Data centers in space: which company is ahead?







Every funding round, every investor, and who is really ahead in data centers, in one document. Have a look ↓
Summary
Starcloud is still ahead in the space data center race, with Axiom Space second and Aethero the strongest smaller challenger. The lead now rests more on money and partners than on new hardware in orbit.
The market is splitting into different races. Starcloud leads orbital AI compute for Earth users, Axiom leads connected infrastructure for spacecraft, Aethero leads productized edge computing and Lonestar leads sovereign storage.
Completed missions still cut through the hype. Starcloud, Axiom, Aethero and Lonestar have operated computing or storage systems beyond Earth. Cowboy put its first satellite in orbit in October 2026, but it tests power beaming, not computing. Sophia and Orbital have not flown yet. Some have flown hardware and some have not, so we sort through who has real customers, and who only promises, in our data center market report.
Starcloud raised a $250 million extension in August 2026 at a $2.3 billion post-money valuation, bringing its total to about $450 million. Its next satellite, Starcloud-2, is now due to be fully operational by 2027. Aethero's Titan, once aimed at October 2026, is now listed for the second quarter of 2027. No startup in this race has flown a new computing payload since March 2026.
Funding still does not map neatly to leadership. Aethero and Lonestar have produced unusual amounts of flight proof with small budgets, while Cowboy has raised about $355 million and still has no computing hardware in space.
Commercial proof remains thin. Lonestar has the clearest named-customer evidence. Starcloud says its 2027 work will serve customers including U.S. government agencies, booked a lunar payload with Firefly for no earlier than 2028 and signed a software deal with Quindar in October 2026. Nobody reports meaningful recurring orbital compute revenue or utilization.
The first viable market is still space-native processing, not moving ordinary Earth-based AI workloads into orbit. Processing satellite data beside the sensor cuts downlink volume; beating terrestrial data centers on cost is much harder.
The big outside players moved closer. Google put its first Project Suncatcher prototype with TPU chips in orbit on 1 October 2026. SpaceX unveiled its AI1 data center satellite in June 2026, and in October Elon Musk was reported as saying a Star Mind space computer co-designed with Nvidia would enter mass production in 2027. China's ADA Space launched eight more computing test satellites on 14 October 2026.
Launch is the new choke point. Starcloud's CEO said in August 2026 that the company needs to book "an enormous amount of launch" while Falcon 9 is due to wind down and Starship is not yet in regular service.
Starcloud keeps first place, but the lead is more fragile than its valuation suggests. Axiom can overtake it with proven customer workloads, Aethero can close the gap if Titan flies in 2027 as planned, and SpaceX or Google could reset the race once they fly real accelerators at scale.
Written by the New Market Pitch teamWe track new markets so founders and investors can move faster. About us
70+pages
This is just a preview page. Don’t worry if a number looks old: the report gets refreshed, and you’ll always get the latest version.
Which companies are really building data centers in space?
Seven startups currently deserve a place in the space data center race: Starcloud, Axiom Space, Aethero, Lonestar Data Holdings, Sophia Space, Cowboy Space and Orbital.
The category covers three related businesses. Starcloud, Cowboy and Orbital want to run large AI workloads in orbit for users on Earth. Axiom, Aethero and Sophia are starting with edge computing for satellites, spacecraft and space stations. Lonestar is focused on sovereign storage, backup and disaster recovery from Earth orbit and the Moon.
We include Axiom even though orbital computing is only one part of its much larger business. Aethero belongs here because it has flown its own computing payloads and is building Titan as a compute-as-a-service platform. We exclude Ramon.Space, Kepler and similar suppliers whose main business is selling processors, links or hosting to other operators. SpaceX, Google and China's ADA Space appear later as major threats, but none is a startup selling an outside orbital compute service today.
