Cowboy Space Corporation has unveiled Big Iron, a stainless steel launch vehicle whose upper stage stays in orbit as a megawatt-class data center, SpaceNews reported on October 27, 2026. The rocket is the delivery system for Stampede, the company’s proposed orbital data center constellation, and Cowboy Space has already filed an FCC application covering 20,000 satellites.
The headline numbers are easy to repeat. Big Iron stands 86 meters tall, uses a reusable booster, and is designed to carry more than 15,000 kilograms to sun-synchronous orbit. The company wants to build the first vehicle by the end of 2027 and fly it in 2028. No price has been confirmed for a launch, a satellite, or a slice of orbital compute.
The timing matters for anyone who buys or builds cloud capacity. Power, land, and chips are the three bottlenecks of the AI buildout, and Cowboy Space is betting that orbit solves the first two. Our earlier look at Musk’s claim that SpaceX closed a $40 billion GPU debt deal showed how fast the money chasing space-based compute is growing. This article breaks down what Cowboy Space actually announced, what the numbers imply, and what stays unproven.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
What Cowboy Space Announced About Big Iron
The announcement bundles three things: a rocket, a constellation, and a commercial program. Big Iron is the rocket. Stampede is the constellation of satellites that will run GPUs in orbit. Big Iron Express is the offer to outside customers who need megawatt-class power in space and do not want to build their own spacecraft.
According to Axios, the booster runs on nine liquid-oxygen-and-methane engines called Clydesdale, and the company aims to finish the first Big Iron by the end of 2027. SpaceNews and Payload Space both describe the same architecture: the second stage is not a throwaway part. It becomes the satellite.
That single design choice drives everything else in the plan. A conventional launch puts a satellite on a rocket and discards the upper stage. Cowboy Space merges the two, so the structure that gets to orbit is also the structure that holds the compute racks, the solar arrays, and the radiators.
Big Iron by the Numbers
The fact sheet from the announcement and the coverage that followed gives a fairly complete spec picture, with one exception on engine thrust that we cover below. The table collects every figure we could tie to the company’s stated plans.
| Item | Reported figure | Status |
|---|---|---|
| Vehicle height | 86 meters | Company-stated |
| Construction | Stainless steel, reusable booster | Company-stated |
| Payload to sun-synchronous orbit | More than 15,000 kg | Company-stated |
| Booster engines | Nine Clydesdale (liquid oxygen and methane) | Company-stated |
| Thrust per engine | 30,000 lbf per Axios, conflicting reports exist | Unresolved |
| Orbital data center footprint | 132 m by 73 m in dawn-dusk orbit | Company-stated |
| Compute payload | Five racks of NVIDIA Vera Rubin GPUs | Company-stated |
| Power class | Megawatt-class (about 1 MW) | Company-stated |
| First vehicle built | End of 2027 | Target |
| First flight | 2028 | Target |
| Launch price | None confirmed | Unknown |
Two points stand out. First, the payload figure of more than 15,000 kilograms to sun-synchronous orbit puts Big Iron in the medium-lift class rather than the heavy-lift class. Second, the 132 by 73 meter footprint works out to roughly 9,600 square meters of deployed area by simple multiplication, which is about the size of a large warehouse floor hanging in space. That area is mostly solar collection and heat rejection, the two things that decide how many watts a spacecraft can use and shed.
Why the Upper Stage Doubles as the Satellite
The company’s own wording is blunt. On its website, Cowboy Space says that in its system the rocket’s upper stage is the satellite itself. That sentence is the whole thesis, and it carries a cost argument. Every kilogram of dead structure that never reaches the customer’s mission is money spent on delivery.
Founder and CEO Baiju Bhatt framed the origin story in the announcement distributed by Business Wire via Morningstar. In his words: “We set out to build data centers in space and found there weren’t enough rockets to get them there, so we’re building our own.”
