Google Looks To Space As AI Datacenters Strain Earth’s Infrastructure

Project Suncatcher aims to exploit energy advances and falling rocket costs

Google is preparing to take its AI ambitions beyond Earth, announcing plans to build the world's first orbital datacenters powered entirely by solar energy.

Under the Project Suncatcher initiative [pdf], Google aims to create tightly packed constellations of solar-powered satellites orbiting about 400 miles above Earth.

The satellites would house the company's tensor processing units (TPUs) and communicate via high-speed free-space optical links.

The first two prototype satellites are scheduled for launch in early 2027. If successful, they could pave the way for a fleet of around 80 orbital datacenter satellites designed to meet skyrocketing AI demand while alleviating pressure on the planet's stretched power grids and water supplies.

"In the right orbit, a solar panel can be up to eight times more productive than on Earth and produce power nearly continuously, reducing the need for batteries," Google says.

The company believes that with rocket launch costs continue to fall, the economics of operating datacenters in space could become comparable to Earth-based facilities by the mid-2030s.

However, that vision faces formidable engineering hurdles.

On Earth, datacenters rely on fiber-optic interconnects capable of transmitting data at blistering speeds. In orbit, maintaining terabit-per-second communication between moving satellites presents a far greater challenge.

Early Earth-based tests have demonstrated bidirectional data transfers up to 1.6 terabits per second, and researchers believe this could scale in space if the satellites remain within about one kilometer of each other.

There are other challenges as well: thermal management, ground communications bandwidth, and long-term system reliability in harsh orbital environments remain open questions.

"Significant engineering challenges remain," Google cautions, describing its findings as "a first milestone towards a scalable space-based AI."

AI's Energy Appetite Strains Earth's Infrastructure

Google's push into space comes as the world faces a mounting AI energy crisis.

A new Turner & Townsend Datacentre Construction Cost Index (2025-2026) highlights severe constraints in power access and supply chains across the globe.

Of more than 300 datacenter projects in 20 countries, nearly half (48 percent) report power availability as their biggest construction delay.

In the U.S., some new datacenters face grid connection queues of up to seven years. The report warns that current infrastructure cannot keep pace with the massive power demands of AI training systems.

OpenAI's known projects alone (some of which are promised for next year, with construction yet to break ground) could require 55.2 gigawatts: enough electricity to power 44 million households. Across the U.S., UK and Europe, datacenters are increasingly competing with residential and industrial users for limited grid capacity.

"Power availability remains a critical barrier," said Paul Barry, Turner & Townsend's North America datacenter sector lead.

"AI datacenters are more advanced, and by extension, costlier. They come with greater power demands and modern cooling solutions. Developers and operators must adapt quickly," he said.

Even where power is available, costs are climbing. The report found that AI-optimized liquid-cooled facilities are now 7-10 percent more expensive to design and build than traditional air-cooled ones.

Meanwhile, 83 percent of industry professionals said local supply chains cannot meet demand for the sophisticated cooling systems AI workloads require.

To address these issues, the report recommends that operators consider on-site generation, energy storage, and grid-independent power, a recommendation that aligns neatly with Google's orbital ambitions.

This article originally appeared on MES Computing’s sister site Computing.