Completed missions shrink the serious group quickly. Starcloud, Axiom, Aethero and Lonestar have operated computing or storage systems beyond Earth. Sophia has funding and a booked demonstration. Cowboy flew its first satellite, Reason-1, on 1 October 2026, but its payload is a laser for power beaming. Orbital picked a satellite builder in September 2026 and targets 2028 for its first spacecraft.
| Company | What it is building | Strongest current proof | Cumulative funding |
|---|---|---|---|
| Starcloud | Solar-powered orbital GPU clusters for AI and cloud computing | Starcloud-1 ran a data-center-class Nvidia GPU and trained a small language model in orbit; Starcloud-2 now planned for 2027 | About $450 million (reported), valued at $2.3 billion in August 2026 |
| Axiom Space | Connected orbital cloud and edge-computing nodes | An ISS prototype plus two free-flying ARC nodes, with node-to-node links confirmed in October 2026 | More than $1.03 billion company-wide |
| Aethero | Space-rated edge computers and a larger compute-as-a-service satellite | Two in-orbit demonstrations, including a 157-TOPS Jetson Orin system | $8.4 million disclosed |
| Lonestar Data Holdings | Sovereign storage, backup and disaster recovery in Earth and lunar orbit | Four test flights, two lunar missions and a commercial storage service whose first orbital payload now targets no earlier than April 2027 | About $15.5 million reported |
| Sophia Space | Modular orbital edge-computing units built around thermal efficiency | An Apex demonstration planned for 2027 and a non-binding $300 million lease-financing letter | $22 million (company-stated) |
| Cowboy Space | Vertically integrated rockets, solar power and megawatt-scale orbital AI systems | Reason-1 power-beaming satellite launched in October 2026; no computing payload yet | About $355 million |
| Orbital | Distributed AI inference satellites, now targeting about 250 kilowatts each | Reflex Aerospace selected to build Orbital-1, targeted for 2028 | $5 million disclosed |
Is Starcloud the space data center leader today?
Starcloud is still ahead overall because it combines the most powerful AI hardware flown by a startup, by far the most dedicated funding and the closest ties to Nvidia.
The lead has changed in nature. In August 2026, Starcloud raised $250 million led by Manhattan West, with Nvidia and Cisco joining, at a $2.3 billion valuation according to Bloomberg and TechCrunch. It is co-developing hardware with Nvidia around the Space-1 Vera Rubin module and building a 100,000-square-foot factory in Woodinville, Washington. But its second satellite slipped from late 2026 to 2027, so its flight record is still one satellite, launched in November 2025. In late October 2026, Starcloud and Quindar said Quindar's mission-management software will run on Starcloud's future satellites, a plan rather than a deployment.
Axiom Space is the closest rival, although it is winning a different race. In October 2026, Axiom confirmed that its two ARC nodes on Kepler satellites can talk to each other and to the ground over optical links. It has more systems in orbit and a better network. Starcloud has the stronger single payload and a direct plan to sell GPU capacity.
Aethero remains third. It has flown two edge-computing systems on $8.4 million in disclosed funding. Titan, its larger cluster, missed its October 2026 target and is now listed on SpaceX's Bandwagon-6 rideshare in the second quarter of 2027. Lonestar follows as the storage specialist. Sophia, Cowboy and Orbital still need their first computing missions.
Starcloud keeps first place for now, but a bigger round is not more proof. Axiom could overtake it by showing real workload volume and paying customers before Starcloud-2 flies.
70+pages
Just a preview. If a number here looks old, no worries: your report comes with the latest data.
Which space data center company has flown the best hardware?
Starcloud has flown the strongest single piece of startup data center hardware, Aethero has the best record of repeated edge deployments, and Google now has data center chips of its own in orbit.
Starcloud-1 carried an Nvidia H100, a chip designed for terrestrial AI data centers rather than the low-power processors normally used on satellites. Starcloud says the satellite ran Google's Gemma model and trained Andrej Karpathy's nanoGPT after its November 2025 launch. Nvidia described the payload as roughly 100 times more powerful than earlier GPU compute flown in space. Starcloud-2, planned as an 8-kilowatt compute satellite with two units on rideshare flights, is now due to be fully operational by 2027, the company says.
Aethero deserves more attention than it usually receives. Its Deimos mission demonstrated onboard inference, software updates and edge training. Phobos then launched in March 2026 with an Nvidia Jetson Orin NX operating at up to 157 TOPS. Aethero and EnduroSat say Titan will cluster Jetson Thor modules to reach up to 16,000 TFLOPS. That figure is a company target until the satellite operates, now in the second quarter of 2027 at the earliest. In August 2026, Aethero also agreed to buy beamed power from Star Catcher for Titan and a future constellation. A target is not a flight record, which is why we judge who is ahead and who is falling behind in our data centers report.