That quote explains why a data center company is now a launch company. Vertical integration is a familiar play in this industry, and the logic here is similar to how SpaceX built its own rockets to fly its own satellites. Cowboy Space is trying to apply the same loop to compute instead of communications.
The configuration is also flexible. Chief operating officer Joseph Yaffe said that the upper-stage payload can be swapped for one that provides a customer with a megawatt of power, or for one that dispenses satellites. That makes Big Iron a rocket with two personalities. Fly it one way and it is a conventional deployer. Fly it the other way and it becomes a power plant with a launch history.
Stampede: 20,000 Satellites and a Pending FCC Filing
Stampede is the product that Big Iron exists to build. The company’s website describes it as a constellation of satellites that uses abundant solar power in low Earth orbit to run on-orbit GPU data centers. The FCC application covers 20,000 satellites.
Two cautions apply. A filing is a request, not a license, and nothing in the fact record shows FCC approval. Also, the 20,000 figure is a proposed ceiling. No source we reviewed confirms how many Stampede satellites exist today, and the fact sheet gives no count of operational units.
The scale is still worth examining. If every satellite delivered the megawatt-class power the company describes, the arithmetic implies roughly 20 gigawatts of aggregate capacity across a full constellation. That is a back-of-envelope product of two company-stated numbers, not a figure Cowboy Space has committed to, and it assumes every platform flies the full compute payload. Treat it as an upper bound that shows the ambition, not a forecast of what ships.
Clydesdale Engines and the Thrust Discrepancy
The two conflicting figures
Axios reports 30,000 pounds of thrust per Clydesdale engine. Other coverage, including reporting that cites Payload Space, points to a figure ten times higher. The announcement coverage we reviewed does not settle it, so we flag it as unresolved rather than pick a side.
What each number would imply
Nine engines at 30,000 pounds each would total 270,000 pounds of liftoff thrust. Nine at 300,000 pounds each would total 2.7 million pounds. The gap matters because a medium-lift vehicle that stands 86 meters tall and carries more than 15,000 kilograms to sun-synchronous orbit needs to lift its own mass off the pad. Only one of the two figures is likely to fit a vehicle of that size, and Cowboy Space will need to publish a specification sheet to end the argument.
Until it does, any analysis that leans on Clydesdale thrust, including payload-to-orbit math and launch cost estimates, rests on shaky ground. We use the 86 meter and 15,000 kilogram figures because the company states them directly, and we avoid deriving anything from the thrust number.
Vera Rubin GPUs in Orbit: What Five Racks Means for Cloud Computing
Each Big Iron data center satellite is designed to carry five compute racks of NVIDIA Vera Rubin GPUs. That is a small footprint compared with a terrestrial hall holding thousands of racks, but it is a large first step for hardware that has to survive launch loads, vacuum, radiation, and thermal swings.
Joseph Yaffe described the goal to Payload Space this way: “What we want to do is put a lot of available power in orbit, in our case to power chips, to power AI data centers.” The sentence puts power first and chips second, which matches how the industry now talks about AI capacity.
The pitch contrasts with the ground. Terrestrial AI sites are running into grid queues, local opposition, and cooling limits. Our explainer on why data centers use 2 million gallons of water a day shows one of the pressure points that orbit would sidestep. Space has constant sunlight in a dawn-dusk orbit, and no water supply to fight over.
Orbit also brings its own problems. Heat has nowhere to go except radiation into space, which is why that 132 by 73 meter footprint is so large. Chips must tolerate radiation. Servicing a failed rack means a launch, not a technician with a cart. None of these are new objections, and the company has not published its answers in the material we reviewed.
Big Iron Express: Selling Power and Launch to Outside Customers
Big Iron Express is the commercial program for outside customers who need megawatt-class power in orbit. The company opened commercial launch to government and third-party payloads, according to the Business Wire headline carried by Morningstar. It gives Cowboy Space a revenue path before Stampede reaches scale.
