Project Suncatcher: AI is going to Space!
Content Team Nov 18, 2025
Did you ever consider that your phone's apps might be powered by the satellites way, far up in space? Sounds like science-fiction you think? But the most renowned tech companies are in the process of developing it! The reason is easy: Artificial Intelligence (AI) requires lots of power. It also uses an enormous amount of power.
To keep AI expanding, we require more energy and better places to house those huge computers, the data centers. This is where the huge concept known as Project Suncatcher comes in. It's a bold plan to build solar-powered AI data centers in orbit around Earth. It's a moonshot idea that aims to solve one of the most pressing challenges that modern technology can solve, that is environmental and energy costs.
Why Go to Space for AI?
You may ask why you don't create another data center on Earth? We've done it for a number of years. But AI requires a lot of power to train and run its large models, such as those that make intelligent apps.
The main thing that space has is that it has the sun! If you place an orbiting satellite in a particular orbit, called a sun-synchronous orbit, it will receive sunlight all the time. On Earth even the most efficient solar panels are not effective at night. However, in space, the solar panels are as much as eight times more efficient than they are on Earth. You'll have unlimited and clean energy.
Project Suncatcher is built on this clever idea. It hopes to make use of the sun's energy continuously to generate electricity to power AI computations without occupying space, requiring plenty of water to cool or causing polluting Earth. It's a way of growing AI while being more eco-friendly to the earth.
The Space Brain TPUs
The computers in these data centers orbiting around won't be normal ones. They will be using custom-designed chips referred to as Tensor Processing Units (TPUs). These chips were developed by Google specifically to be used in machine learning as well as AI tasks.
The goal is to outfit constellations (groups) of smaller solar-powered satellites using these TPUs. The modular design--small interconnected pieces instead of one massive machine -- makes the system more adaptable. The satellites, when working together, will function as one massive, powerful distributed data center. also known as a "space data center."
"For a firsthand perspective on the culture of innovation that fuels Project Suncatcher, review the statement from CEO Sundar Pichai on X: Click Here
The Big Problems Being Solved
The construction of data centers on Earth isn't too difficult But putting one on the moon? This requires some extremely clever engineering. The people who are working on this, including the smart AI developers Sydney has - are tackling huge issues.
1. Talking to each other (Very Very, Super Fast)
For all the satellites to function as one big computer, they must communicate with one another at a lightning-fast speed. That's like, faster than you could imagine! The AI tasks must be distributed across multiple processors that have high-bandwidth and low-latency connections.
The Solution: Project Suncatcher plans to make use of laser beams, also known as free-space optical links to transmit information between satellites. Tests in the lab have shown that they can attain incredible speeds, including more than $1.6trillion per second! However, the signal strength decreases with distance, which means satellites must fly in a tight formation, often just a few meters away. Making sure that the formation stays stable is a major task!
2. Suffering from the Harsh Environment
Space is a tough environment. Things like temperature extremes and radiation can damage electronics.
The Solution: Google has been testing its TPUs to determine how durable they are. The most surprising thing? They are more resistant to radiation than believed! The chips resisted radiation doses three times more than they could have expected during five years of space mission. This provides the engineers with confidence in the chip.
3. Making it affordable
Launching objects into space is still very expensive, which is an obstacle in the way of.
The Solution: This section is an investment in the future. The project's analysis suggests launch costs are dropping rapidly. If launch costs remain at or less than $200/kg by the mid-2030s, then running the cost of an AI data center in space may be economically similar to the energy costs of a similar facility on Earth.
The Next Steps for This Space Dream
Project Suncatcher is in its research phase however, it's progressing quickly. The next step is a critical "learning mission."
Google intends to launch two satellite prototypes in the early 2027s, working with a different company. This mission is about testing. They will test whether optical links operate as quickly like they do in labs, and if the TPUs can cope with radiation effectively in orbit, and whether tightly formed flying could be handled easily. This will set the stage for a future in which AI's full potential can be unlocked by the continuous illumination of the Sun.
This research into space-based, long-term AI compute is fascinating. It demonstrates how big-thinkers are solving huge problems that will shape the future of our technology and the world at large.