Google changed the comparison on 1 October 2026. Its first Project Suncatcher prototype, built with Planet and launched on SpaceX's Transporter-18, carries Google TPUs to test radiation and temperature effects. Google says the satellite is operating as expected. It is a research test, not a service, but it puts a cloud giant's own accelerators in orbit.
No one has released the benchmarks a cloud buyer would need. We still lack sustained tokens per second, energy per task, thermal throttling, error rates and useful uptime for any of these payloads. A successful model run proves feasibility. A competitive cloud service needs sustained performance, day after day.
Axiom has more nodes in space but has disclosed fewer processor details. What it has disclosed is the network: 2.5 Gbps optical links compatible with the Space Development Agency's standard. Lonestar's systems have proved storage, encryption and data transfer during lunar missions, which is valuable but not raw AI compute.
| Company | Best flown hardware | What it proves | What remains unclear |
|---|---|---|---|
| Starcloud | Nvidia H100 on Starcloud-1 | A modern data center GPU can run AI workloads in orbit | Sustained throughput, uptime and whether Starcloud-2 flies on time in 2027 |
| Aethero | Jetson Orin NX on Phobos | Repeatable, radiation-aware edge computing on small satellites | Titan's claimed 16,000-TFLOPS performance has not flown yet; launch now listed for Q2 2027 |
| Axiom Space | ISS unit and two free-flying ARC nodes | Cloud-native nodes can be deployed and linked to each other and the ground | Processor type, power, workload volume and utilization |
| Lonestar | Storage and edge-processing payloads on four missions | Data services can survive cislunar travel and abnormal landings | Capacity, restore speed and long-duration reliability |
| Sophia Space | No hardware in orbit yet | Ground engineering, a Caltech-linked patent and a booked mission | Real thermal and compute performance |
| Cowboy Space | Reason-1 power-beaming satellite, no computing payload | The company can build and fly its own spacecraft | Almost every key computing assumption |
| Orbital | No computing payload in orbit yet | A chosen satellite builder and a 2028 target | Power, cooling, networking and economics |
Does Axiom Space run the most operational orbital data center infrastructure?
Yes. Axiom Space runs the broadest orbital data center setup, with three deployed systems and the clearest plan for connecting computing nodes to other spacecraft.
Axiom sent AxDCU-1 to the International Space Station in August 2025, then launched two free-flying ARC nodes on Kepler's optical relay satellites in January 2026. On 6 October 2026, Axiom said it had confirmed ARC-to-ARC and ARC-to-ground communications and migrated a post-quantum-encrypted workload in a ground test. An on-orbit encryption demonstration is next, with no date given. A larger ISS node with more than 1 petabyte of storage is planned for 2027.
That network design fits the earliest practical market. Earth-observation satellites, research payloads and autonomous spacecraft already create data in orbit. Processing those files nearby reduces downlink volume and delivers answers faster. Axiom is also pitching secure processing to agencies facing new U.S. post-quantum encryption deadlines, a concrete buyer group.
Starcloud remains ahead in individual compute power, and Aethero discloses more processor detail. Axiom's weakness is still transparency: we do not know how often customers use its nodes, what workloads they run or how much revenue they generate.
70+pages
This is a preview page. Spotted an outdated number? That’s normal here. The report you get is always the freshest version.
Which space data center companies have real customers and meaningful revenue?
Lonestar has the clearest named-customer evidence today, while Starcloud has the best chance of building a much larger revenue stream.
Lonestar has handled data for the State of Florida, Valkyrie AI, the Isle of Man government and other organizations across its test missions. Its Freedom payload performed uploads, downloads, encryption, authentication and edge-processing tasks on the way to the Moon.
The company has since launched StarVault as a commercial sovereign-storage service, ordered a second orbital payload from Sidus Space and signed a NASA Space Act Agreement. Timing has slipped: in September 2026, Sidus moved LizzieSat-4, which carries Lonestar's payload, to no earlier than April 2027. Lonestar has a product and named users, but it does not disclose prices, revenue or repeat usage.
