The market has internalised that power is scarce. It has not yet priced that time itself is the constraint. As AI cycles compress and infrastructure absorbs delay unevenly, capital is fragmenting between assets built for variance and those that assume perfect delivery.
2025 was the year AI collided with Earth’s limits. Power queues stretched into years, cooling and land became binding constraints, and time-to-deployment overtook capital. This piece examines how those pressures pushed compute off-planet, and what capital should watch next.
Why orbital compute has crossed from theory into allocation reality. Capital has a long memory for failed ideas and a short tolerance for resolved uncertainty that remains mispriced. In infrastructure markets, the decisive moment is rarely commercial scale. It is the first real-world execution that collapses an entire category of
AI’s constraint is no longer algorithms or capital, but physical infrastructure. Power, cooling, land, and permitting are hard limits. This piece examines why orbital compute has crossed from theory into deployable reality, and what that means for the future of AI infrastructure.