Google Tests Space-Powered AI With Solar Satellites

News

1. California’s Genspark Enters AI Unicorn Club with $200 Million Round

2. OpenAI Taps Intel’s Sachin Katti to Build Compute Backbone for AGI

3. Flipkart Brings Voice-Led Wholesale Ordering to WhatsApp with Sarvam AI

4. Varun Berry Steps Down at Britannia, Rakshit Hargave to Take Over

5. OpenAI Issues Warning on AI’s Acceleration From Inside the Fast Lane

6. Intel Eyes Deeper India Play in Chips, AI

Sundar Pichai, CEO of Google and Alphabet, acknowledged that, like any moonshot, the project will require solving many complex engineering challenges.

“Early research shows our Trillium-generation TPUs (tensor processing units, purpose-built for AI) survived without damage when tested in a particle accelerator to simulate low earth orbit levels of radiation. However, significant challenges still remain like thermal management and on-orbit system reliability,” Pichai said, adding that more testing and breakthroughs are needed as Google counts down to launching two prototype satellites with Planet by early 2027, calling this pilot launch the “next milestone of many.”

He expressed excitement about being part of the innovation happening in space! <blockquote class=”twitter-tweet”><p lang=”en” dir=”ltr”>Our TPUs are headed to space! <br><br>Inspired by our history of moonshots, from quantum computing to autonomous driving, Project Suncatcher is exploring how we could one day build scalable ML compute systems in space, harnessing more of the sun’s power (which emits more power than 100… <a href=”https://t.co/aQhukBAMDp”>pic.twitter.com/aQhukBAMDp</a></p>&mdash; Sundar Pichai (@sundarpichai) <a href=”https://twitter.com/sundarpichai/status/1985754323813605423?ref_src=twsrc%5Etfw”>November 4, 2025</a></blockquote> <script async src=”https://platform.twitter.com/widgets.js” charset=”utf-8″></script>

The next immediate step is to launch two small test satellites with the satellite company Planet by early 2027.

These prototypes will test how well Google’s specialized AI chips and communication technologies perform in space, particularly validating the use of laser links for distributing machine learning tasks. This mission will answer key questions before building large-scale AI satellite networks.

Google’s TPUs are custom AI accelerator chips designed to optimize the complex mathematical computations vital for machine learning and neural networks.

The constellation of solar-powered satellites will be linked by free-space optical connections (laser-based) to other satellites and to ground stations.

“Project Suncatcher” aims to leverage the superior efficiency of solar panels in low Earth orbit, where they can be up to eight times more productive than on the ground.

Google plans to place these satellites in a dawn–dusk sun-synchronous low Earth orbit, where satellites cross the equator near sunrise and sunset local solar times, ensuring their solar panels are almost always bathed in sunlight.

According to Google, this groundbreaking AI infrastructure offers “tremendous scalability” and significantly “minimizes impact” on terrestrial resources, as solar panels can operate nearly continuously and eliminate the need for massive battery storage.

Travis Beals, Senior Director, Paradigms of Intelligence at Google, said, “Artificial intelligence (AI) is a foundational technology that could reshape our world, driving new scientific discoveries and helping us tackle humanity’s greatest challenges. Now, we’re asking where we can go to unlock its fullest potential.”

One critical challenge is establishing high-bandwidth, low-latency connections to distribute massive machine learning workloads across many accelerators.

“Delivering performance comparable to terrestrial data centers requires links between satellites that support tens of terabits per second,” the blogpost said.

Google plans to utilize advanced laser communication technology, known as multi-channel dense wavelength-division multiplexing (DWDM) and spatial multiplexing, which enables multiple data streams to be sent simultaneously using different colors of light and several laser beams in parallel, thereby boosting communication capacity.

To make this ultra-fast communication effective, satellites must fly very close together, only a few hundred meters apart. Because signal strength diminishes sharply with distance, tight formation flying maintains strong signals, enabling large data volumes to be transmitted without loss, effectively “closing the link budget,” or balancing signal power to maintain the connection.

Google has already demonstrated this technology at a small scale, achieving data speeds of 800 Gbps each way (1.6 Tbps total) between a single pair of transceivers.

Flying satellites so close together requires precise control to counter gravity and atmospheric drag effects.

Google’s orbital dynamics models suggest that “with satellites positioned just hundreds of meters apart, we will likely only require modest station-keeping maneuvers to maintain stable constellations within our desired sun-synchronous orbit,” making networked formation flying both feasible and energy efficient.

Google tested its advanced TPU chip, Trillium, by exposing it to harsh proton radiation, similar to the conditions in space. The chips showed remarkable resilience, enduring high radiation levels without serious failures, indicating they are tougher than expected and suitable for satellite use.

Although space launches have traditionally been costly, Google’s research projects aim to lower launch prices to fall below $200 per kilogram by the mid-2030s. This would make space-based data centers economically competitive, comparable to the energy costs of running earthbound data centers.