NextFin News - Baiju Bhatt, the billionaire co-founder of Robinhood, has moved his next venture from pitch deck to orbit. Cowboy Space Corporation, the orbital infrastructure company he founded in 2024, launched its first satellite, Reason-1, on October 1 aboard a SpaceX rideshare, beginning a mission to beam power from space to Earth with a kilowatt-class laser — the opening move in a plan to build AI data centers in low Earth orbit. The company has raised $275 million at a $2 billion valuation, filed with regulators for a 20,000-satellite constellation, and set a target of operating its own rocket by the end of 2028. The question is no longer whether the idea is ambitious — it is whether the economics of orbital compute can beat the grid on Earth.
The First Satellite Is Already in Orbit
Reason-1 lifted off on the SpaceX Transporter-18 rideshare mission from California, with deployment confirmed at 3:41 p.m. ET. The satellite's mission is to collect solar energy in space and transmit it to a ground receiver using a kilowatt-class laser — what the company describes as the first demonstration of laser power beaming from orbit to Earth. In an October 8 interview, Bhatt said the satellite had reached orbit on the SpaceX rideshare and that the team was in daily contact with it, adding that Cowboy Space is developing "an energy grid in orbit."
"Every era of infrastructure, from the railroads to the power grid, has been built by those willing to push into new territory. The next one begins in orbit," Bhatt said in the company's launch statement. "Reason-1 is our first stake in that future, and the first step toward powering humanity from the high frontier."
The launch caps a rapid financing run. On May 11, Cowboy Space closed a $275 million Series B led by Index Ventures at a $2 billion post-money valuation. The round drew Andreessen Horowitz, NEA, Breakthrough Energy Ventures, Institutional Venture Partners, Construct Capital, Blossom Capital, Interlagos, SAIC, and Bhatt himself. It followed roughly $80 million of earlier funding, including a Series A led by Index Ventures in April 2025 — a pace of capital formation rare for early-stage orbital infrastructure. The company was founded as Aetherflux in 2024 with a space-based solar-power thesis and renamed Cowboy Space Corporation in 2026 as it pivoted toward orbital data centers.
The scale of the regulatory filing is the tell. Cowboy Space has asked the Federal Communications Commission for authorization for a 20,000-satellite constellation called Stampede, with each satellite weighing between 20,000 and 25,000 kilograms, generating 1 megawatt of power and supporting just under 800 graphics processing units. The company is collaborating with NVIDIA to deploy NVIDIA Space-1 Vera Rubin Modules in low Earth orbit, and it says its satellites will operate at altitudes of 800 to 1,000 kilometers — higher than the Starlink broadband constellation.
Why Space: The Terrestrial AI Power Crunch
The bet is built on a constraint that every hyperscaler is now feeling: AI data centers need power, and Earth is running short of convenient places to put it. Building a terrestrial data center means securing land, winning grid interconnection, and surviving local opposition to the transmission lines and substations that follow. In the United States, interconnection queues stretch for years, and data-center developers increasingly compete with municipalities and industrial users for the same finite grid capacity.
Orbit offers a different physics problem. Research on space-based solar power has long noted that a satellite in the right orbit can be illuminated more than 99 percent of the time, compared with ground-based panels that collect power for only about 29 percent of the day on average. There is no cloud cover, no night cycle, and no NIMBY opposition at 1,000 kilometers up — and no grid queue, because the power is generated where the compute sits. For workloads that can tolerate the latency of a space-to-Earth round trip, the relevant cost is not the distance the electricity travels but the cost per kilowatt-hour of delivering it to the chip.
Bhatt has framed the scarcity in blunt terms. "There's a lot of new rockets that are coming online, but as we look three, four years out, it's still very, very scarce," he said. "I think that you're going to see a lot of the first-party rocket providers actually specialize into their own payloads." His answer to that scarcity is to become one of those providers: Cowboy Space is building its own purpose-built rocket, targeting a first launch before the end of 2028, and designing its data centers directly into the rocket's second stage rather than treating the payload as a separate object. The metal mass of the upper stage becomes the satellite's heat sink — a piece of hardware that would otherwise be discarded becomes the orbital data center's thermal infrastructure. The resulting vehicle would be slightly more powerful than SpaceX's Falcon 9 workhorse, though still smaller than the Starship system still in development.
The Mechanism: How Orbital Compute Would Actually Work
The architecture breaks into three links, and each one has to work for the model to hold. First, power generation: large solar arrays on a 20-to-25-ton satellite convert sunlight into electricity continuously. Second, thermal management: roughly 800 GPUs packed into a single orbital unit reject heat into the vacuum of space, using the upper stage structure as a radiator. Third, the link to Earth: a kilowatt-class laser beams energy down to a ground receiver, or — in the data-center variant — the compute runs in orbit and results are transmitted down optically.
Reason-2, slated for launch in 2027, is designed to prove the second half of that chain. Sharing many of Reason-1's optical components, it will demonstrate optical data transmission from space to Earth at near-record rates and carry forward the thermal and power lessons from the first mission. The company's roadmap calls for a GPU cluster operating in orbit in early 2027 and orbital data centers by 2028, supported by its own rocket flying before the end of that year.
