Orbital Data Centers Edge Toward Reality

Hot take: this is a real infrastructure thesis, not a near-term replacement for terrestrial cloud. The physics are ugly, but the first wins are probably satellite edge processing and inference, not giant orbital training clusters.

• –Space power and heat rejection are the core appeal: continuous solar input and passive radiative cooling solve two of the nastiest data-center constraints.
• –The category is moving beyond pure hype because there are now actual funded prototypes, launches, and radiation-hard hardware tests.
• –General-purpose training in orbit still looks far off; bandwidth, maintenance, and launch economics make distributed inference much more plausible.
• –For developers, the practical near-term use case is compute tied to space assets, defense, or remote sensing, not a drop-in replacement for AWS or Azure.
• –If launch costs and on-orbit servicing don’t keep improving, this stays a niche, high-cost specialty market.