Tech company Google is now going to build data centers in space. This project has been named ‘Suncatcher’. Google’s CEO, Sundar Pichai, shared this information by posting on X.
Under this project, Google will send satellites equipped with solar panels into space. This means sunlight will be used for electricity. Google’s latest AI chips will be fitted in these satellites. These are called Trillium TPUs. These chips are designed for AI tasks. In simple words, Google wants to make AI superfast by using the sun’s free energy in space to avoid power shortages or other issues on Earth. These satellites sent into space will connect using a technology called Free-Space Optical Links. This means they will share high-speed data wirelessly with the help of laser light. This will make AI’s computing power larger and faster.
What is Project Suncatcher , and how will it work Project Suncatcher is a research idea from Google. Under this, small satellites will be launched into Low Earth Orbit, specifically Sun-Synchronous Orbit (SSO), where they will always receive sunlight.
Each satellite will be equipped with solar panels and Google’s Trillium TPUs chips, which are designed for AI training. These satellites will connect to each other via optical links, providing speeds of terabits per second. Google stated that a cluster of 81 satellites will fly within a radius of just 1 kilometer, to facilitate easy data transfer. Continuous solar power in space will reduce the need for batteries. The company achieved 1.6 Tbps bidirectional speed in initial tests.
Meanwhile, this cluster of satellites flying 400 miles above will handle large Machine Learning (ML) workloads. This will reduce the strain on electricity, water, and land on Earth. Why in space, what’s the problem on Earth? Training AI models requires a lot of energy. On Earth, the problems of electricity, water, and space for data centers are increasing. Google’s Senior Director, Travis Beals, said that the Sun is the ultimate energy source of our solar system, which provides 100 trillion times more power than the world’s total electricity production. Solar panels in space will be 8 times more productive and will provide continuous power. This will also reduce the carbon footprint. Google believes that by 2030, if the launch cost of satellites becomes $200 (₹17,727) per kilogram, then the cost of a space data center will become equivalent to one on Earth. Technical challenges, protecting TPUs from radiation Radiation in space is very high, which damages chips. Google tested the Trillium TPU in a particle accelerator (67MeV proton beam). The result was good – the chip will tolerate radiation up to 15 krad(Si). However, High Bandwidth Memory (HBM) is sensitive. Satellites will have to fly close to each other for optical links to work. For this, Hill-Clohessy-Wiltshire equations and JAX models will be used.
Thermal management and ground communication are also major challenges. Travis Beals said that there are no physics or economic barriers to the core concepts; only engineering challenges remain. First test in 2027, partnership with planet Google will launch two prototype satellites in early 2027 with Planet Labs company. TPU hardware, optical links, and models will be tested in space. Gigawatt-scale constellations will also be built in the future. All its details have been provided in Google’s preprint paper. If successful, AI training will be in space If the project is successful, AI training will be from space. Large ML workloads will be easily handled. Resources on Earth will be saved and the environment will be protected. Space computing will become cheaper as launch costs decrease and solar efficiency increases. Experts believe that space data centers could become a reality by 2035.
The post Google to build AI data center in space:The company announced Project ‘Suncatcher’, and will launch two prototype satellites in 2027 appeared first on Tri-Cities India.
