Google Taps Space-Based Solar Power for Next-Gen Data Centers

Google has launched Project Suncatcher, a research initiative to develop AI data centers in space. The project aims to harness the power of solar energy to run artificial intelligence computations, reducing the need for carbon-based fuels and eliminating the environmental impact of traditional data centers.

The prototype satellite, which is currently orbiting the Earth, contains four Tensor Processing Units (TPUs) designed by Google specifically for machine learning tasks. These chips will run a version of the Gemma AI model to answer simple queries, but due to heat management constraints, they can only operate for 15 minutes at a time.

While there are challenges to overcome, including cooling and maintenance, Google sees this as an opportunity to scale data centers to meet future demand for AI. The company's senior director, Travis Beals, notes that 'the sun puts out almost all of the power in our solar system,' making space-based data centers a more sustainable option.

Beals acknowledges that there are significant costs associated with launching and maintaining satellites in orbit, but believes that the potential benefits outweigh the expenses. He estimates that it will take longer than five years for space-based data centers to become economically viable.