The development has been seeded with Bhatt's own money — roughly $10 million, by his account — and multi-year support from the Operational Energy Capability Improvement Fund, which paid for early personnel, custom components, and the laser and optical payloads on Reason-1. The team remains small — Bhatt has described it as a "pirate ship" startup — and includes Warren Lamont, who previously worked on propulsion at Blue Origin, and Tyler Grinnell, a former launch director at SpaceX. Chithra Perumal joined as chief financial officer in August, and John Sarkis as global head of sales.
The Economics: Launch Cost Is the Whole Game
The entire business case rests on one variable: the cost to put a kilogram into orbit. At 20,000 to 25,000 kilograms per satellite, each orbital data center requires a heavy-lift launch. Cowboy Space expects its own booster to eventually be reusable, which is the only path to the launch cadence a 20,000-satellite constellation would demand. But bringing the rocket in-house is also the hardest playbook in the industry: only a handful of companies in the West consistently launch commercial rockets, and several well-funded startups have spent years without delivering operational systems.
Even so, the math is unforgiving. A terrestrial hyperscale data center delivers power to GPUs through wires at grid prices measured in cents per kilowatt-hour. An orbital data center must amortize the cost of building, launching, and replacing hardware that operates in one of the harshest environments humans engineer for — radiation, thermal cycling, micrometeoroids — and then either beam power down or beam data down, each with conversion losses. For the model to work, the delivered cost per kilowatt-hour in orbit must fall far enough, fast enough, to offset the premium that AI operators will pay to escape terrestrial power constraints.
Bhatt argues the market is large enough for multiple winners. "The prize here, and the size of this market, is big enough that there's room for many players to succeed," he said. "I see the demand for AI getting more and more acute, and I see the options on Earth getting more and more limited." That pitch has attracted capital at a pace unusual for the sector: $275 million at a $2 billion valuation less than two years after founding, putting Cowboy Space in direct competition with SpaceX and Blue Origin, the most advanced and best-funded players in the market.
The Counter-Thesis: Why This Could Fail
The strongest case against Cowboy Space is not that the physics is wrong — space-based solar power and orbital computing have been studied for decades — but that the economics arrive too late. The company's own roadmap puts orbital data centers at 2028 at the earliest, and a full constellation years beyond that. In the same window, Earth-side alternatives are scaling fast: small modular reactors, grid upgrades, demand-response contracts, and data centers sited near stranded renewable generation all aim to relieve the same power bottleneck. Other orbital-compute schemes tell a similar story about timing: Google's Suncatcher project targets the mid-2030s, while players such as Starcloud are starting with edge-processing tasks for space sensors rather than full data centers. If terrestrial power capacity expands faster than AI demand tightens, the premium for orbital compute evaporates.
There is also execution risk on every link of the chain. Reason-1 must prove kilowatt-class laser beaming to a ground receiver; Reason-2 must prove high-rate optical downlink; the GPU cluster must run reliably in orbit; and the company must still design, certify, and mass-produce a rocket it has not yet flown — while competing for launch slots, spectrum rights, and orbital slots against incumbents that already fly regularly. The FCC filing for 20,000 satellites is a request, not a guarantee, and a constellation of that size carries debris and regulatory scrutiny that a young company has never navigated.
And the competitive threat cuts both ways. The very launch providers Cowboy Space depends on today — and plans to compete with tomorrow — could decide to offer orbital compute themselves, with deeper pockets and existing launch capacity. Bhatt's response is vertical integration: build the rocket, own the payload, control the cost curve. That is the same logic that has made SpaceX dominant, and it is the hardest playbook in the industry to execute on a first attempt.
Cyclical Crunch or Structural Shift?
The right read separates two claims that are often blended. The AI power crunch is structural: compute demand has been compounding faster than grid capacity for years, and data-center power is now a board-level constraint at the largest technology companies. That part is not mean-reverting on its own — new generation takes years to permit and build, and demand shows no sign of slowing.
The specific Cowboy Space solution, by contrast, is unproven and sits in the pilot phase. Space-based power beaming at kilowatt scale has never been demonstrated commercially, and no orbital data center has yet run a paying workload. The structural demand is real; whether orbital compute is the answer is a technical and economic question that the next 18 months of demonstrations will begin to settle. The two forces can coexist: a structural shortage of terrestrial power does not guarantee that space is the marginal supplier.
What to Watch
The near-term signals are concrete. First, whether Reason-1 successfully delivers kilowatt-class laser power to a ground receiver — the company has said it will share updates as the satellite reaches each milestone. Second, whether Reason-2 launches in 2027 and achieves near-record optical downlink rates. Third, whether the GPU cluster demonstrator operates in orbit in early 2027 as planned, and whether the NVIDIA Space-1 partnership produces flight hardware rather than a memorandum.
Beyond the demonstrations, the capital and regulatory path matters as much as the technology. A constellation of 20,000 satellites at $275 million per round implies several more billion-dollar fundraisings before revenue arrives. Watch for the next financing, for FCC authorization progress, and for evidence that the purpose-built rocket program is hitting design milestones rather than slipping toward the end-of-2028 target.
The falsifying signal is specific: if Reason-1 fails to beam kilowatt-class power to a ground receiver, or if the delivered cost per kilowatt-hour from orbit remains more than an order of magnitude above terrestrial grid prices after the 2027 demonstrations, the orbital-data-center thesis loses its economic foundation. Until then, Cowboy Space is the most capitalized attempt yet to move AI's energy problem off the ground — and the first one with hardware in orbit.
Data as of October 8, 2026. This article reports on a private company and does not constitute investment advice.
Explore more exclusive insights at nextfin.ai.
