The Solution to the AI Energy Crisis Isn’t on Earth

As Spock would say, “Fascinating”.

__________

Dominick GiuffridaWhy Google, Bezos, and Musk all agree on the next infrastructure trade. The solution to the AI energy crisis isn’t on Earth.

For years, the idea of data centres in space was dismissed as sci-fi. In the last week, it became the new industry consensus.

It is rare to see the biggest names in tech align so perfectly on a single future infrastructure shift. The commentary is no longer about “if”; it is about “who” builds the stack first.

Consider what has hit the market in just the last few days:

Google: Sundar Pichai confirmed “Project Suncatcher,” aiming for TPU constellations by 2027. His rationale is simple: the sun emits “100 trillion times more energy” than humanity produces, and space is the only place to capture it without interruption.

Blue Origin: Jeff Bezos predicts gigawatt-scale data centres in orbit within 20 years, explicitly stating they will beat terrestrial costs because of 24/7 solar access.

Starcloud: While the giants plan, this Nvidia-backed startup just trained the first AI model (NanoGPT) in orbit on an H100 GPU.

SpaceX: Musk is pitching a future where Starship delivers 300GW of solar-powered AI satellites annually.

Why the sudden rush? It comes down to three pragmatic drivers that Earth-based centres cannot solve:

* Energy: Solar panels in orbit are 3x to 8x more productive than on Earth and run 24/7.
* Cooling: The vacuum of space provides free radiative cooling, solving the heat bottleneck that currently caps high-performance compute.
* Speed: Optical laser links in vacuum are faster than fibre on Earth, enabling low-latency global grids.

We are watching the decoupling of compute from the power grid. The next major infrastructure asset class isn’t land, it’s orbit.

Share your thoughts 🤔

See post on LinkedIn