Sundar Pichai just made a wild prediction. Google's CEO believes we're about a decade away from having data centres in space. Not as experiments. As normal infrastructure.

The places that power your Gmail and YouTube could soon be floating above Earth.

Why put data centres in space?

The reason is actually straightforward. Data centres on Earth use massive amounts of electricity and water for cooling. A single large facility can consume as much power as a small city.

AI is making this worse. Training models like ChatGPT requires enormous computing power. Energy consumption is becoming unsustainable.

Space solves both problems. Solar power is constant and intense beyond Earth's atmosphere. No clouds, no nighttime, just endless sunlight. And cooling? The vacuum of space naturally dissipates heat without requiring water or massive air conditioning systems.

The physics work, but the challenges are huge, right?

Solar panels in space generate several times more energy than identical panels on Earth. Heat radiates away naturally in the vacuum. Launch costs have dropped dramatically thanks to reusable rockets.

But building data centres in space isn't simple. Physical repairs become incredibly expensive. Radiation damages electronics faster than on Earth. Data travelling between ground users and orbital facilities introduces delays.

Launch capacity remains limited despite improvements. Building at scale requires launching enormous amounts of hardware.

Who else is exploring this?

Google isn't alone. Microsoft tested underwater data centres, which share similar cooling advantages. Amazon Web Services operates ground stations for satellite data and has partnerships with space companies.

Various startups and research institutions have studied orbital data centres, though most work remains theoretical.

Why this matters for AI

Google's interest connects directly to AI ambitions. Training frontier models requires massive computing clusters running for extended periods. Energy costs and cooling requirements are becoming limiting factors.

If space infrastructure solves those problems, it removes major bottlenecks in AI scaling. Models could grow larger without worrying about power grid capacity or water availability.

The uncomfortable reality

Ten years seems optimistic given the technical, regulatory, and economic hurdles. But Google has a track record of ambitious infrastructure projects. When they commit resources, they follow through.

International space law wasn't written for commercial data centres. Questions about jurisdiction and data sovereignty remain unresolved. If Google processes European data in orbit, does GDPR apply?

Whether we're ready or not, space based computing infrastructure is apparently approaching. The question isn't whether it'll happen anymore. It's who gets there first, and what advantages they'll gain over everyone else still computing on the ground.