Sign in Subscribe
Cloud computing

Starcloud-1 Aims to Revolutionize Computing with Orbs-Based Data Centers

Strategic Insights

— 2 min read

Starcloud-1, a satellite designed to test the feasibility of orbital data processing, is set to launch aboard a SpaceX Falcon 9 with a powerful NVIDIA H100 GPU on board. This mission marks a significant step towards the vision of space-based cloud computing, with Starcloud's long-term goal of building entire data centers in space. The NVIDIA H100 GPU, currently one of the most powerful AI processors in existence, will process large volumes of data from Earth observation satellites, demonstrating the potential of decentralized, orbital computing.

Key Takeaways:

  • The NVIDIA H100 GPU, currently one of the most powerful AI processors in existence, is at the heart of the mission, designed for machine learning, inference, and fine-tuning tasks.
  • The chip aboard Starcloud-1 will process large volumes of data from Earth observation satellites, including high-resolution radar and optical imagery, and will run onboard AI models like Google's Gemma.
  • The mission aims to test real-world AI applications under harsh orbital conditions, processing live Earth observation data and running Google's Gemma LLM in space.
  • The Starcloud-1 mission is a strategic step towards an orbital data center ecosystem, with a long-term plan including building 5-gigawatt orbital data centers, equipped with sprawling solar arrays and radiative cooling panels measuring 4 kilometers across.
  • Launch costs will need to fall significantly for Starcloud's orbital computing vision to become viable, with the company closely watching SpaceX's Starship program, which has the potential to drastically lower the price per kilogram to orbit.
  • Starcloud's CEO, Philip Johnston, has predicted that in 10 years, nearly all new data centers will be built in outer space.

Statistics:

  • The NVIDIA H100 GPU has 544 GB of GDDR6 memory and a peak power consumption of 5.619 kW.
  • The chip is capable of 750 GFLOPS using a 64-bit floating point format.
  • The Starcloud-1 satellite will process large volumes of data from Earth observation satellites, including high-resolution radar and optical imagery.
  • The orbital data centers will be equipped with sprawling solar arrays and radiative cooling panels measuring 4 kilometers across.
  • The cost savings of moving compute operations to orbit will be 10x carbon-dioxide savings over the life of the data center compared with powering the data center terrestrially on Earth.

Sources:

  • [wp-block-quote is-layout-flow wp-block-quote-is-layout-flow]"Starcloud needs to be competitive with the type of workload you can run on an Earth-based data center, and the NVIDIA H100 GPU is by far the most performant in terms of training, fine-tuning and inference," said Philip Johnston, Starcloud's cofounder and CEO.[wp-block-quote is-layout-flow wp-block-quote-is-layout-flow]
  • [wp-block-quote is-layout-flow wp-block-quote-is-layout-flow]"In space, you get almost unlimited, low-cost renewable energy," Johnston explained in a (https://blogs.nvidia.com/blog/starcloud/) recent statement. [wp-block-quote is-layout-flow wp-block-quote-is-layout-flow]
  • [wp-block-quote is-layout-flow wp-block-quote-is-layout-flow]"In 10 years, nearly all new data centers will be being built in outer space," Johnston predicted.[wp-block-quote is-layout-flow wp-block-quote-is-layout-flow]
  • (https://space.com/61942-nvidia-starcloud-1-nasa-orbital-computer-research/) Starcloud-1 mission aims to improve communication between Earth and space-based intelligence systems through decentralized, orbital computing.

Read more