Google Reveals AI Space Plan for Orbital Data Centers by 2027

In space, no one can hear you scheme. But on Earth, and if Google allows it, we can.

Google has unleashed Project Suncatcher, a bold plan to send AI data centers into orbit by early 2027. It reads like a direct answer to the power crunch squeezing the entire AI industry.

The project could reduce carbon emissions tenfold compared to traditional data centers. If that holds, it knocks down one of AIs ugliest environmental problems.

It’s an intriguing idea and ripe for millions of SF wordplay and cultural references. But in the interests of time, let’s stick to the hard facts.

Here comes the Sun

AI, the company says, is a “foundational technology that could help us tackle humanity’s greatest challenges.” But as demand for compute power grows exponentially, energy supply and environmental sustainability have become major bottlenecks. Project Suncatcher aims to address this by moving compute infrastructure off-planet — using solar energy directly from space, without atmospheric loss or land-based limitations.

In essence, Google is envisioning a network of satellites functioning as a distributed data center in orbit. These satellites would process AI workloads while communicating with Earth-based systems, potentially reducing the carbon footprint of large-scale training runs and freeing up terrestrial resources.

Chasing “moonshots”

The concept builds on the company’s tradition of pursuing “moonshots” through its research divisions, such as self-driving cars and quantum computing. Project Suncatcher is the latest in this lineage, representing a long-term bet on a future where computation is no longer constrained by Earth’s physical infrastructure.

Google’s early research, detailed in a preprint paper released alongside the announcement, covers satellite constellation design, communications strategies, and the challenges of operating semiconductor hardware — like TPUs — under the intense radiation conditions of space. The initial findings also suggest that hardware modifications and redundancy mechanisms will be critical to maintaining reliability in orbit.

Partnering with Planet

To bring the idea closer to reality, Google has partnered with Planet, the satellite imaging company, to launch two prototype satellites by early 2027. These will serve as testbeds for the first-generation space-based TPUs, assessing how they perform in real orbital conditions. The prototypes will focus on hardware resilience, data transmission efficiency, and the feasibility of sustained AI workloads powered purely by solar energy.

If successful, the experiment could mark the beginning of a new era of computation — one that decouples data processing from terrestrial limits and opens the door to “off-world” data centers.

A tech odyssey

The potential implications of Project Suncatcher extend far beyond Google. Space-based compute could enable AI research at previously impossible scales, power planetary monitoring systems, or even support autonomous spacecraft navigation and asteroid mining operations. It also raises important regulatory and ethical questions about orbital resource use, data sovereignty, and energy distribution.

While the project remains in its infancy, Google’s announcement underscores a broader shift in thinking about where — and how — the next generation of AI infrastructure will exist.

Things to come? If Project Suncatcher succeeds, the boundary between Earth and the digital cloud may become more blurred than ever before.