Google has officially launched its moonshot project, dubbed “Project Suncatcher,” which aims to move its computing power into space. The project’s goal is to build a scalable AI infrastructure powered by solar energy in space, addressing the increasing energy consumption of AI systems. The idea is to access solar energy in space, which is eight times more efficient than on Earth, and provides a 24/7 uninterrupted power supply. This approach also eliminates the need for land and water consumption, making it a more sustainable option.
The project is driven by the growing demand for data centers, which is expected to exceed the speed of planning new power-generation capabilities. The power consumption of global data infrastructure is projected to be equivalent to that of the entire country of Japan by 2030. Google’s solution is to launch a constellation of satellites powered by solar energy and equipped with its self-developed TPU chips, building an “orbital AI data center” in space.
However, the project faces several challenges, including the need for a “local area network” in space, protection from cosmic radiation, and data transmission back to Earth. Google plans to address these challenges through formation flight and laser communication, radiation-hardened chips, and collaboration with Planet to launch prototype satellites.
The cost of launching satellites into space is a significant obstacle, but Google believes that with the reduction in launch costs, space will become a more viable option for building data centers. The company has adopted SpaceX’s learning-curve assumption, which predicts that the unit launch cost will decrease by 20% every time the total launch mass doubles. SpaceX’s Starship is expected to reduce the launch cost to $60/kg, making space a more competitive option.
Other companies, such as Starcloud, are also working on building data centers in space. Starcloud’s mission is to provide real-time data processing on orbit, and its CEO believes that the cost reduction brought by SpaceX’s Starship is crucial to achieving this goal. NVIDIA’s H100 GPU has also been sent into space for the first time, demonstrating the potential for powerful computing in space.
The development of space-based data centers has the potential to revolutionize the way we process and store data, providing a more sustainable and efficient solution for the growing demands of AI and other applications. With companies like Google, SpaceX, and Starcloud leading the charge, the future of space-based computing looks promising.