital-data-center-race/ (see also Data Center Dynamics, "Aetherflux orbital data center to be operational by Q1 2027," December 2025, https://www.datacenterdynamics.com/en/news/aetherflux-orbital-data-center-to-be-operational-by-q1-2027/)
23. PR Newswire. "Lonestar's Data Center is ready for the Moon: Intuitive Machines and Lonestar Complete Final Installation and Testing of Lonestar's Data Center Payload." February 10, 2025. https://www.prnewswire.com/news-releases/lonestars-data-center-is-ready-for-the-moon-intuitive-machines-and-lonestar-complete-final-installation-and-testing-of-lonestars-data-center-payload-302372021.html
24. Spaceflight Now. "Intuitive Machines' IM-2 Moon mission ends with lander on its side." March 7, 2025. https://spaceflightnow.com/2025/03/07/intuitive-machines-im-2-mission-ends-with-lander-on-its-side-on-the-moon/
25. Andrew Jones. "China launches first of 2,800 satellites for AI space computing constellation." SpaceNews, May 2025. https://spacenews.com/china-launches-first-of-2800-satellites-for-ai-space-computing-constellation/
26. NASASpaceFlight.com. "Amid failure, Wenchang celebrates milestone as China launches AI satellites and prepares for lunar south pole mission." August 12, 2026. https://www.nasaspaceflight.com/2026/08/china-roundup-20260813/
27. Thales Alenia Space. "Thales Alenia Space reveals results of ASCEND feasibility study on space data centers." June 27, 2024. https://www.thalesaleniaspace.com/en/press-releases/thales-alenia-space-reveals-results-ascend-feasibility-study-space-data-centers-0
28. NTT. "NTT and SKY Perfect JSAT Agree to Establish Space Compass Corporation." April 26, 2022. https://group.ntt/en/newsrelease/2022/04/26/220426a.html
29. TechSpot. "Former Google CEO Eric Schmidt wants to put data centers in space." May 2025. https://www.techspot.com/news/107801-former-google-ceo-eric-schmidt-wants-put-data.html
30. Andrew Cavalier. "Why Orbital Data Centers Are Harder Than Silicon Valley Thinks." IEEE Spectrum, June 11, 2026. https://spectrum.ieee.org/orbital-data-centers-heat
31. Taranis. "Datacenters in space are a terrible, horrible, no good idea." taranis.ie, November 26, 2025. https://taranis.ie/datacenters-in-space-are-a-terrible-horrible-no-good-idea/
32. Daniel Nishball, Pranav Myana, Ellie Holbrook, et al. "To Boldly Go: The Case for Space Datacenters." SemiAnalysis, June 3, 2026. https://newsletter.semianalysis.com/p/to-boldly-go-the-case-for-space-datacenters
33. Avi Parrack and Fin Moorhouse. "Will We Really Put Data Centers in Space?" Forethought, May 22, 2026. https://www.forethought.org/research/will-we-really-put-data-centers-in-space
34. Tom's Hardware. "Sam Altman fires back at Elon Musk's proposal for space-based data centers, says orbiting data centers 'ridiculous' for now." February 2026. https://www.tomshardware.com/tech-industry/artificial-intelligence/sam-altman-fires-back-at-elon-musks-proposal-for-space-based-data-centers-says-orbiting-data-centers-ridiculous-for-now-cites-high-failure-rates-and-cost-as-primary-limiters
35. NVIDIA Newsroom. "NVIDIA Launches Space Computing, Rocketing AI Into Orbit." March 16, 2026. https://nvidianews.nvidia.com/news/space-computing
36. International Energy Agency. "Key Questions on Energy and AI." 2026. https://www.iea.org/reports/key-questions-on-energy-and-ai
37. Investing.com. "Deutsche Bank sees SpaceX orbital data centers matching ground costs." 2026. https://www.investing.com/news/stock-market-news/deutsche-bank-sees-spacex-orbital-data-centers-matching-ground-costs-93CH-4790612
38. Luke James. "Elon Musk's first-gen orbital data center craft spans wider than a Boeing 747 and runs an interchangeable chip payload." Tom's Hardware, June 9, 2026. https://www.tomshardware.com/tech-industry/spacex-details-its-ai1-compute-satellite
39. Jason Rainbow. "Starcloud orders Starlink lasers for orbital data center network." SpaceNews, May 26, 2026. https://spacenews.com/starcloud-orders-starlink-lasers-for-orbital-data-center-network/
40. State Council Information Office of China. "From ground to orbit: China eyes computing in space." April 27, 2026. http://english.scio.gov.cn/chinavoices/2026-04/27/content_118464689.html
41. Ben Turner. "China is building a constellation of AI supercomputers in space, and just launched the first pieces." Live Science, May 23, 2025. https://www.livescience.com/technology/computing/china-is-building-a-constellation-of-ai-supercomputers-in-space-and-just-launched-the-first-pieces
42. Tom's Hardware. "SpaceX unveils 11-million-square-foot Gigasat factory, a new manufacturing facility for space-based data centers." June 9, 2026. https://www.tomshardware.com/tech-industry/big-tech/spacex-unveils-11-million-square-foot-gigasat-factory-a-new-manufacturing-facility-for-space-based-data-centers-aims-for-1-gw-year-of-space-ai-compute-by-late-2027-from-its-satellites
43. Luke James. "Elon Musk unveils $20 billion 'TeraFab' chip project to make chips, memory, and package processors all under one roof." Tom's Hardware, March 22, 2026. https://www.tomshardware.com/tech-industry/elon-musk-formally-launches-20-billion-terafab-chip-project
44. Lonestar Data Holdings. "Lunar Data Center Achieves First Success En Route To The Moon." PR Newswire, March 5, 2025. https://www.prnewswire.com/news-releases/lunar-data-center-achieves-first-success-en-route-to-the-moon-302392544.html

