Google has introduced a plan for a space AI hub.

# Google has presented a plan for a space AI hub

Google Corporation will create a satellite system in low Earth orbit to harness solar energy and power data centers. This is mentioned in the company's blog.

Artificial intelligence is a fundamental technology that has the potential to change the world, according to the corporation. However, a lot of power is needed to unlock its potential.

“The Sun is the main source of energy in our solar system, emitting it 100 trillion times more compared to the total production of all humanity. On the right orbit, a solar panel can be eight times more efficient than on Earth and produce energy almost continuously, reducing the need for batteries,” claims the company.

In the future, space may become the best place for scaling artificial intelligence, which is why Google launched the “ambitious research project” Suncatcher. It involves the creation of compact groups of solar-powered satellites equipped with tensor processors (TPU) from the company and connected to each other via optical communication channels.

The proposed system will operate in a constantly illuminated zone of low Earth orbit, synchronized with the Sun. There, devices will be able to receive almost continuous exposure to sunlight and maximize energy collection.

Problems

To create the Suncatcher project, several difficulties must be overcome, Google emphasized.

Ensuring inter-satellite communication channels

Large-scale machine learning workloads require task distribution among numerous high-throughput, low-latency accelerators.

To achieve performance comparable to terrestrial data centers, communication channels between satellites capable of supporting tens of terabits per second are necessary. This is possible with multi-channel dense transceivers using wavelength multiplexing and spatial multiplexing, according to experts at Google.

To achieve the required throughput levels, the accepted power levels must be thousands of times greater than those typical for conventional long-range systems.

“Since the incoming 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. The sample in the laboratory provided data transmission at a rate of 800 Gbps in each direction.

Managing satellites

Inter-satellite communication channels with high bandwidth require very dense placement of devices.

Experts have developed models for analyzing the orbital dynamics of a similar cluster. They show that when satellites are placed just a few hundred meters apart, only small maneuvers will be needed to maintain a stable position.

Radiation Resistance of TPU

For ML accelerators to be resilient, they must withstand conditions in low Earth orbit. The company tested the Trillium chip, and the results were promising.

High-bandwidth memory subsystems began to exhibit instability after the accumulated ionizing dose, which is nearly three times the expected amount.

Experts did not detect any serious failures.

Economic feasibility

Previously, high launch costs were a major barrier to the creation of large-scale space systems. Data analysis has shown that by the mid-2030s, launch prices could drop to less than $200 per kg.

With such figures, the space data processing center could economically become comparable to an equivalent ground data center.

First Launches

Google emphasized that the main calculations do not contradict the fundamental laws of physics and do not face insurmountable economic barriers.

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

Google will start solving problems with a training mission in partnership with Planet, during which the launch of two prototype satellites is planned by the beginning of 2027. The experiment will test the functionality of models and equipment in space.

Recall that in May 2025, China sent 12 satellites into space as part of a project to build a network of orbital supercomputers.