Jeff Bezos: Are Data Centres in Space the Next AI Frontier?

Jeff Bezos believes space may hold the answer to the AI industry’s spiralling energy needs.

Speaking at Italian Tech Week in Turin, the Amazon founder and Executive Chair outlines how solar-powered data centres in orbit could surpass the performance of Earth-based facilities and meet rising demand from artificial intelligence and cloud computing.

“One of the things that’s going to happen next – is we’re going to start building these giant gigawatt data centres in space,” says Jeff.

A gigawatt is equal to one billion watts of power – roughly the output of a large nuclear power plant. At that scale, Jeff argues, orbiting facilities could provide the kind of continuous power that energy-hungry AI clusters now require.

This view comes as the AI industry pushes data centres to expand faster than electricity and water resources on Earth can accommodate.

Why AI infrastructure might move off-planet

In a discussion with John Elkann, who chairs Ferrari and Stellantis, Jeff sets out a vision where gigawatt-scale data centres orbit the planet, using uninterrupted solar power to operate AI workloads around the clock.

“These giant training clusters, those will be better built in space, because we have solar power there, 24/7. There are no clouds and no rain, no weather,” he says.

Training clusters refer to large networks of interconnected servers used to train AI models. These require immense computing power and demand steady, high-voltage energy, something hard to guarantee using terrestrial grids or intermittent renewable sources.

Solar generation in orbit provides a constant energy source, unlike ground-based solar farms which depend on weather conditions and daylight. In time, this could reduce the operating cost of data centres and offset the heavy resource usage of Earth-based sites.

“We will be able to beat the cost of terrestrial data centres in space in the next couple of decades,” Jeff says.

Space infrastructure already supports Earth operations

For Jeff, this move is not speculative. He sees it as a continuation of how space already benefits daily life on Earth through satellite systems.

“It already has happened with weather satellites. It has already happened with communication satellites. The next step is going to be data centres and then other kinds of manufacturing,” he says.

Weather satellites collect atmospheric data that enables accurate forecasts, while communications satellites make global telecommunications and internet access possible. Jeff believes data centres are next in line to follow this orbital trend.

However, major challenges still stand in the way. Maintenance is a primary concern. When equipment fails in space, sending engineers is not an option. Hardware upgrades require space launches, which still carry high costs and technical risks despite reusable rocket technology developed by companies like SpaceX and Blue Origin, which Jeff also founded.

Even with reusable rockets, failures during launch remain a threat, potentially destroying highly valuable components and delaying deployment. These are not minor setbacks for hyperscale operations.

Still, the argument remains that the growing energy footprint of AI – and the limits of Earth-based infrastructure – leave few alternatives.

“It’s hard to know exactly when, it’s 10 plus years – and I bet it’s not more than 20 years,” Jeff says.

With AI workloads only set to increase, and the environmental and economic costs of data centres continuing to climb, Jeff’s prediction points to a future where hyperscale infrastructure moves off-planet to find the resources it can no longer depend on here.