45. European Space Agency. "Knowledge beyond our planet: space-based data centres." August 5, 2024. https://www.esa.int/Enabling_Support/Preparing_for_the_Future/Discovery_and_Preparation/Knowledge_beyond_our_planet_space-based_data_centres

46. European Space Agency Nebula Public Library. "Space based data centres." Contract 4000139754, closed 2022-2024 study record. https://nebula.esa.int/4000139754

47. European Space Agency, IBM Research Europe, and KP Labs. "Space-Based Data Centres: Study Executive Summary." May 6, 2024. https://nebula.esa.int/sites/default/files/neb_tec_study/2473/public/C4000139754_ExS.pdf

48. Ablimit Aili, Jihwan Choi, Yew Soon Ong, and Yonggang Wen. "The development of carbon-neutral data centres in space." Nature Electronics 8, 1016-1026, 2025. https://doi.org/10.1038/s41928-025-01476-1

49. U.S. Government Accountability Office. "Science & Tech Spotlight: Data Centers in Space." GAO-26-109012, April 28, 2026. https://www.gao.gov/products/gao-26-109012

50. Blaise Aguera y Arcas et al. "Towards a future space-based, highly scalable AI infrastructure system design." arXiv:2511.19468 (v1 November 22, 2025; v2 June 17, 2026). https://arxiv.org/abs/2511.19468

51. Federal Communications Commission. "Applications Accepted for Filing: Cowboy Space Corp., SAT-LOA-20260323-00135." Public Notice SAT-02007, June 12, 2026. https://docs.fcc.gov/public/attachments/DOC-422332A1.pdf

52. Jacqueline Feldscher. "Aetherflux Rebrands, Pivots Business and Raises $275M." Payload, May 11, 2026. https://payloadspace.com/aetherflux-rebrands-pivots-business-and-raises-275m/

53. Cowboy Space Corporation. "Orbital Data Centers for AI." Official website, accessed August 17, 2026. https://www.cowboyspace.com/

54. Orbital Compute. "Data Centers in Space." Official website and roadmap, accessed August 17, 2026. https://orbital.inc/

55. Jason Rainbow. "Orbital files plans for 100,000 orbital data centers." SpaceNews, June 30, 2026. https://spacenews.com/orbital-files-plans-for-100000-orbital-data-centers/

56. Aethero. "Deimos Satellite Launch." August 16, 2024. https://www.aethero.com/news/press-release-deimos-satellite-launch/

57. Aethero. "Phobos Satellite Launch." March 30, 2026. https://www.aethero.com/news/press-release-phobos-satellite-launch/

58. SpaceX. "Smallsat Rideshare Program." Official website, accessed August 17, 2026. https://www.spacex.com/rideshare/

59. EnduroSat. "EnduroSat Makes Orbital AI Compute a Standard Satellite Capability with NVIDIA." August 28, 2026. https://www.endurosat.com/news/endurosat-makes-data-center-class-ai-compute-standard-on-serial-production-satellites-with-nvidia/