Starcloud's commercial story is bigger but less proven. Crusoe announced plans to offer Starcloud's orbital GPU capacity through its cloud, with limited capacity originally expected from 2027. That plan depends on satellites that have not launched yet. In August 2026, Starcloud said its 2027 orbital inference work would serve customers including U.S. government agencies, without naming them or disclosing contract values. In September 2026, Firefly signed Starcloud as a commercial customer: an SC-1L compute payload will ride on Firefly's Elytra spacecraft in lunar orbit, on a mission targeted no earlier than 2028. Here Starcloud is the buyer of a ride, and Firefly disclosed no contract value.
Aethero may reach revenue through a simpler model: satellite operators buy its edge computers or rent hosted capacity. It says multiple software customers already run on Phobos, without naming them. Axiom says it has more than $2.2 billion in contracts across its wider business, but it does not separate orbital computing from astronaut missions and station work.
No company reports recurring orbital compute revenue, gross margins or meaningful utilization. Lonestar leads on visible commercial activity, Starcloud on revenue potential, and Aethero may have the shortest path to practical hardware sales.
| Company | Best customer or revenue evidence | How strong is it? |
|---|---|---|
| Lonestar | Several named users, a commercial storage service, a second orbital payload and a NASA agreement | Strongest proof of real commercial activity, though small and now delayed to 2027 |
| Starcloud | Crusoe distribution plan, unnamed U.S. government customers for 2027 and a Firefly lunar payload booked for 2028 or later | Best distribution path, but no disclosed bookings or contract value |
| Axiom Space | Large company-wide contract base and government support | Credible buyer access, but orbital data center revenue is hidden |
| Aethero | Commercial compute products, unnamed software customers on Phobos | Straightforward path to hardware sales, but named end users are scarce |
| Sophia Space | Non-binding $300 million lease-financing letter with SLI for 10 satellites | A financing route, not proven demand |
| Cowboy Space | Nvidia relationship and major investors | Strong backing, while customer demand remains unproven |
| Orbital | Investor-backed plan and a satellite builder under contract | No public customer proof yet |
Which space data center startup is moving fastest right now?
Starcloud is moving fastest on money, partners and factories, while Aethero has achieved the most per dollar. On computing flights, no startup has launched a new payload since March 2026, and the next two big ones have both moved into 2027.
Starcloud went from about $30 million raised to about $450 million in five months, with a $170 million Series A in March 2026 and a $250 million extension in August. It has 25 employees, according to TechCrunch, and is building production lines for its larger Starcloud-3 spacecraft. It added a Firefly lunar booking in September and the Quindar software deal in October. The weak point is timing: Starcloud-2 slipped to 2027, and Starcloud-3 depends on Starship.
Aethero's pace is quieter, and now slower. Phobos launched in March 2026 and the company signed a power deal with Star Catcher in August, but Titan, built on EnduroSat's FRAME-15 platform, missed its October 2026 target and is now listed for the second quarter of 2027. It has done this with about 2% of Starcloud's funding.
Cowboy is now flying hardware. Reason-1 launched on 1 October 2026 to test laser power beaming, and the company opened a 291,000-square-foot plant in Kent, Washington, in September to build rockets, engines and solar systems. It says its own rockets could fly from 2028. Its first computing system still has not flown.
Sophia added a $7 million SAFE in June, a patent with Caltech in July and the SLI financing letter in September. Orbital chose Reflex Aerospace in September and raised its power target per satellite to about 250 kilowatts. Lonestar slowed, with its next payload pushed to 2027.
70+pages
Preview only. Some numbers may have changed since we made this page, but your report will be up to date.
Has any space data center company solved orbital cooling?
No space data center company has solved cooling at commercial scale. Starcloud and Aethero have proved only that useful processors can run within small orbital power budgets.
Vacuum removes convection, so heat must travel through the spacecraft and escape from radiators as infrared energy. A 2026 technical study estimated that a representative one-megawatt orbital system could need about 2,500 square meters of radiator area and 5,640 square meters of solar panels. Fixed spacecraft hardware pushed the estimated mass to roughly 34 to 59 kilograms for every delivered kilowatt of computing power. Radiator area and mass are big hurdles up there, and we cover the biggest unsolved problems in our report covering data centers.
