Cowboy Space Corp. has raised $275 million in Series B funding to build orbital data centers powered by sunlight and launched on a rocket the company is designing itself. The round signals investor appetite for a thesis that has moved from science-fiction adjacent to venture-fundable in roughly eighteen months: if Earth cannot deliver power to AI fast enough, put the compute where the power is.
The company plans to deploy a constellation called Stampede. Each satellite will double as a data center, with the rocket’s upper stage itself converting into orbital infrastructure once deployed. A first demonstration satellite is targeted for later in 2026. The company’s own rocket is scheduled to fly in 2028, according to Space.com.
The pitch: the grid is the bottleneck, orbit is the workaround
Cowboy’s framing is blunt. According to the company, Earth’s energy grid infrastructure cannot keep pace with AI’s growing power demands, which is why they believe orbital data centers offer a solution. The company cites grid connection lead times of five to seven years or more in major US markets for new data centers.
That number is not marketing exaggeration. Amazon Web Services has acknowledged the same constraint in Europe, where Pamela MacDougall, head of energy markets and regulation for AWS EMEA, told Reuters that securing power for a new facility can take up to seven years, with hubs like Frankfurt, London, Amsterdam, Paris and Dublin sometimes stretching to a decade. In US hot spots like Northern Virginia, similar multi-year waits have been documented.
The macro picture supports the diagnosis even if it does not validate the cure. Global data center capacity is projected to roughly double from 103 GW to 200 GW by 2030, requiring up to $3 trillion in investment, according to JLL’s 2026 Global Data Center Outlook. AI workloads are expected to consume half of all that capacity. The same report identifies grid interconnection — averaging more than four years in primary markets — as the binding constraint on the entire build-out.
What Cowboy is actually building
The technical architecture is unusual. Rather than separate the launch vehicle from the payload, Cowboy intends to convert the rocket’s upper stage into a functioning orbital data center hub. The satellites would be powered by solar arrays and connected via high-speed optical links. A second mission would test a cluster of satellites for laser communications between platforms.
The first demonstration launch later this year is also designed to test power beaming from space to Earth — a separate capability that, if it works at scale, would invert the architecture entirely: orbital generation feeding terrestrial compute rather than orbital compute avoiding terrestrial generation.
The engineering team is drawn from leading aerospace and technology companies including SpaceX, Astranis, Amazon’s Project Kuiper and NVIDIA.
The numbers don’t quite add up — yet
One megawatt is a useful demonstration. It is not a serious data center. Modern AI training racks already pull substantial power, with next-generation racks expected to draw even more. A single megawatt would power only a handful of such racks. Hyperscalers are now planning facilities measured in gigawatts.
That gap matters because the economics of Cowboy’s approach depend on solving a problem at scale, not as a tech demo. AWS has added gigawatts of capacity in recent quarters. To meaningfully offset terrestrial grid pressure, Stampede would need to deploy thousands of satellites — or scale individual platforms by orders of magnitude — while keeping launch costs low enough to compete with natural gas, solar-plus-storage and small modular reactors as the alternatives terrestrial operators are already pursuing.
Cowboy’s bet is that the rocket-and-payload integration eliminates enough cost to make the math work. Whether that bet survives contact with thermal management, radiation hardening, and the bandwidth required to actually move AI training data to and from orbit is the open engineering question.
The vertical integration play
The most interesting strategic feature of Cowboy is its insistence on building its own rocket. SpaceX has made third-party launch capacity cheaper than ever, and most satellite operators are happy to ride along. Cowboy’s choice to build a homegrown vehicle — with first flight not scheduled until 2028 — adds years of risk and hundreds of millions in capex.
The logic appears to be cadence and control. If Stampede requires frequent, dedicated launches with upper stages purpose-built to become data centers, a generic ride-share model doesn’t work. The upper stage isn’t a delivery mechanism; it’s the product. That makes Cowboy structurally closer to SpaceX’s Starlink model — vertically integrated launch and payload — than to a conventional satellite startup.
It also places the company in the same category as the broader trend of operators building infrastructure they once would have rented. As Space Daily has reported, the dominant pattern in commercial and government space alike is now constellation-scale buildouts that demand custom platforms.
What investors are actually buying
Index Ventures and the Series B syndicate are not betting on Cowboy successfully replacing terrestrial hyperscale capacity by 2030. The math doesn’t permit that. They are betting on something narrower and more plausible: that some workloads — sovereign AI compute, latency-tolerant batch training, defense-adjacent inference — will pay a premium for power that is not gated by a seven-year grid queue.
Sovereign infrastructure investment represents billions in capex opportunity by 2030, with AI training facilities commanding substantial lease rate premiums over traditional data centers. If Cowboy can capture a sliver of that premium tier — particularly customers willing to pay for orbital placement as a feature rather than a workaround — the unit economics could work even at one-megawatt scale.
The other thing investors are buying is optionality on power beaming. If the first demonstration this year shows orbital solar can be transmitted to terrestrial receivers at meaningful efficiency, Cowboy becomes a different kind of company entirely — an energy infrastructure play that happens to launch rockets.
The broader signal
The fact that a $275 million round can be raised on this thesis says something about how the AI infrastructure crunch has reshaped what counts as a reasonable adjacent investment. Three years ago, orbital data centers were a slide in a futurist’s deck. Now they are a funded company with engineers from SpaceX and NVIDIA and a launch manifest. The grid bottleneck is doing the work that no amount of pitch deck enthusiasm could.
Whether Stampede flies on schedule is a separate question. First-flight rockets often experience delays, and Cowboy is attempting both a novel vehicle and a novel payload architecture simultaneously. The 2028 launch date should be read as aspirational.
But the underlying conditions Cowboy is responding to are not aspirational. They are documented, persistent, and getting worse. The grid will not catch up to AI demand this decade. Someone is going to monetize that gap. Cowboy has now raised enough money to find out whether the answer is in orbit.
Articles cover space industry news and the Mind & Meaning pillar (human psychology, ambition, isolation, meaning under extremes). The Space Daily Editorial Team produces content through a collective process: research, source verification, drafting, technical review, and editing. Articles under this byline reflect the team’s editorial judgment rather than a single writer’s. Space Daily takes editorial responsibility for content published under this byline. See our editorial policy for more on how we work.
Photo by Zelch Csaba on Pexels