60. EnduroSat. "Aethero Deploys Next-Generation In-Orbit Computing on EnduroSat FRAME-15." April 16, 2026. https://www.endurosat.com/news/aethero-deploys-next-generation-in-orbit-computing-on-endurosat-frame-15/

61. NVIDIA. "Space Computing." Accessed September 3, 2026. https://www.nvidia.com/en-us/edge-computing/space-computing/

62. Travis Beals. "Behind Project Suncatcher, our moonshot to put AI in space." The Keyword (Google), September 24, 2026. https://blog.google/innovation-and-ai/models-and-research/google-research/google-project-suncatcher-facts/

63. SpaceX. "Transporter-18 Mission." Official SpaceX website, accessed September 28, 2026. https://www.spacex.com/launches/transporter18

64. Ryan Whitwam. "Google's first Suncatcher orbital data center test launches October 1." Ars Technica, September 24, 2026. https://arstechnica.com/google/2026/09/googles-first-suncatcher-orbital-data-center-test-launches-october-1/

65. Akash Sriram. "Google plans first test of AI chips in space under Project Suncatcher." Reuters, September 24, 2026 (via StreetInsider). https://www.streetinsider.com/Reuters/Google+plans+first+test+of+AI+chips+in+space+under+Project+Suncatcher/27100094.html

66. Tim Fernholz. "Starcloud raises $250 million for orbital data centers as launch options dry up." TechCrunch, August 21, 2026. https://techcrunch.com/2026/08/21/starcloud-raises-200-million-for-orbital-data-centers-as-launch-options-dry-up/

67. Alan Boyle. "Starcloud raises $250M to support the creation of data center satellite network in league with Nvidia." GeekWire, August 21, 2026. https://www.geekwire.com/2026/starcloud-250m-data-center-satellite-network-nvidia/

68. Starcloud. "Starcloud-1." Official website, accessed September 25, 2026. https://www.starcloud.com/starcloud-1

69. NVIDIA. "SpaceXAI Adopts NVIDIA Vera CPU to Accelerate Agentic AI at Massive Scale." Press release, August 24, 2026. https://investor.nvidia.com/news/press-release-details/2026/SpaceXAI-Adopts-NVIDIA-Vera-CPU-to-Accelerate-Agentic-AI-at-Massive-Scale/default.aspx

70. Gyana Swain. "China takes edge computing to orbit with first space-based processing network." Computerworld, May 20, 2025. https://www.computerworld.com/article/3990472/china-takes-edge-computing-to-orbit-with-first-space-based-processing-network.html

71. Elon Musk (@elonmusk) on X. "SpaceX, in partnership with Nvidia, has designed a space-optimized Vera Rubin NVL72 system for launch to orbit in Q4 next year, with significant scale in 2028." August 24, 2026. https://x.com/elonmusk/status/2091939113008238838

72. SatNews. "SpaceX Submits FCC Re-Entry Debris and Collision Risk Analysis for Starmind Orbital Data Center Constellation." September 19, 2026. https://satnews.com/2026/09/19/spacex-submits-fcc-re-entry-debris-and-collision-risk-analysis-for-starmind-orbital-data-center-constellation/

73. Payload. "Orbital Partners with Reflex Aerospace on ODC Platform." September 10, 2026. https://payloadspace.com/orbital-partners-with-reflex-aerospace-on-odc-platform/

74. The Register. "Jeff Bezos' rocket company Blue Origin applies to launch 51,000 datacenter satellites." March 20, 2026. https://www.theregister.com/2026/03/20/blue_origin_project_sunrise_orbital_datacenter/

75. Axiom Space. "Orbital Data Centers." Official website, accessed September 25, 2026. https://www.axiomspace.com/orbital-data-center

76. SatNews. "Kepler Communications' next-generation optical data relay constellation launched." February 9, 2026. https://satnews.com/2026/02/09/63641/

77. China in Space. "New Orbital Compute Demonstrating Spacecraft Arrive in Space [Kinetica-1 Y18]." September 20, 2026. https://www.china-in-space.com/p/new-orbital-compute-demonstrating