A commercial operator would need to keep thousands of chips near their target temperature for years. Radiator mass, deployment mechanisms, coolant loops and radiation damage could erase much of the benefit of abundant solar power.
Sophia has made thermal design the center of its TILE architecture and secured a patent on passive cooling with Caltech in July 2026. Starcloud says it is working with Nvidia on the trade-off between chip temperature and radiator size. SpaceX's AI1 design uses a deployable liquid radiator, and ADA Space says its new test satellites carry 10-kilowatt liquid-cooling loops. None of these has been shown at scale.
Can space data centers beat Earth data centers on cost?
Space data centers are still too expensive for ordinary terrestrial AI workloads, while space-native edge computing can already make economic sense in selected cases.
The strongest independent cost study published in 2026 reached a blunt conclusion. It assumed about 40 kilograms of orbital hardware for each delivered kilowatt. At that weight, spacecraft and launch costs would need to fall to roughly $250 to $1,000 per kilogram to match terrestrial infrastructure costing $10,000 to $40,000 per kilowatt. That allowance was 3.4 to 13.5 times below the public dedicated Falcon 9 benchmark, before spacecraft construction, communications, operations and replacement.
Starcloud argues that constant sunlight and simpler cooling can cut electricity costs sharply. Electricity savings alone cannot cover the system. Orbital operators must pay for launch mass, radiation protection, radiators, optical terminals, ground infrastructure and replacement of hardware that cannot be repaired cheaply.
The calculation changes when the data already sits on a satellite. A processor can turn a large image or sensor stream into a compact result before downlink. Some research demonstrations have cut transmitted data by more than 99% for specific Earth-observation tasks. Avoiding downlink congestion can justify orbital compute even when each unit of processing costs more than on Earth.
Axiom, Aethero and Sophia therefore have the clearest near-term economics. Starcloud's larger vision, and SpaceX's, depend on Starship making launch far cheaper.
70+pages
This is just a preview page. Don’t worry if a number looks old: the report gets refreshed, and you’ll always get the latest version.
Which space data center company can move enough data?
Axiom Space has the best networking position among the startups because its plan is built around optical links to other satellites, and it has now shown those links working between its nodes.
Orbital computing becomes useless when inputs cannot arrive and outputs cannot leave. Terrestrial AI clusters exchange data internally at enormous rates, while most satellite links run many orders of magnitude slower. Optical inter-satellite links help, but they need precise pointing, routing and enough ground gateways.
Axiom's ARC nodes use 2.5 Gbps optical links, and the company confirmed node-to-node and node-to-ground traffic in October 2026. Satellites can send data to nearby nodes, process it and return smaller results. Aethero follows similar logic by placing compute beside the sensor.
Starcloud needs a harder model because Earth users may send large datasets into orbit and expect fast results back. Starcloud-1 did not demonstrate a commercial high-capacity network. SpaceX has the obvious edge here: it says AI1 satellites will relay data over Starlink laser links. Lonestar can tolerate slower links because archival storage is less interactive.
For now, the best space data center workload is the one that creates high-value answers from few transmitted bits. Axiom and Aethero are closest to that rule.
Can Starcloud get enough launches to build orbital data centers?
Not yet with certainty. Launch has become the main constraint on Starcloud and every other startup that wants large orbital data centers, while Axiom and Aethero need far less of it.
In August 2026, Starcloud CEO Philip Johnston told TechCrunch the company needs to book "an enormous amount of launch" and that launch is among its biggest costs. SpaceX plans to retire Falcon 9 around 2028 in favor of Starship, which is not yet flying regular missions. Blue Origin's New Glenn and ULA's Vulcan are not flying often, and Rocket Lab's Neutron has not launched. Johnston said failing to book SpaceX capacity for 2029 would be challenging. Starcloud-3, its large spacecraft, is designed for Starship. Its September 2026 deal with Firefly adds one ride to lunar orbit no earlier than 2028, a single demonstration slot, not recurring capacity.
That creates an awkward dependence. SpaceX is both the likely launch provider and a direct competitor, with its own AI1 satellites and a Gigasat factory in Bastrop, Texas, due to start production by the end of 2027. A launch provider that is also a rival makes the race lopsided, and we look at how far the challengers are behind in our report on data centers.
