The cloud is leaving the ground. Starcloud just raised $250 million to build data centers in orbit, betting that terrestrial energy and land constraints will make low-earth real estate the next hot asset class. The timing is uncomfortable: TechCrunch notes that launch options are already drying up, meaning the capital race is outrunning the logistics. This is the structure of every infrastructure bubble, money moves faster than pipe.

Why Orbital Infrastructure Is the New Fiber Trench

The 1990s fiber overbuilding boom is the obvious comp. Capital flooded into undersea and terrestrial cables on the assumption that data demand would be infinite and immediate. It was, eventually, but the timeline was wrong and most of the early investors were wiped out. Orbital data centers carry the same asymmetric risk profile: the demand thesis (AI compute needs somewhere to go, energy is constrained on earth) is structurally sound, but the execution window is narrow and the failure modes are catastrophic and non-recoverable. You cannot fix a broken orbital data center with a truck roll. Meanwhile, Bloomberg reports that memory stocks are losing smart money momentum as AI infrastructure investment cools at the margins. Starcloud is raising into a window that may already be closing on the public market side.

The Certification Gap in Autonomous Infrastructure

There is a governance dimension here that nobody is talking about. A 2026 position paper in arXiv CS.AI by Riemer et al. argues that AI agents with chain-of-thought reasoning are predisposed to collusion when making market decisions, and that certification requirements are justified. Orbital data centers will almost certainly run autonomous AI workloads with minimal human oversight. The regulatory frameworks for what happens when those systems make consequential decisions from low-earth orbit do not yet exist. The infrastructure is getting funded. The governance is not.