The dual configuration is what makes the offer interesting. A customer could buy a ride for its own satellites, or buy the power-and-compute platform itself. For a defense agency or an Earth-observation operator, processing data near where it is collected can cut the volume of raw data sent to the ground.
Cloud buyers should read this as an early signal rather than a purchasing option. No price list, service-level terms, or customer names appear in the verified record. For readers tracking how orbital capacity might slot into hybrid architectures, our cloud computing coverage will follow the story as terms emerge.
From Aetherflux to Cowboy Space: Funding and Timeline
Cowboy Space did not start as a rocket company. Via Satellite reported in May 2026 that Aetherflux rebranded as Cowboy Space, expanding its plans to include in-orbit compute and launch. The company’s own site also carries an announcement of a $275 million Series B, titled Introducing Cowboy Space: A $275 Million Series B.
Reports on the launch also describe a first satellite, Reason-1, flown on October 1, 2026 to test laser power beaming from orbit to Earth, and a follow-on Reason-2 planned for 2027 to run GPUs in orbit. We could not tie those details to the primary announcement text, so treat them as reported rather than confirmed.
| Date | Milestone | Source |
|---|---|---|
| May 2026 | Aetherflux rebrands as Cowboy Space, adds compute and launch plans | Via Satellite |
| Oct. 1, 2026 | Reason-1 reportedly launches to test laser power beaming | Reports |
| Oct. 26-27, 2026 | Big Iron and Stampede unveiled | Business Wire, SpaceNews, Axios |
| 2027 | Reason-2 planned, first Big Iron targeted for end of year | Company targets |
| 2028 | First Big Iron flight targeted | Company target |
The schedule is aggressive for a new orbital-class vehicle. Building the first rocket by the end of 2027 and flying it in 2028 leaves little room for the usual test slips. The company has not said that Big Iron has launched, finished testing, or entered commercial service, and none of those milestones is confirmed.
How Big Iron Compares With Other Orbital Compute Plans
Cowboy Space is not alone. Industry coverage points to several rivals chasing space-based compute, including Starcloud, Google’s Suncatcher initiative, SpaceX’s launch and satellite business, and Blue Origin’s infrastructure ambitions. We have not seen verified spec sheets for the competing efforts, so the table compares approach rather than performance.
| Player | Approach | Launch link in the reported plan |
|---|---|---|
| Cowboy Space | Upper stage is the data center satellite | Own rocket (Big Iron, 2028 target) |
| SpaceX | Large-scale launch plus satellite infrastructure | Own launch and satellites |
| Blue Origin | Launch and space infrastructure ambitions | Own launch business |
| Google (Suncatcher) | Space-based compute initiative | Not covered in the announcement |
| Starcloud | Orbital data center plans | Not covered in the announcement |
The difference is the merge of launch and satellite. Any operator that buys rides from another company faces the capacity squeeze satellite builders already feel. Telesat, for example, needs 14 SpaceX Falcon 9 launches to deploy 225 satellites, and that dependence on a competitor’s manifest is exactly what Cowboy Space is trying to avoid.
Historical Context: From Reusable Boosters to Orbital Compute
Orbital computing has been a talking point for years, but the economics only began to make sense after reusable boosters cut the price of reaching orbit. Cowboy Space is building on that shift with a reusable first stage, and then pushing further by reusing the upper stage’s mass as the product.
Launch access remains a choke point for the whole sector. Canada’s experience shows the risk of leaning on one provider. Our report that Canada relied on SpaceX for 12 of 12 launches captures why governments and operators want alternatives. Big Iron Express targets that appetite directly by opening commercial launch to government payloads.
Power is the other thread in the company’s story. Yaffe’s emphasis on available power in orbit, and the reported Reason-1 power beaming test, suggest the company sees energy as the product and compute as the customer for it.