Cowboy's answer is to own the rocket. It opened its Kent plant in September 2026 and plans first engine hot-fire next year, with launches from 2028. That could lower cost, or multiply what can go wrong. Axiom, Aethero and Sophia ride on proven buses and rideshares; Aethero uses EnduroSat's platform and a SpaceX Bandwagon rideshare, and Sophia an Apex bus, so their capital goes to computing and thermal design. Rideshare is cheaper, but it ties their dates to someone else's manifest.
Regulation adds another constraint. The FCC has accepted SpaceX's application for up to one million satellites for public comment, which is not approval, and petitions argue that very large fleets need cumulative environmental reviews covering debris, astronomy and atmospheric effects.
70+pages
Just a preview. If a number here looks old, no worries: your report comes with the latest data.
What can Starcloud, Axiom Space or Aethero do that competitors cannot easily copy?
Axiom is hardest to copy because of its orbital relationships, Starcloud has the deepest Nvidia ties and the most valuable high-power flight data, and Aethero has the clearest repeatable edge-computing product.
Axiom's position comes from years of ISS operations, government relationships, station development and mission management. Another startup can buy similar chips, but it cannot quickly reproduce access to the same customers, orbital infrastructure and operating history.
Starcloud's advantage will come from learning, not a single patent. Every mission produces data on power, radiation, cooling, software recovery and accelerator behavior, and Starcloud now shares that data with Nvidia as it designs its first space GPU module. That edge stays shallow until Starcloud-2 proves the lessons transfer beyond one demonstration. Nvidia also works with SpaceX, so the relationship is not exclusive. The October 2026 Quindar deal adds operations software, which any rival can also buy.
Aethero has taken a practical product route. Its modules fit several satellite sizes, support common software tools and can be bought without adopting a proprietary constellation. Two completed demonstrations give it credibility with spacecraft makers that need flight history more than grand ambition.
Lonestar owns the strongest niche around lunar and sovereign storage. Sophia now has patent protection for its passive-cooling design. Cowboy could lower costs by combining launch and compute, though that advantage exists only on paper so far.
Large companies can copy visible hardware more easily than flight records, customer workflows, network access, manufacturing yield and years of operating data.
Does Starcloud's $2.3 billion valuation make it the leader?
Partly. The valuation shows that investors back Starcloud more than any other startup in this race, but Starcloud leads because of flown hardware and partners, not because of the price tag.
Starcloud has raised about $450 million, according to Bloomberg; TechCrunch reported the $250 million extension without giving a total, but the known rounds add up to the same figure. Nvidia put about $25 million into the August round, according to a person cited by TechCrunch. Almost all of Starcloud's money is tied to orbital computing.
Axiom has raised more than $1 billion company-wide, including a round above $525 million in 2026, but most supports its station, spacesuit and human-spaceflight programs. Its orbital computing work also drew a $5.5 million grant in June 2026. The balance sheet helps, but does not tell us what Axiom spends on data center nodes.
Cowboy has raised about $355 million, including its $275 million Series B, and has now flown one power satellite. For now, that capital measures investor belief more than delivery.
Aethero is the leanest, with two computing missions and a third, much larger one prepared on $8.4 million in disclosed funding. Lonestar has completed four flights on roughly $15.5 million of reported capital.
Funding still matters because the next phase is brutally expensive. Starcloud has the best mix of money and proof. Axiom has the deepest resources. Aethero and Lonestar have achieved the most per dollar. Cowboy has the largest pool of capital that still needs to become computing hardware.
70+pages
This is a preview page. Spotted an outdated number? That’s normal here. The report you get is always the freshest version.
Can Starcloud and the startups hold off Google, SpaceX and China?
Only for a short time. SpaceX is the biggest threat to every startup here, Google now has hardware in orbit, and China already flies a computing constellation.
