Starcloud’s Bet: Data Centers Move to Space When Launch Gets Cheap Enough

Published 2026-03-17 - Runtime about 44 min - Watch on YouTube

Philip Johnston’s core claim is blunt: once launch gets cheap enough, data centers move to space because Earth is running out of easy power sites, while orbit removes land, batteries, and most cooling infrastructure. The bet is not speculative sci-fi; it is a cost crossover driven by solar, vacuum, and Starship-scale launch rates.

What Matters

• Johnston says terrestrial data-center buildouts face 5-10 year permitting lead times for 100 MW projects in North America.
• His break-even model puts space at around $500 per kilo of launch cost, rising toward $1,000 per kilo as land constraints worsen.
• Space cuts three major terrestrial costs: permitted land, batteries and backup power, and solar overbuild; he says one square meter gets 8x the energy in orbit.
• Thermal design is the hard problem: about 70% of engineering time goes to heat, because vacuum only allows infrared radiation to dump waste heat.
• StarCloud’s first satellite used phase-change material; the next uses liquid cooling plus a large deployable radiator, with on-orbit validation planned later in 2026.
• The first commercial workloads are edge and cloud services for other spacecraft, especially military, government, and Earth-observation satellites.
• He says StarCloud 3 is designed for 200 kW each, 50 per Starship, or about 10 MW per launch, with sub-50 ms latency through Starlink-like links.
• Johnston’s end-state view is extreme: within 5-10 years, at least half of new compute capacity moves to space, and within 10 years capex could approach $1 trillion a year.