78. Mike Wall. "SpaceX launching prototype Google AI satellite next week." Space.com, September 25, 2026. https://www.space.com/space-exploration/satellites/spacex-launching-prototype-google-ai-satellite-next-week

79. Planet Labs PBC. "Planet Launches Suncatcher, Tanager-2, and 18 SuperDove Satellites." Press release, October 1, 2026. https://investors.planet.com/news/news-details/2026/Planet-Launches-Suncatcher-Tanager-2-and-18-SuperDove-Satellites/default.aspx

80. Travis Beals. "Our Project Suncatcher prototype satellite is in orbit." The Keyword (Google), October 1, 2026. https://blog.google/innovation-and-ai/models-and-research/google-research/project-suncatcher-prototype/

81. Blaise Aguera y Arcas et al. "Toward a future space-based, highly scalable AI infrastructure system design." Joule, article 102678, published online October 1, 2026. https://doi.org/10.1016/j.joule.2026.102678

82. Jonathan McDowell. "GCAT current catalog" (General Catalog of Artificial Space Objects; entry S66303, Starcloud-1). Updated October 15, 2026. https://planet4589.org/space/gcat/data/derived/currentcat.html

83. SpaceX. Letter to the FCC responding to Space Bureau questions, ICFS File No. SAT-LOA-20260108-00016. September 16, 2026. https://api-prod.fcc.gov/icfs-attachment/exp/api/v1/a81df57a93db4f10ac9c39bfa8373c08

84. Federal Communications Commission, Space Bureau. Letters requesting additional information from SpaceX (SAT-LOA-20260108-00016), Starcloud (SAT-LOA-20260202-00073), Blue Origin (SAT-LOA-20260310-00118), Cowboy Space (SAT-LOA-20260323-00135) and Orbital Compute (SAT-LOA-20260624-00251). August 28, 2026. https://api-prod.fcc.gov/icfs-attachment/exp/api/v1/06bf7724cf4b4f107cd9f9c41f851c98 (see also https://api-prod.fcc.gov/icfs-attachment/exp/api/v1/27ad3720cf4b4f107cd9f9c41f851c67 , https://api-prod.fcc.gov/icfs-attachment/exp/api/v1/79d59a6d2f038310c7a543ca6fa4e3f1 , https://api-prod.fcc.gov/icfs-attachment/exp/api/v1/46d83328cf0b4f107cd9f9c41f851ce5 and https://api-prod.fcc.gov/icfs-attachment/exp/api/v1/367eff60cf4b4f107cd9f9c41f851c39 )

85. Federal Communications Commission. "FCC Modernizes its Approach to NEPA Environmental Regulations." News release on Report and Order and Further Notice of Proposed Rulemaking FCC 26-64, September 30, 2026. https://docs.fcc.gov/public/attachments/DOC-425491A1.pdf

86. Federal Communications Commission. "Applications Accepted for Filing." Public Notice, Report No. SAT-02024, July 31, 2026. https://docs.fcc.gov/public/attachments/DOC-424016A1.pdf

87. Zhao Chenchen. "China launches eight satellites for space-based computing." CGTN, October 14, 2026. https://news.cgtn.com/news/2026-10-14/China-launches-eight-satellites-for-space-based-computing-1ReYc4EvEQ0/p.html

88. Andrew Jones. "Chinese launches boost Guowang constellation, add space computing sats." SpaceNews, October 15, 2026. https://spacenews.com/chinese-launches-boost-guowang-constellation-add-space-computing-sats/

89. Fu Yifei. "Kinetica-1 'one rocket, nine satellites' launch succeeds" (in Chinese). Science and Technology Daily, September 20, 2026. https://www.stdaily.com/web/gdxw/2026-09/20/content_584397.html

90. CGTN. "China launches integrated comms-sensing-compute demo satellite." September 20, 2026. https://news.cgtn.com/news/2026-09-20/China-launches-integrated-comms-sensing-compute-demo-satellite-1QAMNqSkW6A/p.html

91. "双星发射！三体计算星座加速组网 共商共建释放聚变效能" (Two satellites launched as the Three-Body Computing Constellation speeds up networking), Science and Technology Daily, July 24, 2026. https://www.stdaily.com/web/gdxw/2026-07/24/content_552976.html

92. Zhu Han and Song Chen, "具备太空在轨计算能力 我国太空计算卫星星座发射看点解析" (China's space computing satellite constellation launch explained), Xinhua via China Securities Journal, May 14, 2025. https://cs.com.cn/5g/202505/t20250514_6491297.html