SpaceX unveiled its first orbital data center satellite, AI1, in June 2026. SpaceX says it delivers 150 kilowatts peak and uses a deployable liquid radiator. Separately, in October 2026, Elon Musk was reported as saying that Star Mind, a space computer co-designed with Nvidia, would enter mass production in 2027; the same reports describe each satellite as carrying the equivalent of one NVL72 rack. SpaceX has applied to the FCC for up to one million satellites. Beyond that production goal, Musk gave no date for deployment in orbit and no fleet size, and AI1 has not flown. Google and SpaceX both have big plans on paper, so we weigh who the likely future winners are in our data centers report.
Google's Project Suncatcher moved from paper to orbit on 1 October 2026 with one TPU prototype built with Planet. Google's earlier plan calls for a two-satellite test next. Google can design the chips, software and cloud interface together, but it depends on outside launch and satellites.
China is further along than most Western coverage admits. ADA Space and Zhejiang Lab launched 12 computing satellites in May 2025, each reported at 744 TOPS, for a planned 2,800-satellite network. On 14 October 2026, a Zhuque-2E rocket carried eight more ADA Space test satellites to orbit to test cooling under load and laser routing. Counted by spacecraft, China is ahead: about 20 reported computing satellites against one for Starcloud, though they are test spacecraft, not a commercial service.
Starcloud and Aethero keep a time advantage among Western startups because their hardware has flown. Axiom has a network and customer advantage in space-native work. Those leads could disappear quickly once SpaceX flies real accelerators or Google proves linked TPU satellites. Blue Origin remains a longer-term threat through its 51,600-satellite filing from March 2026, with no hardware announced.
Does one company lead every part of the space data center market?
No company leads every part of the space data center market because orbital AI, satellite edge computing and sovereign storage require different products.
Starcloud leads the startup race to provide powerful AI compute for users on Earth. Axiom leads connected infrastructure for satellites and space stations. Aethero leads the smaller, productized edge-computing segment. Lonestar is furthest ahead in off-Earth backup and sovereign storage. Outside the startup field, China's ADA Space leads on the number of computing satellites in orbit.
Sophia could become a strong supplier of modular compute if TILE proves its cooling design, and the SLI letter offers a way to finance satellites without selling more equity, if it becomes binding. Cowboy is trying to build the most parts itself, from launch to power to processing. Orbital makes the boldest small-company claim, up to 100,000 satellites, on $5 million in disclosed funding.
A single average score would hide these differences. Starcloud's powerful GPU says little about lunar backup. Lonestar's named customers say little about model training. Axiom's optical network does not prove the lowest compute cost. We rank Starcloud first overall because it leads the largest potential segment with strong evidence across several dimensions. Specialists still beat it in several niches.
70+pages
Preview only. Some numbers may have changed since we made this page, but your report will be up to date.
How reliable is the evidence behind the space data center race?
The broad ranking is reliable. Precise revenue, performance and cost comparisons are not.
We have high confidence in launches, funding rounds, named chips and regulatory filings. Those can be checked through launch records, company releases, the FCC and financing disclosures. Starcloud's GPU mission, Axiom's nodes, Aethero's Phobos launch, Lonestar's lunar tests, Google's prototype and Cowboy's Reason-1 are real milestones.
We have medium confidence in technical performance. Claims such as 100 times more compute, 16,000 TFLOPS or 150 kilowatts per satellite use different definitions and conditions. None of the contenders publishes an independent benchmark covering sustained throughput, power draw, radiation errors, cooling limits and uptime.
Commercial evidence is weaker. No company reports recurring orbital compute revenue, gross margin, utilization or a clean customer count. Funding needs care: Axiom's billion-dollar total covers several businesses, Sophia's $300 million is a non-binding letter, and Lonestar's reported total is not confirmed in full by public filings.
Schedules deserve caution. Starcloud-2, Lonestar's next orbital payload and Aethero's Titan all slipped to 2027; Titan's October 2026 target passed without a launch, and Aethero has not said publicly why. Constellation sizes deserve the lowest weight: Starcloud has filed for up to 88,000 satellites, Cowboy for 20,000, Orbital for 100,000 and SpaceX for one million. These describe regulatory ambition, not funded production.
Which companies are actually ahead in space data centers?
Starcloud is still the space data center leader among startups, with Axiom Space second and Aethero the strongest smaller challenger. The real pressure now comes from SpaceX, Google and China.
