Google Unveils Plan for Space-Based AI Hub

Google to create satellite system for solar energy in orbit.

Google plans to establish a satellite system in Earth’s orbit to harness solar energy for powering data centers, as detailed in the company’s blog.

The corporation views artificial intelligence as a transformative technology with the potential to change the world. However, unlocking its full potential requires significant computational power.

“The Sun is the primary energy source in our solar system, emitting energy 100 trillion times greater than the total production of all humanity. In the right orbit, a solar panel can be eight times more efficient than on Earth, producing energy almost continuously and reducing the need for batteries,” the firm asserts.

In the future, space could become the optimal location for scaling artificial intelligence, prompting Google to launch the ambitious research project Suncatcher. This initiative involves creating compact groups of solar-powered satellites equipped with the company’s tensor processing units (TPUs) and interconnected through optical communication channels.

The proposed system will operate in a sun-synchronous low Earth orbit, where devices can receive nearly continuous sunlight exposure and maximize energy collection.

Challenges

Google highlighted several challenges that must be overcome to realize the Suncatcher project.

Ensuring Inter-Satellite Communication Channels

Large-scale machine learning workloads require task distribution among numerous accelerators with high bandwidth and low latency.

To achieve performance comparable to terrestrial data centers, communication channels between satellites must support tens of terabits per second. This is feasible through dense multi-channel transceivers with wavelength division multiplexing and spatial multiplexing, according to Google experts.

Achieving the necessary bandwidth requires levels of received power thousands of times greater than typical for conventional long-range systems.

“Since input energy is inversely proportional to the square of the distance, we can overcome this issue by launching satellites in a very dense formation,” the blog states.

The team has already begun testing this approach. A laboratory prototype achieved data transmission speeds of 800 Gbps in each direction.

Satellite Management

High-bandwidth inter-satellite communication channels require very dense device placement.

Experts have developed models to analyze the orbital dynamics of such a cluster. They indicate that with satellites placed only a few hundred meters apart, minor maneuvers will be necessary to maintain stable positioning.

Radiation Resistance of TPUs

To ensure the resilience of ML accelerators, they must withstand the conditions of low Earth orbit. The company tested the Trillium chip, with promising results.

High-bandwidth memory subsystems began to show instability after accumulating an ionizing dose nearly three times the expected level.

Experts did not observe any significant failures.

Economic Viability

Previously, high launch costs were the main obstacle to creating large-scale space systems. Data analysis suggests that by the mid-2030s, launch prices could drop to less than $200 per kg.

At such rates, a space-based data center could become economically comparable to an equivalent terrestrial data center.

Initial Launches

Google emphasized that the primary computations do not contradict fundamental physical laws and do not face insurmountable economic barriers.

“However, significant engineering challenges remain, such as thermal regulation, high-bandwidth ground communication, and orbital system reliability,” the blog notes.

Google will begin addressing these issues with a training mission in partnership with Planet, planning to launch two prototype satellites by early 2027. The experiment will test the functionality of models and equipment in space.

Back in May 2025, China launched 12 satellites as part of a project to build a network of orbital supercomputers.