93. "国星宇航一箭八星发射任务取得圆满成功" (ADA Space's eight-satellite launch succeeds), Science and Technology Daily, October 14, 2026. https://www.stdaily.com/web/gdxw/2026-10/14/content_596205.html

94. "一箭十星！国星宇航'星算'02组8星入轨，太空计算星座再扩容" (Ten satellites on one rocket: eight Star-Compute Group 02 satellites reach orbit), Cailian Press, October 14, 2026. https://www.cls.cn/detail/2504855

95. "Chinese scientist eyes AI-powered satellite constellation," CGTN, August 5, 2026. https://news.cgtn.com/news/2026-08-05/Chinese-scientist-eyes-AI-powered-satellite-constellation-1PmwuyTSETe/p.html

96. Carolyn A. Mahoney, letter to Joseph Yaffe, Cowboy Space Corp., extending the response date, ICFS File No. SAT-LOA-20260323-00135, Federal Communications Commission, Space Bureau, September 11, 2026. https://api-prod.fcc.gov/icfs-attachment/exp/api/v1/ca34c7813bd38b101c9b37ac24e45a80

97. Cowboy Space Corp., supplement responding to the Space Bureau's August 28, 2026 letter, ICFS File No. SAT-LOA-20260323-00135, Federal Communications Commission, September 18, 2026. https://api-prod.fcc.gov/icfs-attachment/exp/api/v1/76a19353cfdf8b104470b6770f851c29

98. Cowboy Space, "Our First Mission Takes Flight," Cowboy Space, October 1, 2026. https://www.cowboyspace.com/news/reason-1-takes-flight

99. Victoria Woodburn, "Transporter-18 Tests the Building Blocks of Orbital Data Centers," Payload, October 1, 2026. https://payloadspace.com/transporter-18-tests-the-building-blocks-of-orbital-data-centers/

100. Sidus Space, "Sidus Space to Launch LizzieSat-4 and LizzieSat-5 Together on SpaceX Bandwagon-6," Sidus Space, September 28, 2026. https://investors.sidusspace.com/news-events/press-releases/detail/294/sidus-space-to-launch-lizziesat-4-and-lizziesat-5-together

101. Lonestar Data Holdings, "StarVault: The World's First Commercial Data Storage Service From Space," Lonestar, April 15, 2026. https://lonestar.space/press-releases/starvault-the-world-s-first-commercial-data-storage-service-from-space

102. Mike Wall, "Polish startup Ares Shield hired to protect data center satellites with high-power microwave weapons," Space.com, August 25, 2026. https://www.space.com/space-exploration/satellites/satellite-defense-ares-shield-contract-lonestar-data-holdings

103. Axiom Space, "Axiom Space, Kepler Ready Orbital Computing for Quantum Era," Axiom Space, October 6, 2026. https://www.axiomspace.com/release/axiom-space-kepler-ready-orbital-computing-for-quantum-era

104. Axiom Space, "Axiom Space, Spacebilt Announce Orbital Data Center Node Aboard International Space Station," Axiom Space, September 16, 2025. https://www.axiomspace.com/release/axiom-space-spacebilt-announce-orbital-data-center-node

105. Will Robinson-Smith, "SpaceX deploys NASA's Pandora, other smallsats amid 1st 'Twilight' rideshare mission," Spaceflight Now, January 10, 2026. https://spaceflightnow.com/2026/01/10/live-coverage-spacex-to-launch-1st-twilight-rideshare-mission/

106. Kepler Communications, "Kepler Announces On-Orbit Compute Capacity on Optical Data Relay Network," Kepler, April 7, 2025. https://kepler.space/kepler-announces-on-orbit-compute-capacity-on-optical-data-relay-network/

107. Travis Beals. "We've received our first response from the Project Suncatcher prototype satellite." The Keyword (Google), October 30, 2026. https://blog.google/innovation-and-ai/models-and-research/google-research/project-suncatcher-satellite-response/

108. Google. Post on X, October 30, 2026. https://x.com/Google/status/2116244898269610456

109. Tim Fernholz. "Google's first data center satellite is online and writing poetry." TechCrunch, October 30, 2026. https://techcrunch.com/2026/10/30/googles-first-data-center-satellite-is-online-and-writing-poetry/