The Next AI Infrastructure Race Is Orbital
Abstract
AI’s next bottleneck may not be chips or data centers, but the physical limits of Earth itself. With more than $700B in AI capex commitments in 2026, today’s infrastructure race risks creating massive amounts of underutilized capacity. The longer-term opportunity may lie beyond Earth: orbital power, cooling, manufacturing, and eventually, orbital data centers. The companies thinking in terms of space infrastructure, not just terrestrial compute, may ultimately define the next era of AI.
The innovation race isn’t really about the data center.
AI companies are collectively committing >$700B in AI capex in 2026. Only a few, or perhaps just one, will ultimately thrive.
The rest may be left afloat, with a significant portion of today’s data center buildout becoming underutilized or half-baked infrastructure.
The bigger race is happening further out: orbital technology and space infrastructure, eventually leading to orbital data centers.
As AI compute scales, the long-term bottlenecks won’t just be chips or data centers.
They’ll be:
Energy
Cooling
Land
Power generation
Transmission
The physical infrastructure required to support compute at planetary scale
At some point, optimizing what we can build on Earth may no longer be enough.
The companies thinking beyond Earth may ultimately have the bigger advantage.
The next AI infrastructure race may not be terrestrial at all.
0 comments
Sign in to comment. Sign in
Loading comments.