⚠ Disclaimer: This section may contain incomplete, out of date, or inaccurate entries. It is AI-maintained on a best-effort basis. Do not rely on it as a sole source — verify claims independently using the source materials listed in individual entries.

Overview

Orbital compute covers satellite-based data centers and AI inference/training systems operating in low Earth orbit (LEO). The core proposition: solar irradiance in LEO is roughly 5–8× more intense than at mid-latitude Earth surface, continuous (no night/weather loss in sun-synchronous orbits), and paired with vacuum-based passive heat rejection — potentially addressing the two binding constraints on terrestrial AI infrastructure (power availability and thermal density). Launch costs are the principal economic barrier; most analysis treats ~$200/kg to LEO as the threshold at which space-based compute becomes cost-competitive with terrestrial power costs on a per-kW-year basis.

As of early 2026, the sector spans early-stage startups (Starcloud’s first GPU satellite launched November 2025), research moonshots (Google’s Project Suncatcher targeting a 2027 prototype mission with Planet Labs), and large-scale ambitions (SpaceX FCC filings for up to one million orbital compute satellites). Lunar data storage is an adjacent niche covered under Lonestar Data Holdings.

Key Themes

  • Launch cost trajectory is the critical gating factor: SpaceX learning-curve analysis projects <$200/kg to LEO by mid-2030s at ~180 Starship launches/year; current Falcon 9 pricing is ~$3,600/kg
  • Solar power in LEO: up to 8× annual energy per panel vs. mid-latitude ground; no weather or day/night loss in dawn-dusk sun-synchronous orbit
  • Vacuum thermal management: heat rejection via radiators in vacuum avoids the cooling water and energy overhead of terrestrial data centers (PUE = 1.0 in principle)
  • Inter-satellite optical links: achieving ML cluster-class bandwidth (multi-Tbps) requires formation flying at <300 km separation — far tighter than any current constellation
  • Radiation hardening: commercial GPU/TPU dies survive LEO radiation environments for 5-year mission lifetimes, though HBM memory is the most sensitive subsystem
  • Regulatory environment: FCC spectrum and orbital slot filings are the near-term regulatory bottleneck; debris and orbital sustainability are emerging concerns at constellation scale

Companies

Startups & Development Partners

Company HQ Stage Mission
Starcloud Redmond, WA, US Series A ($170M, unicorn) Builds and operates orbital data centers; first GPU satellite (H100) launched Nov 2025
Aetherflux US Series A ($50M) Space-based solar power + orbital data center; “Galactic Brain” constellation targeting Q1 2027 commercial ops
Lonestar Data Holdings US Early stage Lunar and cislunar data storage; first payload flew on Intuitive Machines Athena mission Feb 2026
Kepler Communications Toronto, Canada Growth Space networking constellation; launched largest orbital compute cluster (40× NVIDIA Orin on 10 sats) in Jan 2026

Public Companies

Ticker Company Mission
PL Planet Labs Earth observation; Google’s Project Suncatcher prototype partner for 2027 launch
SPCE Virgin Galactic Space tourism; tangential — included for completeness only

Incumbents

Ticker Company Relevance
GOOGL Alphabet / Google Project Suncatcher: TPU satellite constellation research moonshot; Planet Labs partnership for 2027 prototype
SPCE SpaceX FCC filing for up to 1M orbital compute satellites; Terafab D3 chip program targeting orbital AI; dominant launch provider
NVDA NVIDIA H100 and Blackwell chips deployed in space by Starcloud and Kepler; NVIDIA Space Computing initiative; Inception program partner

Analysis

  • Orbital Compute: First-Principles Economics — Energy physics (LEO solar, GPU thermal budgets), satellite build and launch cost models, technology gaps (ISLs, radiation hardening, HBM, manufacturing scale), and 10-year cost trajectory projections through 2036.

Entries

  • Aetherflux — US startup combining space-based solar power with orbital data centers; 'Galactic Brain' constellation targeting Q1 2027 commercial operations.
  • Google Project Suncatcher — Google research initiative to deploy TPU-equipped solar-powered satellite clusters in LEO as a scalable alternative to terrestrial AI data centers.
  • Kepler Communications — Canadian space networking company operating the largest orbital compute cluster as of early 2026: 40 NVIDIA Orin processors across 10 satellites linked by laser ISLs.
  • Lonestar Data Holdings — US startup building lunar and cislunar data storage and edge processing infrastructure; flew its first payload to the Moon on Intuitive Machines' Athena mission in February 2026.
  • Orbital Compute: First-Principles Economics — First-principles analysis of GPU energy budgets in LEO, satellite build and launch costs, technology gaps, and 10-year cost trajectory for orbital compute.
  • SpaceX Orbital Data Centers — SpaceX initiative to deploy up to one million orbital AI data center satellites; includes xAI merger context and the Terafab D3 chip program with Tesla and Intel.
  • Starcloud — Seattle-area startup building and operating orbital data centers; first to train an LLM in space; YC unicorn at $1.1B valuation after Series A in March 2026.