What happens when demand for AI compute meets the harsh reality that there are only so many rockets?
Cowboy Space just raised $275 million to find out. The company, led by Robinhood co-founder Baiju Bhatt, is making a wild bet: the real constraint on orbital data centers isn't technology or power—it's launch capacity.
This isn't another "let's beam solar power from space" pipe dream. Cowboy pivoted hard from their original Aetherflux identity after realizing something crucial about the AI boom.
<> "We're standing up our own rocket program" to bypass mid-2030s delays, Bhatt told TechCrunch. The timeline crunch is real—competitors like Google's Suncatcher are "targeting the mid-2030s" while Cowboy wants first launches by end of 2028./>
The numbers tell the story. Launch costs hit $50M+ per Falcon 9 flight with 2-3 year backlogs. Meanwhile, AI compute demand is growing at 40%+ CAGR through 2030. Something's gotta give.
Cowboy's answer? Build rockets where the upper stage doubles as a 1-megawatt orbital data center. No separate payload delivery. No waiting for someone else's launch manifest. Just integrated launch-plus-compute from day one.
The Vertical Integration Play
This feels familiar. Remember when hyperscalers got tired of buying servers and started designing their own chips? Now we're seeing the same pattern in space.
The technical specs matter here:
- 1MW per orbital unit (massive by space standards)
- 24/7 solar power (vs. Earth's 20-30% duty cycle)
- Radiation-cooled in vacuum (bye bye, cooling bills)
- Sub-$1K per kg launch costs if they nail the economics
Key hires signal serious intent: former Blue Origin propulsion engineer Warren Lamont and ex-SpaceX launch director Tyler Grinnell. They're not just buying engines off the shelf—they're building their own from scratch.
The funding round tells its own story. Index Ventures led, with backing from a16z (AI-focused), Bill Gates' Breakthrough Energy (climate tech), and SAIC (defense contractor). That's a coalition betting on AI-space convergence.
Hot Take: This Is Brilliant Timing
Everyone's missing the real play here. While terrestrial data centers hit energy grid limits and cooling nightmares, orbital compute sidesteps both problems entirely.
Unlimited solar power. Zero cooling infrastructure. No NIMBY permitting hell.
But the launch bottleneck was killing the economics. SpaceX prioritizes Starlink. Blue Origin keeps slipping timelines. Russian and Chinese options are geopolitically toxic.
Cowboy's vertical integration could crack this wide open. Their $2B post-money valuation assumes they can execute, but the market opportunity is massive—we're talking about a $500B+ AI compute market with orbital niches barely tapped.
The Execution Risk
Here's where it gets spicy. Building rocket engines is the hardest part of aerospace. SpaceX took 10+ years to nail Merlin engines. Blue Origin's BE-4 engines slipped repeatedly.
Cowboy wants to demo satellite power beaming in late 2026, then full rocket launches by end of 2028. That's aggressive bordering on delusional.
But maybe that's exactly what this market needs. Incremental thinking won't solve launch capacity constraints. Sometimes you need Robinhood-level audacity to break through entrenched bottlenecks.
The real question isn't whether orbital data centers make sense—they obviously do. It's whether a 2024-founded startup can execute rocket development faster than Blue Origin managed with infinite Bezos money.
My bet? They'll hit delays, but the vertical integration thesis is sound. And in a world where AI training runs cost $50M+, orbital compute starts looking less crazy and more inevitable.
