Starcloud
Starcloud
One-line summary: Private orbital-data-center startup (CEO philip-johnston); flew the first Nvidia H100 in orbit (Starcloud 1, Nov 2025) and supplies the sharpest operator-grade numbers yet on the orbital-compute capex arbitrage — a ~$500/kg launch break-even and a radiator 100x cheaper per watt than the ISS's.
What it is
~20-engineer team in Redmond, WA (half ex-SpaceX, half ex-AWS/Azure), founded Jan 2024 out of a space-based-solar white paper: since 90-95% of beamed power is lost in transmission and the marginal terrestrial energy project powers data centers anyway, move the data center to the power instead. Raised ~$200M at >$1B valuation. All facts from 2026-07-01-podcast-moonshots-sonnet-5-drops-fable-5-will-return-fusion-s-first.
Key facts (from 2026-07-01-podcast-moonshots-sonnet-5-drops-fable-5-will-return-fusion-s-first, speaker philip-johnston)
- Break-even: space-based solar needs ~$50/kg launch cost; orbital data centers only ~$500/kg — "we came to a launch cost break even around $500 a kilo if we had a cheap way to get the data center space."
- Core IP is the radiator: "10 times less mass per watt of dissipation than the ISS radiator. And about 100 times less cost per watt" — a manufacturing problem, not new physics.
- Roadmap: Starcloud 2 (Jan 2027 booked, ~100x Starcloud 1 power, largest commercial deployable radiator, H100s + Nvidia Blackwell chip + AWS Outpost); Starcloud 3: 200 kW, 3 t, 50 per Starship → "about 10 megawatts of new compute capacity per Starship launch."
- Business model: energy/infrastructure landlord, not a cloud — agreement with Crusoe: "we have a box, and that box has power cooling and connectivity... you pay that kind of rental fee to us and you finance the chips" (Equinix-of-space).
- Phase 1 (launch still expensive): edge inference for DoD/Earth-observation constellations — e.g. processing SAR data on orbit and downlinking only the tank coordinates instead of 100 GB of raw imagery. Phase 2 (~3-4 yrs, Starship cadence): compete with terrestrial DCs on energy cost.
- Binding constraint is launch: "if you want to book anything for 2028 on Falcon 9, there's just nothing available. The government's just plonked down 20 launches." Booking on Relativity Space's first launch; SpaceX contract for Starlink plug-and-play laser terminals on its next 25 satellites.
- Orbit choice: dawn-dusk SSO; "the ideal altitude to fly is actually 1200 kilometers" (lowest with zero eclipse year-round) — at 600 km there is still a ~10% blackout window one month/year.
Why it matters to stock-market
Not tradeable (private), but it is the operator-evidence layer under rapid-reusability-to-orbital-compute-capex-arbitrage and terrestrial-power-flat-to-orbital-dc-arbitrage: a third independent seat (founder-operator, after investor Fox/Baker and operator Marshall) naming a launch-cost crossover, plus concrete unit economics (radiator cost, MW-per-Starship). Its launch-scarcity testimony reinforces SpaceX's toll-road position (elon-web-services-to-spacex-hyperscaler-rerate).
Caveat: Johnston is pitching his own book on every one of these numbers; the $500/kg break-even and radiator multiples are company-model claims with no third-party verification.