Starcloud ranks first because it has powerful hardware in orbit, about $450 million raised, close work with Nvidia and a focused commercial plan. Its next flight slipped to 2027, so its lead now rests more on resources than on new proof. Axiom ranks second because it has the broadest deployed infrastructure and has shown its nodes linking in orbit, though actual usage remains hidden.
Aethero is third with two completed edge-computing missions and Titan pushed to the second quarter of 2027. Lonestar ranks fourth as the storage specialist, with named users, a commercial service and a NASA agreement, but a payload delayed to 2027.
Sophia takes fifth with a focused design, $22 million, a patent and a 2027 demonstration. Cowboy ranks sixth: it flew its first satellite in October 2026, but for power, not computing. Orbital remains seventh until Orbital-1 flies, now targeted for 2028.
The ranking could change within a year. Starcloud must fly Starcloud-2 and turn Crusoe and government interest into contracts. Axiom needs to show customers and workload volume. Aethero must deliver Titan's claimed performance. SpaceX remains the outside force most capable of reshaping the field once its data center satellites reach production.
| Rank | Company | Why it ranks here now |
|---|---|---|
| 1 | Starcloud | Strongest flown AI compute among startups, about $450 million raised, Nvidia partnership and a route to cloud distribution; next flight now in 2027 |
| 2 | Axiom Space | Most deployed orbital data center infrastructure and confirmed node-to-node optical links, with limited visibility into usage and revenue |
| 3 | Aethero | Two successful edge-computing missions, a productized hardware strategy and a much larger cluster prepared on very little capital; Titan now due in Q2 2027 |
| 4 | Lonestar Data Holdings | Clear leader in sovereign space storage, with four flights, named users, a commercial service and a NASA agreement; next payload delayed to 2027 |
| 5 | Sophia Space | Serious thermal-first design, a patent, adequate demonstration funding and a 2027 mission, but no flight proof yet |
| 6 | Cowboy Space | Exceptional funding and a first satellite in orbit, but no computing hardware has reached space |
| 7 | Orbital | Clear vision and a satellite builder under contract, though its first spacecraft is not due before 2028 |
Our methodology
This analysis tests which company is ahead in space data centers by comparing what each contender has actually built, operated and started to commercialize. We treat orbital AI infrastructure, satellite edge computing, connected orbital cloud systems and sovereign storage as related but distinct parts of the market.
We assessed leadership across deployed hardware, technical performance, operational infrastructure, customer evidence, commercial readiness, execution speed, thermal design, economics, networking, manufacturing and launch access, and defensibility. No single metric decides the ranking: funding can show confidence without proving delivery, and a powerful payload can prove feasibility without proving a business.
Completed missions, operating hardware, named customers, signed agreements and observable deployment progress received the most weight. Future missions, constellation filings, non-binding financing letters, company performance claims and announced capacity targets were included when useful, but they received less weight until demonstrated in orbit.
We also separated dedicated orbital-computing funding from company-wide capital. This is especially important for Axiom Space, whose funding supports human spaceflight, spacesuits and station development as well as orbital data centers. SpaceX, Google and ADA Space are treated as outside threats rather than ranked startups.
The final ranking aggregates the evidence across all dimensions while preserving specialist leadership. That is why Starcloud ranks first overall even though Axiom leads deployed infrastructure, Aethero leads productized edge computing and Lonestar leads sovereign storage.
Key sources include Starcloud's Starcloud-1 mission record, TechCrunch's reporting on Starcloud's August 2026 round and launch constraints, Axiom Space and Kepler's October 2026 orbital computing release, EnduroSat's Titan mission announcement with Aethero, Aethero's platform page with the current Titan schedule, Starcloud's Starcloud-2 mission page, Firefly's September 2026 announcement of Starcloud's lunar payload, Sidus Space's LizzieSat-4 launch update, Sophia Space's SLI financing announcement, Cowboy Space's Reason-1 launch note, Payload's reporting on Orbital and Reflex Aerospace, Google's Project Suncatcher prototype announcement, and the independent study on orbital data center economics.
70+pages
This is just a preview page. Don’t worry if a number looks old: the report gets refreshed, and you’ll always get the latest version.
Written by the New Market Pitch teamWe track new markets so founders and investors can move faster. About us









