Talk Around Town: Cowboy Space Claims Its Rocket Upper Stage Turns Into an AI Data Center in Orbit

What Folks Are Hollering About

Well, slap a saddle on a server rack and call it a spacecraft — that's essentially what Cowboy Space is claiming it has done. At an event the company called 'Big Sky Rising,' held October 26, 2026, in Palo Alto, California, Cowboy Space — run by Robinhood co-founder Baiju Bhatt — unveiled what it calls 'Big Iron,' a rocket the company describes as turning its own upper stage into an orbiting AI data center after reaching space, according to Space.com and Axios.

The constellation the company says it aims to build with these rockets is called Stampede, and Cowboy Space claims it intends to populate low Earth orbit with roughly 20,000 of these combined rocket-satellite rigs. That figure, Lord have mercy, comes from an FCC application filing — which, as Space.com noted, sets a regulatory ceiling and starts a process rather than authorizing a dadgum thing. Four separate editorial outlets — Space.com, Axios, Payload Space, and the Los Angeles Times — all covered the announcement, lending solid footing to the fact that the event and the claims happened, even if the hardware hasn't left the ground yet.

What Is Actually Known for Certain

Here's what's got nails in it: Cowboy Space is a real company, formerly known as Aetherflux before rebranding, and it previously raised $275 million in a Series B funding round led by Index Ventures, as reported by Space.com in May 2026. That money is corroborated and about as solid as a cast-iron skillet.

The October 26 unveiling event happened and was covered by multiple independent newsrooms. Consistent across those outlets are the hardware figures the company announced: Big Iron, as described by Cowboy Space, would stand 282 feet tall, be constructed of stainless steel, and carry nine liquid-oxygen and methane engines on the first stage — which the company calls Clydesdale engines — plus one on the upper stage, mirroring the same 9+1 arrangement as SpaceX's Falcon 9. The company also says each resulting satellite would carry five compute racks of NVIDIA Vera Rubin GPUs and deliver roughly 1 megawatt of computing power. The company is also opening Big Iron to outside payload customers through a program it calls 'Big Iron Express,' offering access to up to one megawatt of power per Payload Space.

The core engineering concept, as Bhatt described it, is that the upper stage of the rocket does not separate and burn up on reentry — instead, the company says it transforms into the data center satellite, making the stage and the spacecraft one unified system. That's the pitch, and those are the company's words, consistent across multiple outlets that covered the event.

What Nobody Has Verified Yet

Now here's where the barn door's still wide open, friends. Big Iron has never flown. Not once. The only engine test the company has publicly pointed to is a combustion chamber firing at NASA Stennis — a component test, not a full engine run, and the milestone itself originates entirely from Cowboy Space. COO Joseph Yaffe told Payload Space that a team of roughly 30 people built and tested the Clydesdale combustion chamber in 90 days and that the company is, in Yaffe's words, almost a year ahead of schedule. Those are self-reported claims with no independent engineering corroboration, like a fella grading his own hog at the county fair.

The 20,000-satellite Stampede constellation is another critter that hasn't been roped yet. Analyst Douglas Messier, writing on his Substack, pointed out the plain arithmetic: 20,000 satellites divided by one satellite per launch equals 20,000 launches — a cadence that has no established precedent in the history of spaceflight, not even close. The company has also argued that orbital solar-powered data centers are a better answer to terrestrial energy grid constraints than building on the ground, but that claim has not been independently validated at any meaningful scale.

A first Big Iron launch is targeted by the company for December 2028, which is more than two years off as of this writing. The timeline is a company target, not a regulatory authorization or an engineering certification. Anyone expecting this rocket to be rolling out to the pad next Tuesday is going to be waiting a good long spell.

Our Analysis: The Vision Is Big; the Gap Is Bigger

Setting aside for a moment that Big Iron sounds like what happens when a data center engineer and a rodeo announcer collaborate on a napkin sketch — the underlying concept is legitimately interesting from an engineering standpoint, and that's worth saying. The idea of eliminating the separation between rocket stage and orbital payload to reduce mass overhead is not science fiction; it's a logical extension of integrated design thinking. Whether it works at scale is the trillion-dollar question, and right now there is no independent answer.

What concerns us analytically is the sheer mathematical audacity of 20,000 launches. Even if Cowboy Space executes flawlessly, even if Big Iron flies on schedule and performs exactly as the company describes, achieving 20,000 operational satellites would require a manufacturing and launch tempo that dwarfs anything humanity has ever managed. Messier's point isn't a gotcha — it's a reminder that ambition and logistics are two different pastures, and you can't just wave your hat and expect the cattle to sort themselves.

The $275 million Series B is real money, and Bhatt has a track record of building companies people initially laughed at. Index Ventures backing a venture at this scale suggests sophisticated investors see something worth betting on. But sophisticated bets and proven rocketry are different animals entirely, and readers should keep their boots firmly on the ground until Big Iron at minimum completes a full engine test, let alone a flight. We're watching, hats in hand, but we're not saddling up just yet.