Market Impact for Cloud Providers and Data Center Builders
Near term: little changes for buyers
No orbital data center from Cowboy Space can serve a customer before 2028 at the earliest, and likely later. Hyperscalers will keep spending on land-based capacity. The financing pressure behind that spend is real, as our coverage of the AI cash crunch, with Meta seeking $95 billion, makes clear. Anything that promises cheaper power will get a hearing from investors.
Longer term: a new tier of capacity
If Stampede ever reaches meaningful scale, it would add a distinct tier to the cloud market. The likeliest early workloads are the ones that already start in space: satellite imagery, defense sensing, and communications processing. General-purpose AI training in orbit faces harder odds because of bandwidth, latency, and repair constraints.
NVIDIA also stands to benefit. A Vera Rubin deployment in orbit is a showcase customer win, even at five racks per platform. Chip suppliers get another channel, and launch providers get another reason to compete on mass to orbit.
What Remains Unconfirmed
The gap between announcement and operation is wide. Reports describe design intent, and several claims can circulate that the record does not support. The table lists what holds up and what does not.
| Claim | Verdict |
|---|---|
| Big Iron has launched | Not confirmed |
| Big Iron has completed testing | Not confirmed |
| FCC approval for 20,000 satellites | Not confirmed, application filed |
| Big Iron in commercial service | Not confirmed |
| Exact launch date | Not confirmed, 2028 target only |
| Count of operational Stampede satellites | Not confirmed |
| Per-engine Clydesdale thrust | Conflicting reports |
| Launch price | None identified |
Engineering unknowns pile on top of the regulatory ones. Thermal management, radiation tolerance, inter-satellite networking, and orbital servicing all remain open questions in the coverage we reviewed. Each could slip the schedule or cap the power each platform can actually use.
Five Predictions for Cowboy Space and Orbital Compute
These are our analysis, not reported facts, and each rests on the verified record above.
- The thrust figure gets corrected within weeks. A tenfold gap between published numbers is too visible to linger, and Cowboy Space will likely issue a spec sheet or correction.
- The 2028 first flight slips. New vehicles rarely hit their first target, and a build date of late 2027 leaves almost no test margin. Expect a 2029 debut to be the more common planning assumption.
- Government payloads arrive first. Big Iron Express courts government customers, and defense and Earth-observation buyers have the clearest use for processing data in orbit.
- Rivals respond with launch deals. Operators that lack their own rockets will lock in capacity or acquire launch assets, because the Cowboy Space model exposes that dependence.
- FCC action stays pending through 2027. A 20,000-satellite filing invites interference and debris questions, and regulators rarely move fast on large constellations.
Frequently Asked Questions
What is Cowboy Space’s Big Iron?
Big Iron is a stainless steel, 86-meter launch vehicle with a reusable booster. Its upper stage is designed to stay in orbit as a megawatt-class data center instead of being discarded.
What is the Stampede constellation?
Stampede is Cowboy Space’s proposed constellation of satellites that use low Earth orbit solar power to run on-orbit GPU data centers. The company has filed an FCC application for 20,000 satellites, and approval has not been confirmed.
When will Big Iron fly?
Cowboy Space aims to build the first vehicle by the end of 2027 and fly it in 2028. No exact launch date has been confirmed.
What GPUs will the orbital data center use?
The upper-stage data center is designed to carry five compute racks of NVIDIA Vera Rubin GPUs.
How much payload can Big Iron carry?
The company describes a capacity of more than 15,000 kilograms to sun-synchronous orbit. The upper stage can also be configured to provide one megawatt of power or to dispense satellites.
What does Big Iron Express offer?
Big Iron Express is the commercial program for outside customers that need megawatt-class power in orbit, including government and third-party payloads.
How much will a Big Iron launch cost?
No confirmed price has been identified. Cowboy Space has not published launch or compute pricing in the coverage we reviewed.
How many thrust pounds does each Clydesdale engine produce?
Axios reports 30,000 pounds per engine, while other coverage reports a figure ten times higher. The discrepancy is unresolved until the company publishes a specification.
